@@aoeu256 and where does that money go? Cause it sure as hell ain't going to schools or serving the public good in anyway. Hell of a lot of corruption at local government level.
Me being a new viewer of BOTH is really enlightening. I went from knowing nothing about regenerative ai to learning more about "extended reality". And I'm in the small community that has been using IMVU, Secondlife, Sims throughtout my whole tween to adult years and slowly realizing their expansions on complete human extended interactions via online. Its wild just how fast technology has come. But its also been very slow for a long time. AI is just a scifi fantasticle escapade. Another use of 0's and 1's . Those who freak out about it also freaked out about who virtual chat sites would take over the minds of the youth. But here I am, an adult after using new technology and nothing beats real life experiences.
I honestly believe Adam fails to understand the real potential and potential harm that artificial intelligence represents human intelligence is not as unique as he wants to make it out to be, the reality is that human creativity is nothing more than the ability to blend concepts and ideas, and in that, LLMs are incredibly powerful and we are only scratching the surface. I think your whole concept of skepticism and discovering the truth in everyday deception is very valid and necessary, BUT I think you are really losing sight of the kind of paradigms that LLMs represent, I really think you underestimate the potential existential risks and it is annoying and irresponsible for you to indirectly attack the voices that have been raised to warn about it. you treat people like Eliezer yudkowsky and the like as doomsayers, who are motivated by some kind of financial gain.
@@cyndaguy fair, but this is pretty nitpicky. Call it a "perspective" or whatever. Every discussion on these topics has to mention 0% interest rates and Dan did it because he's bringing some bigger picture thinking.
Agreed. This interview in particular was a real treat. I've not previously encountered Adam's work but he seems to understand the subject matter very well.
"They say it's AI and it turns out the AI is a sweatshop" That's sort of what the Amazon Mechanical Turk program is tbh. Like no, AI didn't read your product receipt, I did, for two cents.
Really? Reading a receipt is something even AI from the 90s should've been really good at. It's not a very difficult task, especially because a computer printed the receipt.
@alexc4924 Actually, I can tell you from experience that you are right, but it doesn't matter. I did a project when I was unemployed to try and track what I was buying. You can get the text of the receipt pretty easily, but, receipts vary massively from vendor to vendor and within each vendor the text is highly abbreviated, and it is not always obvious what the abbreviation means. Also, sometimes, you actually can't even get the text off the receipt. Remember that these things start off as pristine OCRable text, but then they get crumpled up and shoved in a pocket, get water damaged by sweat, and then the friction/heat sensitive paper gets to fade for three weeks before you remember to scan it.
@@thewhitefalcon8539 Not really, it was possible from 2011 onwards, as there was hardware restraints, and Machine Learning plus newer hardware made reading receipts at scale easy. Ai, there is no such thing.
I preferred his makeup tutorial era. Cateyes _on point!_ Joking ofc, but all his stuff is great. The film concepts, as noted, & his film reviews that go deep but often hilariously cut deep (notably _The Snowman,_ his _50 Shades_ trilogy, _Su[¡c]¡de Squad_ - also not nitpicky nor cheap; he's clearly knowledgeable & passionate about film, but his approach is done in a way that's intelligent but avoids being tedious)... Also niche/cult-classic film history stuff like "Bakshi's LOTR" and his really cool "bigger picture" type essays like the one centred around "Contagion", released during the height of 'that plague' (which still hasn't ended, just a reminder) and the one about Jamie Oliver's weird quest against nuggies "to make school lunches healthy" (it's about more than just "kids being fed healthy food", much more). And I'm always pleased to hear Dan's voice doing a quote or something on a fellow RUclipsr's video. Anyway.
Adam Conover & Dan Olson, the crossover I definitely wanted to see! Honestly, I saw "Dan Olson" in the title & thought, "It couldn't be...? I'm not that lucky. Is it?" Twas! What a legend! Been watching his channel *Folding Ideas* for years. They're really exceptional.
As a User Experience Designer and Researcher who has a background in Anthropology, Adam and Dan are spot on in pinpointing what's wrong with tech thinking when it comes to people using technology as tools/tool use. I'm thrilled that this is *finally* being discussed!!
Yep. They perfectly describe the pattern playing out over and over. I've felt absolutely insane watching it constantly repeat without any serious pushback!
As a UI/UX Engineer, I wish all Designers had Anthropology degrees. Most of the problems tech creates is due to the people designing the systems having low social intelligence.
@@thatadamwhitleyThey should at the very least be encouraged to go to social events regularly. Because lets be honest, these guys spend hours in the gaming chair cut off from the world every day.
The more I learn about the modern video game/“web 3”/metaverse tech stuff, the more obsessed I am with the connections to Roblox. Roblox is almost EXACTLY what all these people are trying to create - it’s a platform with an extremely wide reach like Fortnite, it’s a vertically-integrated content host/provider like social media, it has a speculative NFT market with items that can be carried through every world and game within the space. Roblox is also huge, and preying directly on children. Game development via child labor, wage theft, gambling on NFTs, 3rd party gambling websites based on the NFTs, all packaged and sold in the same way as a Fortnite battle pass. It’s such a huge blind spot for industry experts and I’m blown away by how much they’re seeming to influence and how much they’re getting away with.
@@chase-warwick i'll have to direct you to People Make Games's documentary on the child labour stuff. like yes it is like what you say in a fashion, at the same time it's telling literal 10 year olds they can have 100s of dollars if they crunch all their free time away, not to mention the real messed up creator group hierarchies that can abuse adult content creators and there's no vetting system for minors in these groups. the edge cases are dangerous
@@jacksim5759@jacksim5759 That's true, but the comment has a point: for all it's flaws, Roblox isn't as bad as web3 because it's confined to virtuality. Those kids aren't actually living day to day on scrip like they did in mining towns. Web 3 is comparatively more dangerous.
That's fair. I really, really dislike Roblox, but it's nowhere near as bad as cryptocurrencies. The very premise of cryptocurrencies and anything based on them was fundamentally broken right from the start. They were never going to have net legitimate utility.
the proof of this fabricated hype is in how Altman keeps saying how "AI will turn the world to ashes," and "we need to regulate AIs cause they are too awesome", but then when Europe starts thinking of applying copyright laws to AI, or regulate machine learning, he goes all like "fine, no chatgpt to you, loser". like if it is dangerous as he says, then he should be happy for regulations.
of course, and the reason they're spinning that "ai will destroy us" thing as if it's the terminator, and not the obvious "we're going to steal everyone's information so we can profit off of it and soak up all the revenue from various industries" is the misdirection, and by getting up in front of this shows his intent is regulatory capture. They are spinning this as if it's skynet, when it's an advanced autocomplete scrapping data. They want to be given the means to control things themselves which is the entire point.
You are severly mistaken in your assesment, You think too much of yourself and of your race (mine too, i am a human), we are not even close to being smart or intelligent, and there are things that can and will be smarter and more powerful than us and no self-imposed virtual and ethereal grandness will save us. You wont even have time to realize how wrong and overconfident you were.
@@user-yl7kl7sl1gThe issue you're describing isn't "LLMs being more intelligent than human beings," it's "humans using LLMs' vast information repository to learn things they can use to harm others."
@@user-yl7kl7sl1g the thing is language models aren't intelligent. They aren't even true AIs. Like a lot of "sentience" people see in AI is their mirror neurons at work. The only way they could be more "intelligent" (which translates to more time-efficient and less expendable) is through Data, as well as more complex hardware and more complex algorithm. These AIs are made and run by people, they aren't independent. The only real danger would be one company having our collective data stored to change and modify. What Altman fears is competition, a lawsuit, and loss of money, not the end of the human race like Matrix. Like, the only "AI apocalypse" I can imagine is if we completely substituted human work with AI. Generally AI needs data to be on the same level of humans. Constant data, since we humans are constantly evolving and creating stuff. but if nobody creates nothing in the future and it is all AI generated what we would create is synthetic data: data created by an AI. This synthetic data theoretically can be taught to the machine, but it is proven to create this "cannibalisation effect" where the language model decreases in quality because it copies the biases and errors AI-generated content possess. So it would eventually collapse to the point of being unusable. If our entire systems are all AI based, this eventual cannibalisation could mean loss of not only of modern infrastructures (all AI based in this hypothetical future), but of our entire digitally recorded history. And if that wasn't enough, we must imagine these future humans to have had relied on AI for their entire lifetime. Therefore we would have completely underskilled, mentally underdeveloped humans thrown to square 1 of human civilization.
He only wants regulation for cuckoo fantasy scenarios like AIs becoming so smart they transcend reality, not for actual thinhs his company might do in reality.
Could you imagine paying hundreds of thousands of dollars for an NFT sword and then you bring it into Breath of the Wild and you hit one dude and it shatters?
@@TheManinBlack9054 I think its use outside of the scientific sphere is largely a net negative for humanity and society to be honest, however it has some important uses in data analysis and simulating protein folding in the sciences and medicine.
To be fair, it's not hard to ruin these amazingly tone-deaf ideas that they mention. As soon as you start to read into it and understand the functioning behind all this recent techno-crap, you realize it's either much, MUCH less than what they sell, or simply a rehash of something that already existed way before.
What I feel the Metaverse thing misunderstand the most and everyone should have learnt by now in the post-pandemic world is that people do want to go outside and do things with their friends and families.
Metaverse is the real life story of Sword Art Online. Bunch of gamers get trapped in their favorite game world. Spend the next two years fighting to get the heck out of the game world XD
@@Bustermachine Hell, SAO even showed you why being permanently stuck in a virtual world is not physically viable or good for you: your IRL body becomes malnourished and weak, because virtual foods, no matter how many signals they send to your brain, don't provide nutrition.
Oh, I'm very happy to stay inside! But if I'm going to be doing something with graphics I want it to LOOK PRETTY (which the Metaverse doesn't), be able to accept ANY KEYBOARD (Lost Ark, I'm glaring at you), and, you know, other things which apparently the teams headed by the richest dudes in the world have apparently never heard of.
Jesus, I remember when Dan-O was "just" my boy down in Calgary having Socratic discussions with a milk-carton-turned-hand-puppet about the narrative possibilities of a Dwarf Fortress playthrough. I feel so weirdly proud and happy seeing him here now. (parasocial relationships be like that)
I think Dan is on of the most impressive transformations I've seen on RUclips. Starting as just another faceless avatars taking apart games and movies, through being a knowledgable movie buff, now creating those elaborate, multi-hour videos that help millions of people understand very complex subjects. He could have stopped so many times, and just keep making content in the same niche. But he keeps being hungry and ambitious. Such an incredible dude.
@@andrzejsugier Dan Olson is just one of THE GUYS who helps people understand whatever he’s discussing His The Wall video is one of my favorite cause he actually engages with the metaphorical medium with good faith and critiquing what’s a lil juvenile or poorly explored
I've been a teacher for 30 years and every few years we get trained on another transformative method or technology that will completely change our teaching. Spoiler: they never do, though occasionally we get a nice tool. Effective teaching is about relating to students.
Enterprise tools, teaching tools, and tools for government employees. They all have the same flaws, the person buying them isn't the person using them, so they're optimized for the best "on paper" abilities, rather than caring about the people who are actually going to use them. I don't envy you. When I hear what teachers have to put up with, it makes me so sad.
When Adam talks about AI NPCs I can't help but think of the "games" my 5 yo invents. The rules are constantly changing and updated, the goal is unclear, and it ends up going nowhere, without anyone knowing who won.
The bigger issue with AI NPCs, IMO, is that you don't typically want too much creativity from a game AI. You want something that behaves in 'clever' ways relative to the bounds of the intended gameplay experience. Think about how much the AI, in say, Halo, of Half Life, or FEAR, was praised. But in reality what made these AI seem so convincing was that they had well telegraphed, reliable, behaviors. You may not know if a combine soldier was going to chuck a grenade through the window, but you leaned pretty quickly that it was a thing the Combine's AI could do. You knew when you stickied an Elite that he was going to get angry and then charge at you in a Kamikaze rush. Etc . . . To a degree, you want consistent behavior, you just want very complex and consistent behavior.
Like, anyone who talks about AI generative NPCs should be forced to watch a Let’s Play of Facades in order to show them exactly why that’s a terrible idea.
"Xbox, smoke break?" took no longer to say than "Xbox, pause game" (though what happened to just pressing the pause button?), but it sure was more fun. I don't trust "listening" devices, but it was fun to do at a friend-group's house full of Xboxes with the voice command stuff not disengaged.
@@kevinhengehold4387 After like 24-48 hours, as I recall, they shut it down cos it was already trained by trolls to say awful shit. Pretty hilarious, honestly, cos these tech giants sunk so much into it yet just forgot about 4chan?
A disheartening quirk of platforms who've taken measures to at least hide AI stuff that's flooded them that I've seen is that on pixiv, a japanese site for posting art, has the option to turn off the ability to see art tagged as AI. While you might think people would just not tag their AI stuff as such to get past the filter would be the problem, the bigger problem I've seen is that people are just straight up stealing people's actual art and tagging it as AI to minimize the chance they get caught.
The sensible thing would be to create a separate website for aggregating AI content, but of course for many people that would defeat the purpose. (of tricking people)
Literally the only thing Driveshare apps did that made me want to use them over taxis is just not having to make a phone call and knowing how much a ride would cost before you summon the car.
Exactly. Where I am from, the taxi business had a weird aversion against any type of renewal that Uber seemed like a God send just because it was easy to use and you could minimize scams because it was all monitored on your phone.
Yeah, that's the thing: Uber actually did have a product, it was just a quite small app. Calling cabs was terrible, back in the day: you had to call and talk to someone, and then they told you "five to fifteen minutes," and then maybe a cab would show up, or maybe not. And maybe they could find you, and maybe not. And twenty minutes goes by, and you call back, and you get "five to fifteen minutes." Cab company+uber app would have been a legit product! A money-maker, even! Just, without the labor violations and exploitation, not the kind of money-printing scam machine they wanted.
There are some old stream highlights on his RUclips that sound like he scripted them because of how intelligently he's able to talk about some subjects off the cuff.
The physical reality of looking at a product in a store also lets you see things about that individual product that you'll never see in a Metaverse. Physical defects like scratches, holes, discolorations, loose joints, smells, everything. There is no way in hell that any Metaverse version of a store will upload full 5d (throw in some "d"s for smell and other stuff) scans of every single product that goes through their warehouse, tagged so that a user can pick out a hyper-specific product out of the stock. For everything where that specific ability to judge a specific instance of a product doesn't matter, web pages are fine.
Yep. At that point it's just cheaper to set up a real shop. Why recreate all that stuff digitally? The whole point of digital spaces is to not have specifics. To be interchangeable and non-specific. That's where the savings come in.
The thing is, the people who are coming up with these idiotic ideas don't _know_ what people do in physical stores because they themselves are so far removed from it. Do you really believe someone like Zuckerberg understands that when a person picks up a carton of eggs and opens it that they aren't simply going through the motions of a societal ritual but are checking for broken eggs?
@@IshtarNike I'm almost surprised that they aren't trying to gig-ify the whole endeavor by forcing doordash shoppers to wear a body cam connected to the user's VR headset as a cheap way to "virtualize" a whole store. Now that shopper can have the customer directly in their ear telling them where to go and what to pick up, so convenient! Wait, hang on, that was a joke Meta, don't actually
I love that in response to AI stealing art to train their algorithms, people invented Glaze to put art through basically a noise filter. Barely changes the original art, but introduces that noise into the datasets the AI trains on. Thereby corrupting the output, as this digital noise compounds to the point the creations are unusable. Effectively poisoning the well. It's fantastic, and I hope AI art programmers never figure out a way around it.
EverywherecI looked, the newest AI is openly described as a "pastiche machine" - I've no idea why any of you seem to think pointing this out breaks some kinda "official narrative", it doesn't lol.
@@alejandrotuazon4831 More precisely, AI doesn't suffer or benefit from the interplay between skill and psychology. For example, the effects depression, schizophrenia, or joy have on artwork. The mistake a lot of tech bros make is thinking art is all about skill and technique, but disregarding the heart, soul, and all the other messed up crap inside of us that makes art less about the picture we see and more about the story it tells us.
I remember the day FB changed to Meta. It was because of the whistle blower. I worked in FB ads at the time, their whole platform was down for 24 hours right after that hearing. They probably thought the whistle blowers evidence was too substantial. She primarily talked about how Meta uses engagement data to make people sad, and mostly teens. They want people to be sad and angry because it makes people buy things for their ad platform. It's disgusting. I know this because I remember watching the hearing live, and working in FB ads at the time.
@@fangirlfortheages5940 i think it was Zuckerbergs way to smoke and mirrors the whole situation. People seem to forget things when there's diversions over and over again. They changed the whole company direction when they announced their name as Meta. Many celebrities do the diversion after diversion tactic as well lol. I remember their servers were down all day - atleast 13 hours. I think they were wiping data the whole time to ensure they'll get away with it. The hearing was very convincing and I think they needed a reason to turn the servers off for a while. The whole meta thing was probably in the works for years and the hearing triggered a perfect time to do it. I could be wrong, but everything seemed to line up perfectly. The whistleblower must've been too right. It would make sense since facebook could barely keep top talent at their company. Watch the documentary The Social Dillema!
I legitimately laughed when they called Zuckerberg a loser for renaming his company meta and now nobody is talking about the metaverse. It's so true tho. A fat L for the robot fratboy
But overall they are all there to actually engage with you and the story and campaign, of course some people will want to test limits but everyone is on the same page...i.e. we are here to role play and have fun right? If someone comes into the game and starts questioning physics and asking why about everything and running around slapping all the villagers you are quickly going to have the in stocks by the city guard.
Lol it will just turn into the NPC's nervously looking around "hey guys I am really here to talk to you about the dark Mage of Azimoth, not the Minions movie..."
I’ve introduced multiple NPCs that I didn’t know would immediately be latched onto by my buddy, and both times they’ve been taken - one was a married lesbian woman and the other is betrothed to one of the party’s allies lol
Im a CS student with a background in ML and a 3d printing hobby, and hype is so real and theres constantly claims that fundamentally misunderstand the capabilities and practicalities of the technology
Whether your expertise is in ML or in the thing they're trying to shove ML into, it's really jarring to see how little the hype artists understand about either.
The whole impracticality of AI reminds me of how I was browsing someone's art blog on tumblr, went to their "about" page, saw a picture of their business card which had a QR code on it, and read, "scan this QR code with your phone to go to my personal website!". I was like... Are you kidding me? That is so convoluted. Just put a link to your website there. Not to mention, what is someone supposed to do if they're viewing that site on their phone?
@@Troisnyx still a pain in the ass nonetheless also i hate this idea so much like i put so much effort into not needing to use my phone for everything and honestly if i came across this i just like wouldnt
Well chat gpt is a quite useful tool but you can't trust it. You have to expect that it might tell you nonsensical stuff. I use it to do math stuff for me, so i don't have to waste my time on it, and also to pitch me the idea how to solve something when i get stuck. It usually gives the right direction, but can't properly solve stuff. The gpt3 is mostly useless, but gpt4 is actually an useful tool. Even for advanced math. It can do the easy stuff well, but for more advanced things, it usually takes few tries to get the right direction for something. Its a great tool for doing proofs, because it spits out kinda related ideas and that is often what you are missing. It can cut down the time to figure out the right approach massively. But still it can't do it fully atm. Maybe gpt5 will actually be able to do most of the stuff correctly. Lets hope
This video is great. I'm a developer myself, and recently I've been floating my resume around. Lots of companies won't even look at my resume unless I include something about GPT and other LLMs. I have used the systems in my work, but they really can't be used to write entire applications without a chaperone and they can't be used for anything that's not surface level. They don't understand performance, idioms, maintainable code, efficiency, etc. The reality is that these tools effectively have just supplanted search engines and other websites like stack overflow. They also help with some of the grunt work but that's about it. And forget writing something new with these systems, they can't really create anything new because they just regurgitate patterns based off of the data that they were trained on. There's no creativity or understanding of the topics that they cover. None of these models even understand what language even is in the first place. What's really troubling though, is the fact that these systems are effectively black boxes. In any other domain of programming, if you created an app that was a black box then you would be in a lot of trouble. If you can't have any kind of introspection into the workings of the system itself, then there's no way to verify that it's doing exactly what you think it's doing. For all you know the code could be firing missiles or hacking your system etc. In the case of these LLMs, not only is a functionality a black box, the concept itself is also a black box. Even the most intelligent researcher does not fully understand how these systems work completely. As far as their concerned we threw a bunch of linear algebra, probability and calculus together with a ton of data and somehow that produced a system that could replicate patterns.
I think it's a bit much to say the concept itself is a black box. After all, you just described the basic principle of what's going on and I'm incredulous that you can make something this complex by just throwing math together and seeing what sticks. It's very similar conceptually to certain fields of statistical linguistics. Not to mention implementing guard rails/directing the output would not be practical if the developers didn't have some idea of what the model was doing. The system isn't conceptually a black box. But rather, getting to the change of events behind any one output is a black box because it's so damned convoluted. And therein lies the danger that you end up with this pooping its pants because it suddenly thinks the Brothers Gracchi, the Heimlich Maneuver, Critical Race Theory, and National Economic Planning are all deeply interrelated and the solution to life the universe and everything is 42+Pi. Edit : Another issue, of course, is epistemology. Namely, as you mentioned, this is being regarded as a stand in for using a search engine. But ChatGPT is no less credulous than google if you feed it bad information.
@@Brandon82967 just because it can regurgitate information doesn't mean that it actually understands any of that information. There is a reason why it hallucinates so often. I don't care how much data they feed these current models, they will never be able to get rid of the hallucinations with the current generation of AI. There are many research papers on this subject, though I think the most important one is one that came out around the time that GPT started becoming very popular. I'm not sure if you're familiar with alphago, it's a neural network based on the same technology as these LLMs that was able to beat the world's foremost go player consistently over a series of matches. The paper highlights that the algorithm itself doesn't really understand what go even is on a fundamental level. Essentially, they found a strategy where you attempt to capture large sections of the board in a way that allows you to take all of the stones of alphago. The model seemingly ignores that you are winning until you finally captured all of its stones by which point the game is already over. One of the main rules of go is that if you surround a section of stones, then those stones are captured. What this means is that the researchers were able to find a strategy that used a very basic concept of the game to beat this algorithm because the neural network did not understand the game. The only thing that the neural network understands is the patterns of play that can arise from playing go, and that's where the important distinction comes from. If you try to use the same strategy against a human being, the human being will immediately counter it.
I'm listening to this conversation while prepping vegetables and saying out loud, "But the problem with that business model is -" and then Adam or Dan finishes the sentence for me! What an amazing conversation. Thank you and I'll see you in Baltimore.
Things are, in the strictest sense of the word, defined by their boundaries. (If you walked into a room where light was evenly and equally projected from all angles and directions, with no shadows or obscurants, you would, quite literally, not be able to see a single thing)
It’s not just constraints, even if you roleplay and say exactly what the npc wants, someone still has to make that mission, someone still has to make the “correct” parameters
Dan has stared into the Sun long enough to obtain wisdom but somehow kept his eyesight. I am cannot imagine how terrifying it is to know as much as him.
Everything he says sounds like obvious common sense but only after you've heard it, before that you couldn't even imagine. Like the whole "flat earthers are all into QAnon now" thing.
The notions discussed around 37:00 about Tech Billionaires desperately trying to build new Walled Gardens to entice us into (and then, brick up the walls behind us) is... Both resonating and terrifying.
I worked Uber/Lyft for 3 years. I loved the job itself; meet all kinds of people, play my own music and podcasts, I was one of the top 10 rated Uber drivers in my city, it was great. BUT, after taking out for gas, carwashes, etc, I was making $10/hr plus paltry, infrequent tips. I was also putting 50k miles a year on my car. The final straw was when I got sideswiped on the highway. His fault 100%, his insurance covered everything. But I was completely unemployed for two weeks while my car was being fixed. No safety net or job security is, imo, the biggest problem with all the gig economy work.
@@rakino4418 the fact that they were all blogposts and none referred to the available economists who did research into the matter or even the color papers lol. he mentions zero-knowledge proofs zero times lol
@@snowballeffect7812 If I clicked on a video about the interface between people and crypto and the creator started talking about zero knowledge proofs, I'd probably stop watching because it would be readily apparent that the video wasn't ever going to actually get to any sociological subject matter with any kind of depth (unless it was like 8 hours long). Not everything has to be a 3blue1brown math explainer; there's something called scope that good writers pay attention to.
@@tracyh5751 he talked about mining and minting and DAOs but somehow the absolutely amazing notion that you can prove you know information without revealing any information is too much for you? lol. If he's going to make claims, he needs to cite them. He didn't and he made false ones. His opinions are his own and i generally agree with him, but it was disappointing to see him not do any research and being wrong about what he was talking about. Misinformation is always bad.
Omg omg! I'm so proud of Dan, as a fellow Calgarian it's so cool to see him gain this kind of prowess among the likes of Adam and others. He deserves it! Great work both of you
Fellow Calgarian in the digital marketing industry and man everything they talk about is the nonsense I have to mentally filter out on a daily basis. Every anti-hype video I watch from one of these 2 is such an exercise in affirmation because I feel my cynicism is justified in so many ways. Sometimes I feel insane because all the rhetoric swirling around on social media is so counter to my gut instinct, but guys like Adam and Dan keep me feeling sane and valid.
Same here, it's awesome having someone as well spoken and observant and just generally engaging as him repping our town especially since a lot of Canadians play their hometowns very close to the chest.
The best justification I've seen for Zuck diving into Meta, was that he desperately wants to establish his own version of the Apple platform. Basically, create a platform (IOS and IOS-based hardware) where he can charge 3rd party developers a cut if they want their app to be included on the platform (the App Store). It's an enviable 'passive income' model, that many businessmen drool over.
I wish there was more discussion on non-creative jobs being overtaken by automation just as predicted. I'm a window cleaner by trade and over the past couple of years it's become very clear that we won't have a call center in the next 3 years. Not only will things be scheduled, invoices sent, and routes planned by the computer but will also have less drive time less drags in the system and more money on the books. I'm personally not a fan of a society willing to slash jobs without compensation outside of the labor market, merely because the incentive of monetary wealth is too damn high in today's world, but that's a whole other discussion. This isn't AI like the language model they've been discussing (for the most part) but multiple algorithms working together to produce better results than humans ever could
I want every job that can be done by a machine to be done by a machine - a good part of what makes me good at my job (word processing) is my ability to make the machine do the repetitive stuff, or boil down a series of steps into a single button-click. The problem with labor-saving technology is that the benefits don't accrue to labor, in the form of a shorter workday/week/life with better living conditions, but solely to capital, in the form of increased concentration of wealth and power in the hands of those who own the machines, and immiseration of everyone else.
@@dwc1964Hmmm, I agree the benefits will go disproportionately to those who own the capital, but to say it will go _solely_ to those who own the capital is perhaps a bit of a reach. It is possible for everyone to benefit at the same time that inequality gets worse. The rich get a lot richer, the poor see moderate gains. Ultimately, labour are _also_ consumers. People who currently work in call centres also need to buy food and pay for utilities and services (which probably cost more than they theoretically could with AI reducing inefficiencies). Some portion of those who work in call centres _will_ retrain as something else, another proportion will be retained doing something difficult-to-automate in the call centre, and might get a pay bump if their job is higher skilled. Some amount of the additional wealth generated by AI _will_ be taxed by governments and put back into public goods and welfare. Fundamentally, it's kinda hard to imagine a scenario where human society gets more efficient at delivering goods and services, and the average persons quality of life does _not_ improve.
@@merrymachiavelli2041 I think the argument is that the machines will take the work and there won't be enough demand of skilled labor and society won't adjust fast enough before we have unprecedented inequality, or worse a mass culling through poverty or civil unrest. You can look at most of the job market post Covid, nobody wanted to work minimum wage being a waitress or cashier when they had been getting more than that from the government in stimulus and yet businesses were starving for employees but refused to pay livable fair wages. You had a few businesses "experiment" by paying low skill jobs $15/hr and suddenly that business boomed. Shocker, we learned that both business CAN provide fair wages AND the government didn't implode by providing a basic income for everyone. I agree though, automation and general efficiency increases will reduce the price of everything; so everyone benefits. However, we have to leave behind some draconian concepts that contribute to inequality. The future is going to be more socialized or it's going to be a war from the poor. Looking at history, I don't have a lot of faith our economic systems will catch up or change fast enough.
@@merrymachiavelli2041while I'm not saying that this is exactly what's going to happen, what you're describing isn't really the trend we've seen. As more and more jobs get shipped overseas or automated, what we've seen is that "excess labor" (i.e. the people who lost their jobs) have been increasingly pushed into the prison-industrial complex. When you turn on employed people into prisoners, you can charge them for their consumption and offset those charges on to taxpayers while suppliers, politicians, owners of private prisons, prison guard unions, etc make money off of every additional prisoner put into the system. Essentially, anybody who gets put out of work by AI would more likely than not contribute to a similar system we're rather than have their quality of life improved, they will be forced into a less-than-ideal situation where their exploitation can be maximized while all the costs associated with a get pushed onto the remaining high-skill wage laborers. This doesn't necessarily mean everybody made unemployed through automation we'll get thrown in jail, although I'd be willing to bet a pretty sizable chunk of the poorer ones will be, but they will probably be thrown into a similar system where they can't actually enjoy the benefits of automation. Honestly, I think what you'll probably see people who are made unemployed through automation will likely get put into the prison-industrial complex, the military-industrial complex, or some other similar system where they are basically forced to consume products. The problem is that fundamentally, corporations only have the incentive to extract profit. This can sometimes mean forcing people to consume goods and services through force (like in a prison or through imperialism), and so even though society might be progressing, ultimately they have to expand. Capitalism is based on the assumption of infinite growth, so they either have to destroy themselves the recession or war, or find a new market through imperialism. It's an inherently self-destructive system that maintains itself through boom-bust cycles, where the average person's quality-of-life only marginally improved compared to those at the top, and suffering is exported to the people farthest at the bottom like in the developing world.
Came here for the topic, met Dan, and left with opened eyes and my feet comfortably on the ground. Thanks Adam, Dan, & their teams for the great conversation! (edited for grammar b/c it bothered me).
About food delivery apps, especially in rural areas: Here's a story I'm still annoyed about months later. I encountered a new nightmare where I used Pizza Hut's own app to order food but their system handed my delivery off to a Doordash driver. I'm certain the workers in the kitchen had my food prepped and in that dasher's hands pretty quickly, but then it lived in their car for over 90 min. I temp checked my food as soon as it was delivered and it was obviously below a safe temp for food that had at this point been out of the oven for almost 2 hours. They also have a call center that calls that don't pick up fast enough at the local store cover over to and they just place your order online for you. But if you have an issue they transfer you to the local store which then covers back over to the call center. So now I'm 2 apps deep and there's no way to get assistance. When they first started doing this they would at least text you a notification that they had given it to a dasher and provide their name, but that has stopped and so there's no accountability for this anonymous third party. I drove the mile and a half down the only highway through my small town to return the cold food for a refund because I wasn't going to make these poor workers have to add a rush remaking my order on top of their already busy plates because their corporate policies are to abuse these delivery apps instead of staffing properly. I left feedback on my experience with their corporate contact and basically got back a boiler plate response that yes they're handing overflow off to third parties due to staffing. I miss the early days of delivery apps when businesses wanted to disallow third party deliveries because they couldn't assure the quality would meet their expectations. Now they don't care because they can run delivery service without employing delivery drivers the same way Uber doesn't run cars and AirBnB doesn't own hotels.
This is such a pure example of the trust thermocline. Companies that automate responding to customer complaints WILL fail, sooner or later, and when it happens they won't even know why.
Part of the problem is that no business worth their salt has ever done delivery to rural areas. It's a bad idea, and in addition to not getting products delivered properly, I bet everyone lost money in the process. Let's have a bike route, ride in safely, sit at a table, say hi to someone in your community, and cut out the middle corporation completely.
@@Praisethesunson That is a hard one! They are all very varied in topic! I think a fun one to start with if you like gaming would be "Why it's rude to suck at Warcraft". But honestly every single one of them is great lol
@@PraisethesunsonI’ve been listening to his decentraland video a lot lately, but his nft video and flat earth video are both classics imo. I do also really like his older media analysis stuff, but I think he’s found an amazing niche recently with his anti-grift focus so I want to recommend those first.
He said that and I immediately said to myself "I'm stealing that!" In truth I'll probably credit him, especially since I'm a fan so it'll be easy for me to remember where I got it from.,
dan lost a lot of credit with me with his completely uncited nft video. i was frankly surprised no one else called him out on his lack of citations on that one, but i guess no one really cared that much.
As someone in the pagan community, I've seen a few pagans using AI like ChatGPT or Midjourney as a divinatory tool. So, they're very self-consciously doing the thing Dan says.
@@arturoaguilar6002 it's truly no different than using chicken bones or tea leaves. It's extrapolating "information" out of stochastic chaos, the point is the human interpretation not the input.
Presumably because of my ADHD, a lot of times I will jump from thought A straight to thought J in a few seconds and in order to explain my reasoning I need to slow down and go back and try to construct my thought patterns in a sensible way as if my thought patterns were sensible in the first place, and it's a drag. But sometimes I will get into a conversation with someone else who is able to see my train from thought A to thought J without explanation (usually someone who is also ADHD and familiar with the relevent subjects) and respond with thought M and then the conversation is no longer serially processed but parallel processed and there's this certain energetic vibe to it. And that is the vibe I got from this video. Excellent work. I love it.
13:42 "It's a failure to think through the basics of what your own proposal entails" That statement applies to so much viral information nowadays. Whether it is hype, a conspiracy theory, a rumor. It's like there is an epidemic of people just escalating and hyping up ideas without thinking about them. Part of it I think is due to the social norms of the internet. We operate on clickbate, headlines and memes. And all you have to do to perpetuate an idea and advertise it to all of your friends and followers is click a button. So something with no substance can easily become the thing everyone is talking about.
the metaverse hype was legit hilarious. I worked in a call center in 2022 for a financial firm (ik gross, but i gotta eat), and the hype this got from top ppl and the company was hilarious. We're talking articles, meetings, increased investments. Had they never talked to someone about it under 22? It was just shitty, monetized VR chat for thousands lmao.
I do not worry too much about AI disrupting creative professions. I worry a lot more about MY profession. When I entered the workforce, I was able to support myself pretty well, even without a roommate, working as clerical support (word processing, spreadsheet use, filing, etc). But tech completely disrupted that (particularly with the increased use of the Internet, autocorrect and predictive text) and since about Y2k, the only career path open to me has been customer service. It's a soul-crushing job that has taken its toll on my mental health, but I feel like the position I have right now is as good as it gets in that field, and I'm pretty happy right now. But it's exactly the kind of thing that LLM is designed to do. Most of it is answering questions and following instructions from a particular knowledge base. The only thing the AI couldn't do is the human side, which only becomes critical when a member escalates. So 90% of my job could be done by a computer with a sophisticated enough verbal I/O matrix... what happens when that results in a 90% reduction in the workforce in my field? Sure *some* of us can be reskilled into other jobs, but if this were not an effective way to devalue labor, the ownership class wouldn't be so keen on it, would they?
This is the part the carriage-driver argument never acknowledges. Cars put carriage drivers out of good, skilled work and denoted them to one factory worker with a dozen others waiting behind to replace him
Automation is already killing so many jobs. I know this as I work(ed? - currently not working) in automating people's jobs. Not entirely. But step by step the "workload reducing" scripts I write & end user facing catalogs I cobbled together are setting the stage to cut out jobs... and no one even seems to blink. Cause those people don't really get the tech, just want the workload reduction. The thing: I've long left those companies. Many of my good solutions live on, years later, still having removed that part of someone's job. In a generation or two there will be enough little pieces like that in place to remove people. So no need to worry about language models per se. They're just one out of many pieces. It's the constellation of many small pieces we should freak out over. But each little step is too small to really see. AI won't take your job alone. Together with all the automation steps? Yeah. Get ready to be replaced in the near-ish future.
In a dane world the reduction of work load created by aitomatization would just make work days shorter. We don't need work 8hrs/day and yet some people still work 12hrs. It's ridiculous!
Because I'm a software architect, I normally see immediately through every one of these over-hyped software products. But I always find the sales pitches absolutely shocking and the fact that the "media" seems to accept them with no research at all most of the time and changed the way I think about TV (and online) news.
58:52 I worked for a tech startup that was trying to pass off the work of humans in South Asia as “AI”. A now (or soon-to-be) defunct “shopping app” called Nate. Their “AI” could barely perform its basic functions, so a bunch of humans had to step in and the credit would still go to the shopping robot (that barely worked) and founder of the company (who is an asshat). If more of that happens, it will be up to consumers to call bullshit and employees to be whistleblowers.
I'm so glad Dan O's crypto video got so huge (>10mil views) cos it's an excellent piece (like everything he does). And his trilogy on "50 Shades"; he absolutely eviscerated it, from wonky fanfic to cobbled mess of a book to film 1 (which mercifully removed Anna's inner monologue - "double crap!") to the baffling sequel films, and so much more... Plus the guest voices reading out bits from the book, and the song. Oh, the song! (I genuinely love it!) Speaking of messing with the voice commands, I had a houseful of friends who didn't take long to find "other" phrases. "Xbox, go to the bathroom" opened the home page, and "Xbox, smoke break?" made it pause the game. There were plenty of others, but it worked even when I tried it, so it wasn't just specific voice or accent or anything. P.S. I was shown an advert recently here on RUclips for "Metaverse doing new & great things". "We've got legs!" Not trying to even take advantage of the way people were (reasonably) trashing the lack of legs, just literally faking excitement about having legs. Never mind how useless in the big picture the "feature" is. _There's no other word for it: cringe. It's cringe._
9:34 "It turns out that if we just create a flat page with a grid..." What a lot of people tend to forget is that text and grids and lists are all man-made and made for a reason. The goal of technology is efficiency often providing convenience. If you're data gathering, like say, trying to find how much calories is in a serving of cereal, it's more efficient to pull up a list of cereals and its calories per serving than walking into a store, walking down the aisle, scan the shelf and find the cereal you're looking for, picking up the box, finding the nutrition information box and finding the calories per serving in said box. The HTML format was made to be able to provide a simple, condensed sheet of information for people. If you force it beyond that, you're going away from said efficiency and, hence, possible convenience. Basically, imagine you buying a car. And then someone comes along and goes, "You know what cars lack? Tactile feedback. That's why our cars will force you to pedal a certain speed to maintain a certain higher speed. This technology will revolutionize the car industry!" The pedaling voids the point of the car.
Yeah, and like peak efficiency isn't always desirable -- it's fine to make something just because you think it'd be neat if it existed or because you want to -- but that's never how techbros and hype-runners frame these things. They almost invariably frame things like the metaverse as being innovations in efficiency, or depth of information, or in some way, quantifiably superior to the existing options, when in reality what they're pitching is almost always needlessly complex, less-efficient, and their desire to make it is based on "wouldn't it be neat" or "I could monetize this more easily," but they'll never admit as much. Like just admit your desire to make a VR metaverse is based on the fact that you read "Snowcrash" and thought it was cool. It'll be way more respectable than the idiot attempts to convince everyone that uncanny VR stores are somehow a better way to commerce that we're actually getting.
@@fanboy50 No offense, but if you're starting off a reply to my comment with "...and like peak efficiency isn't always desirable ", you've missed the point of my comment. Like, I guarantee that if these techbros (well, they're not really tech, they're just bros) successfully make the metaverse for their own personal enjoyment, they'll get bored with it and move on because it doesn't provide them with anything they couldn't already do better, fast and with more enjoyment. As humans, as living creatures, we're always looking for the next more efficient thing. The question is, in what way is something efficient? Power? Speed? Delivery speed? Fun? Less clicks on a webpage? Conveyance of an idea? Etc.
What a great pair, I love their dynamic. Adam is energetic and snappy. Dan is more low key and has a dry wit. I wish they would do a podcast or show together, It would be so good.
The amount of money Zuck has lit on fire with the Metaverse is fucking depressing when you think about what else could be done with that money. Also, Zuck was smart once, but power has been scientifically proven to completely break your brain.
Ironically, Meta put out a TON of great AI work during that timeperiod.... and released it to the public for free ;) Including some libraries that are like the backbone of a lot of AI projects (such as xformers). And most indie LLMs these days are now based off of Meta's LLaMA model.
I remember when i was a kid, we had a toy that was very much like ChatGPT. It was called the Magic 8-Ball. Wasn't as high tech, but you asked or questions, and you would get an answer.
To be fair, I think these tools have a lot more utility than a magic 8 ball, or crypto currency. But at the same time, they can do a lot more damage. Not because of 'Rogue AI' or what not. I'm much more worried about people doing dumb things with them because they think they're smarter than they are. It's a tool, and the good or harm it will do depends on how one uses it. I have a table saw in my garage, for instance, but that saw is not a woodworker. Nor would it know to stop if my thumb was about to strike the blade (Yes, I know saw stop is a really brand if you have thousands of extra dollars XD)
I think the greatest example of Clark's Law and magical thinking at the moment is the belief not that AI will "transform" X Y or whatever, but that AI will "just keep improving." My suspicion is that, while current tech might not be its final form, Machine Learning Algorithms will eventually hit a wall and they repeatedly come up against the problem of conceptualization, because I don't think more powerful computers or larger data sets are going to be enough to get over that hill.
I think you're right. AI is a mockingbird device, it can replicate whatever you feed into it, very impressively so. But mockingbirds haven't turned into Skynet just because they're good at copying things.
It's probably because if you want to use human thinking as a base, you have to understand how humans think, and guess what? Turns out that's nowhere near as easy as it sounds: it goes so far beyond magical thinking that it isn't even funny.
2019: VR will change the world. 2020: Crypto will change the world. 2021: NFTs will change the world. 2022: The metaverse will change the world. 2023: AI will change the world. 2024: Wifi connected buttplugs will change the world.
Adam and Dan are on very similar wave lengths but their personalities couldn’t be more different (in a good way!) they have the making of a good podcast duo.
The college in my hometown had "robot food delivery" where it turns out the robots were just remotely driven by some underpaid worker, somewhere, possibly overseas. For me the weird thing was the robots weren't vandalized or covered in stickers
The problem that stuck out to me with the LLM video game NPCs is that you're putting all the onus on the player to tell the game's story. The NPCs can say anything, but you have to know what to ask it to get it to say something interesting and tell the player a story thats meaningful. Replicating a life-like conversation system in a video game fails to consider that I engage in conversations IRL for a totally different reason than I do it in games.
When talking about the race to the bottom, it's especially noteable that the big argument for mass produced garbage is that "it's good for the people who can't afford the nicer handmade sruff". Meaning that we're aware the reason people go along with it is because they don't have the money to choose otherwise. The reason this has people so uneasy with LLMs especially, is because following that thought will turn *human interaction* into a luxury service. And whether or not the machines will live up to the hype, companies have already tipped their hands that this is an outcome they'll willingly push onto us if it becomes cost-effective
Companies won't hire "real" people even for cheap because they'll always seek profits. This means more people will be jobeless which means more people won't be able to afford real human interaction I think capitalism might not survive automatization. How do you sell anything if nobody has s job?
The first time I interacted with someone in VR, I was like, woah. This could be a game changer. Even with crude avatars, there was a physicality that you don't get from other forms of long-distance communication. Instead of talking to a disembodied voice or a small rectangular video feed, it felt like talking to a person. And this was just with hand and head movements. I saw a future where we got better tracking (face, eyes, fingers, legs), and better headsets (more comfortable, more convenient, cheaper, without requiring expensive hardware and external sensors), and this could become a common way to interact with people long-distance. Not the primary way, but another option, like how video calls are something we save for special calls. I think that's why Zuckerberg went all in on the concept. He saw that too, also thought it might be the future, and wanted to get in on the ground floor to be the market leader when it exploded. But in doing so, in tying the concept of social VR to boring office use cases and metaverse nonsense that ranged from embarrassing to scams, he basically ruined the public perception of the entire concept and defused that explosion before it could happen. I think that's one thing I hate most about this tech hype stuff. There's often something cool at the center of it, like VR and generative AI. But then these tech guys get in there and ruin it by trying to turn it into some unethical profit machine.
Pretty much yeah. Like when I think about ChatGPT and how I interact with it, what it reminds me of is not Data, from Star Trek, but the SHIP'S COMPUTER from Star Trek. The Ship/Station computers are never treated as people or characters, because they are not, but they are capable of all sorts of things due to their enormous databases, including generating images such as Holograms! And yet this never detracts from the 'human' characters or their own experiences/betterment. To me, that is the desirable outcome of this sort of tool. It's easy to be cynical though, given how much of an abusive relationship we currently have with technology.
An aspect of AI hype that I don't see discussed much is that it's a sink for all the GPU compute that's no longer useful for crypto - which is why nVidia is so in on it
What a great crossover! Dan brings a lot of calm and collected energy when he speaks, as well as being articulate and witty, so it can be a real treat to hear him speak. You two played off of each other very well! I hope to see you do this again sometime!
The argument that Uber will not switch over to AI drivers so long as there are human drivers desperate enough to work for cheap is more an indictment of the social contract than it is evidence that a robot taxi service is not in our future.
I like that Dan distinguished between the Metaverse and Crypto hype phenomenons and the AI hype boom we're seeing in that the AI hype is actually based on a real-world useful set of tech that people are just blowing up into a new gold rush, because I think Adam has been doing his Adam schtick with AI where he just hyperbolically declares something total BS without leaving room for the rational kernel of counterpoint.
@@MorgenWhite I can certainly understand the potential for bias in this regard. But I've also been playing around quite a bit, prompting various versions of GPT to write stories, and I don't think I really agree. One thing that I've been noticing is that while it can generate convincing sounding content. When I reprompt it on the same kind of story, I get similar superlatives, similar language, similar structures, over and over again. And not just in a way that rhymes but is almost outright word substitution for a previous generation. I don't write for money, but I do write for fun, and the forms are very very obvious in that regard. The model is decently reliable at determining a coherent series of sentences. But I think there's a lot more human programmed structure going on under the hood which is obfuscated by the part the LLM can do on its own. Which is hard to really say for sure because the later iterations of ChatGPT are much more mum about the actual goings on inside their software. But it also makes sense because one of the first user friendly applications of language models was sussing out the most common questions users will ask. Which means OpenAI has a ready made list of what things people are going to ask their language model and can implement guard rails as needed when testing shows the model doesn't give a coherent answer. A much simpler predecessor to these models is how your phones text prediction works. It's hard to say. The tool is certainly very powerful. But like Adam mentions, very similar AI tools have been implemented all over for decades now. The only thing that makes ChatGPT so startling is that it's an application that is public facing and user friendly. An example would be Bloomberg launching 'Bloomberg AI'. A Large Language Model that actual guts are just . . . the pre-existing Bloomberg Control Panel . . . They even went so far as to emphasize that Bloomberg AI is trained off of Bloomberg's industry leading data sets . . . which are the real prize of a machine learning model.
Thank god someone pointed this out. The entire discussion, Adam kept trying to group AI in with crypto and the metaverse, and you could tell that Dan was trying to take a much more nuanced approach. It's also very strange that Adam seems to think that LLMs (large language models) are what the new AI wave is all about, and keeps trying to draw this line of logic that basically says "No, AI can't do [thing X] because AI is just LLMs and [thing X] has nothing to do with language / text." LLMs are just one of many uses of neural nets and (un/semi)supervised learning. The question isn't whether LLMs can do the things that Adam seems to think that people think they can do. The question is whether the TOOLS that built LLMs can do a whole fuckload of other stuff.
Yeah, there's maybe 10-20% of an actual product with current AI, but at least that's way above the 1% of a product that the metaverse is (for a random member of the public, at least), and the ... 0%? of blockchain tech.
Freelance writer chiming in. I understand where you're coming from. I was mortified by ChatGPT's development, but decided to play around with it a bit and learn to use it at the very least. Yet, no matter how much I tried, I just couldn't get anything useful out of it. I got to the point where I was still doing all of the research, writing and editing, but trying to pare down information or summarize details to speed up my writing process, using VERY carefully constructed prompts and I realized it just wasn't worth it. Trying to use it to help me LITERALLY WRITE STUFF was actually just a waste of my time. And that's an LLM's prime directive, really. If they aren't good enough to help with that, what are they good for?
1:09:05 Speaking from experience, Dan just hit the nail on the head. The common advertisement in Uber Eats amd Doordash was "Up to $25/hour part-time!" That was true, actually! If you managed to spend 50/60 minutes in an hour delivering, you'd make an average of $25 from about 3 deliveries. But if you get stuck ***anywhere*** in the process, you watch your pay literally disappear with every passing second. I stopped delivering all Wingstop orders because I waited for 2 different orders for 45 minutes each. That turned my Pay-per-hour into $7, less than half of California's current minimum. The gamification of maximizing your time means it can be a slot machine on whether you're making $25/hour or $5.
The joke Adam made about the name of the Lakers Arena has already played out here in Miami. The AAA Arena, the Heat’s stadium, was renamed the FTX arena. The name lasted less than a year, I believe 😭😭😭
Thanks Adam! We love you and appreciate someone willing to call out the BS- if I see another Joe Rogan video w/Elon Musk talking about how AI is scary good! I'm going to freakout- and everytime I get an argument with someone in those comments sections I ordered them to do their own independent research or refer them to your channel 💜
All the big innovations that the internet allowed for came from a time when competition existed in that space. Stagnation is all you can get out of the same 4 or 5 companies that control the infrastructure of the internet now
As a PhD student who does research on experimental artificial intelligence algorithms, it makes me so happy to see other people see through the hype. Like it is SO DIFFICULT to get people in my own field to understand how much of it is a scam
Knowing the hype is fake actually makes me feel worse tho. People aren't going to lose their jobs to machines that do it better; they're going to lose their jobs to machines that do it *worse*
What a lot of tech bros miss when spinning these hype trains for technology is that in a lot of the sci-fi media they're pulling from, technology is so advanced that there are several steps before we get even close, if we even can. I mean, Star Trek's humanity is only able to live in a post-scarcity utopia because they did the scientific equivalent of inventing magic. The foundational principle of multiple technologies in Star Trek is the conversion of matter into energy and vice versa. I don't know if that's something that humans will ever be able to achieve and we're at the very least nowhere close to being able to do that. So until we can get to the underlying principles that make these technologies function, any attempts to just recreate them will be guaranteed to be sub-par to expectations.
Not just converting matter and energy, but the way energy was created was something so near-perfect in it's output that they essentially had perpetual motion. Star Trek tech isn't unfathomably complex stuff we just haven't invented yet, it requires us to break some of the most fundamental concepts of physics
"If you want to abuse the commons, I'm going to chuck your stuff into the river"
Dan Olsen is fucking based.
hopefully the Bow River
Those companies got permission from the gov and probably pay taxes and fees to the local gov, which makes money on rentable vehicles.
@@aoeu256 and where does that money go? Cause it sure as hell ain't going to schools or serving the public good in anyway. Hell of a lot of corruption at local government level.
@@motalux and for e-scooters you don't have to glue the pages shut.
Would make for a great T-shirt... _for permanent purchase_
I would literally watch Adam and Dan have an hour long conversation every week for years. this was awesome.
Me being a new viewer of BOTH is really enlightening. I went from knowing nothing about regenerative ai to learning more about "extended reality". And I'm in the small community that has been using IMVU, Secondlife, Sims throughtout my whole tween to adult years and slowly realizing their expansions on complete human extended interactions via online. Its wild just how fast technology has come. But its also been very slow for a long time. AI is just a scifi fantasticle escapade. Another use of 0's and 1's . Those who freak out about it also freaked out about who virtual chat sites would take over the minds of the youth. But here I am, an adult after using new technology and nothing beats real life experiences.
I'd pay for it.
There's an AI project that's an endless conversation between Werner Herzog and Slovaj Zizek. It's fascinating for about 20 minutes.
They have excellent chemistry!
I honestly believe Adam fails to understand the real potential and potential harm that artificial intelligence represents
human intelligence is not as unique as he wants to make it out to be, the reality is that human creativity is nothing more than the ability to blend concepts and ideas, and in that, LLMs are incredibly powerful and we are only scratching the surface.
I think your whole concept of skepticism and discovering the truth in everyday deception is very valid and necessary, BUT I think you are really losing sight of the kind of paradigms that LLMs represent, I really think you underestimate the potential existential risks and it is annoying and irresponsible for you to indirectly attack the voices that have been raised to warn about it.
you treat people like Eliezer yudkowsky and the like as doomsayers, who are motivated by some kind of financial gain.
The Perpetually energetic Adam and the absolutely deadpan Dan have simply the best complimentary energies.
I'm shipping
I completely agree. I want a show that's just these two guys talking about stuff.
I want a buddy cop movie with them, of course the bad guy would be some scaming billionare from the tech industry
Like a Cox'n'Crendor who actually know social psychology and social engineering.
@@JoshSweetvale i did not make that connection but yeah you're 100% right
For those of us who appreciate Dan's content, finding interviews about Dan's content where he goes even deeper is a beautiful thing.
I can't get enough, for real. He needs a podcast of his own, but then we wouldn't get any more videos
content isn't real stop using corporate language
@@cyndaguy fair, but this is pretty nitpicky. Call it a "perspective" or whatever. Every discussion on these topics has to mention 0% interest rates and Dan did it because he's bringing some bigger picture thinking.
Have you found other good interviews with him that you could share?
Agreed. This interview in particular was a real treat. I've not previously encountered Adam's work but he seems to understand the subject matter very well.
"They say it's AI and it turns out the AI is a sweatshop" That's sort of what the Amazon Mechanical Turk program is tbh. Like no, AI didn't read your product receipt, I did, for two cents.
Really? Reading a receipt is something even AI from the 90s should've been really good at. It's not a very difficult task, especially because a computer printed the receipt.
@alexc4924 Actually, I can tell you from experience that you are right, but it doesn't matter. I did a project when I was unemployed to try and track what I was buying.
You can get the text of the receipt pretty easily, but, receipts vary massively from vendor to vendor and within each vendor the text is highly abbreviated, and it is not always obvious what the abbreviation means.
Also, sometimes, you actually can't even get the text off the receipt. Remember that these things start off as pristine OCRable text, but then they get crumpled up and shoved in a pocket, get water damaged by sweat, and then the friction/heat sensitive paper gets to fade for three weeks before you remember to scan it.
@@thewhitefalcon8539 Not really, it was possible from 2011 onwards, as there was hardware restraints, and Machine Learning plus newer hardware made reading receipts at scale easy. Ai, there is no such thing.
Mechanical Turk might be the single most on-the-nose name they could give it.
I'm really loving this anti-hype arc of Dan's career
Big contrast to the Hypebeast persona he cultivated early on.
I loved all of the earlier film editing stuff and hilarious reviews but the new era is easily the best stuff on RUclips.
@@TheK3vinfair
I preferred his makeup tutorial era. Cateyes _on point!_
Joking ofc, but all his stuff is great. The film concepts, as noted, & his film reviews that go deep but often hilariously cut deep (notably _The Snowman,_ his _50 Shades_ trilogy, _Su[¡c]¡de Squad_ - also not nitpicky nor cheap; he's clearly knowledgeable & passionate about film, but his approach is done in a way that's intelligent but avoids being tedious)... Also niche/cult-classic film history stuff like "Bakshi's LOTR" and his really cool "bigger picture" type essays like the one centred around "Contagion", released during the height of 'that plague' (which still hasn't ended, just a reminder) and the one about Jamie Oliver's weird quest against nuggies "to make school lunches healthy" (it's about more than just "kids being fed healthy food", much more).
And I'm always pleased to hear Dan's voice doing a quote or something on a fellow RUclipsr's video. Anyway.
@@TheK3vinwhen he had a hat?
Adam Conover and Dan Olson is the crossover I didn't know I wanted to see.
Now kiss!
SO excited to see my faves talk ❤
I clicked so hard my keyboard broke.
I clicked on this video so fast!
Adam Conover & Dan Olson, the crossover I definitely wanted to see! Honestly, I saw "Dan Olson" in the title & thought, "It couldn't be...? I'm not that lucky. Is it?" Twas! What a legend! Been watching his channel *Folding Ideas* for years. They're really exceptional.
As a User Experience Designer and Researcher who has a background in Anthropology, Adam and Dan are spot on in pinpointing what's wrong with tech thinking when it comes to people using technology as tools/tool use. I'm thrilled that this is *finally* being discussed!!
Yep. They perfectly describe the pattern playing out over and over. I've felt absolutely insane watching it constantly repeat without any serious pushback!
Same here! It’s the things we yell at our stakeholders and they never hear 😂
As a UI/UX Engineer, I wish all Designers had Anthropology degrees. Most of the problems tech creates is due to the people designing the systems having low social intelligence.
❤❤❤
@@thatadamwhitleyThey should at the very least be encouraged to go to social events regularly. Because lets be honest, these guys spend hours in the gaming chair cut off from the world every day.
The more I learn about the modern video game/“web 3”/metaverse tech stuff, the more obsessed I am with the connections to Roblox.
Roblox is almost EXACTLY what all these people are trying to create - it’s a platform with an extremely wide reach like Fortnite, it’s a vertically-integrated content host/provider like social media, it has a speculative NFT market with items that can be carried through every world and game within the space.
Roblox is also huge, and preying directly on children. Game development via child labor, wage theft, gambling on NFTs, 3rd party gambling websites based on the NFTs, all packaged and sold in the same way as a Fortnite battle pass.
It’s such a huge blind spot for industry experts and I’m blown away by how much they’re seeming to influence and how much they’re getting away with.
@@chase-warwickPeople Make Games has a great video on why Roblox is preying kids: ruclips.net/video/_gXlauRB1EQ/видео.html
Roblox is in its own world man it pulls billions a year and it's easily the biggest game medium in the world
@@chase-warwick i'll have to direct you to People Make Games's documentary on the child labour stuff. like yes it is like what you say in a fashion, at the same time it's telling literal 10 year olds they can have 100s of dollars if they crunch all their free time away, not to mention the real messed up creator group hierarchies that can abuse adult content creators and there's no vetting system for minors in these groups. the edge cases are dangerous
@@jacksim5759@jacksim5759 That's true, but the comment has a point: for all it's flaws, Roblox isn't as bad as web3 because it's confined to virtuality. Those kids aren't actually living day to day on scrip like they did in mining towns. Web 3 is comparatively more dangerous.
That's fair. I really, really dislike Roblox, but it's nowhere near as bad as cryptocurrencies.
The very premise of cryptocurrencies and anything based on them was fundamentally broken right from the start. They were never going to have net legitimate utility.
the proof of this fabricated hype is in how Altman keeps saying how "AI will turn the world to ashes," and "we need to regulate AIs cause they are too awesome", but then when Europe starts thinking of applying copyright laws to AI, or regulate machine learning, he goes all like "fine, no chatgpt to you, loser". like if it is dangerous as he says, then he should be happy for regulations.
of course, and the reason they're spinning that "ai will destroy us" thing as if it's the terminator, and not the obvious "we're going to steal everyone's information so we can profit off of it and soak up all the revenue from various industries" is the misdirection, and by getting up in front of this shows his intent is regulatory capture. They are spinning this as if it's skynet, when it's an advanced autocomplete scrapping data. They want to be given the means to control things themselves which is the entire point.
You are severly mistaken in your assesment, You think too much of yourself and of your race (mine too, i am a human), we are not even close to being smart or intelligent, and there are things that can and will be smarter and more powerful than us and no self-imposed virtual and ethereal grandness will save us. You wont even have time to realize how wrong and overconfident you were.
@@user-yl7kl7sl1gThe issue you're describing isn't "LLMs being more intelligent than human beings," it's "humans using LLMs' vast information repository to learn things they can use to harm others."
@@user-yl7kl7sl1g the thing is language models aren't intelligent. They aren't even true AIs. Like a lot of "sentience" people see in AI is their mirror neurons at work. The only way they could be more "intelligent" (which translates to more time-efficient and less expendable) is through Data, as well as more complex hardware and more complex algorithm. These AIs are made and run by people, they aren't independent. The only real danger would be one company having our collective data stored to change and modify. What Altman fears is competition, a lawsuit, and loss of money, not the end of the human race like Matrix.
Like, the only "AI apocalypse" I can imagine is if we completely substituted human work with AI. Generally AI needs data to be on the same level of humans. Constant data, since we humans are constantly evolving and creating stuff. but if nobody creates nothing in the future and it is all AI generated what we would create is synthetic data: data created by an AI. This synthetic data theoretically can be taught to the machine, but it is proven to create this "cannibalisation effect" where the language model decreases in quality because it copies the biases and errors AI-generated content possess. So it would eventually collapse to the point of being unusable. If our entire systems are all AI based, this eventual cannibalisation could mean loss of not only of modern infrastructures (all AI based in this hypothetical future), but of our entire digitally recorded history. And if that wasn't enough, we must imagine these future humans to have had relied on AI for their entire lifetime. Therefore we would have completely underskilled, mentally underdeveloped humans thrown to square 1 of human civilization.
He only wants regulation for cuckoo fantasy scenarios like AIs becoming so smart they transcend reality, not for actual thinhs his company might do in reality.
Could you imagine paying hundreds of thousands of dollars for an NFT sword and then you bring it into Breath of the Wild and you hit one dude and it shatters?
That would be pretty funny
It shattering in one to two hits is such a funny thing to think about
I cannot, no.
I can imagine an NFT bro doing that, and it certainly puts a smile on my facr.
Thats the really insane stuff, transfering weapons from game to game is an absolute balancing nightmare without any potential payoff
With Adam and Dan's powers combined, we truly can ruin everything, or at least everything horrible.
I'd ship it
Is AI horrible (plz dont call me an AI bro simply because i think there might be good uses for it)
We truly live in the best of times and the worst of times. ❤
@@TheManinBlack9054 I think its use outside of the scientific sphere is largely a net negative for humanity and society to be honest, however it has some important uses in data analysis and simulating protein folding in the sciences and medicine.
To be fair, it's not hard to ruin these amazingly tone-deaf ideas that they mention. As soon as you start to read into it and understand the functioning behind all this recent techno-crap, you realize it's either much, MUCH less than what they sell, or simply a rehash of something that already existed way before.
What I feel the Metaverse thing misunderstand the most and everyone should have learnt by now in the post-pandemic world is that people do want to go outside and do things with their friends and families.
Metaverse is the real life story of Sword Art Online. Bunch of gamers get trapped in their favorite game world. Spend the next two years fighting to get the heck out of the game world XD
@@Bustermachine Hell, SAO even showed you why being permanently stuck in a virtual world is not physically viable or good for you: your IRL body becomes malnourished and weak, because virtual foods, no matter how many signals they send to your brain, don't provide nutrition.
Even if you don’t want to leave your house much most people want to be connected to real people, real experiences and reality.
Oh, I'm very happy to stay inside!
But if I'm going to be doing something with graphics I want it to LOOK PRETTY (which the Metaverse doesn't), be able to accept ANY KEYBOARD (Lost Ark, I'm glaring at you), and, you know, other things which apparently the teams headed by the richest dudes in the world have apparently never heard of.
What I really want is better public transportation and infrastructure so I can see my friends and family
Jesus, I remember when Dan-O was "just" my boy down in Calgary having Socratic discussions with a milk-carton-turned-hand-puppet about the narrative possibilities of a Dwarf Fortress playthrough. I feel so weirdly proud and happy seeing him here now. (parasocial relationships be like that)
My introduction to Dan was watching him chug cough medicine after watching Suicide Squad 2016. That's my permanent mental image of him.
I still miss the early jank of being able to hear the puppet's cardboard mouth flapping
I think Dan is on of the most impressive transformations I've seen on RUclips. Starting as just another faceless avatars taking apart games and movies, through being a knowledgable movie buff, now creating those elaborate, multi-hour videos that help millions of people understand very complex subjects. He could have stopped so many times, and just keep making content in the same niche. But he keeps being hungry and ambitious. Such an incredible dude.
@@andrzejsugier
Dan Olson is just one of THE GUYS who helps people understand whatever he’s discussing
His The Wall video is one of my favorite cause he actually engages with the metaphorical medium with good faith and critiquing what’s a lil juvenile or poorly explored
I've been a teacher for 30 years and every few years we get trained on another transformative method or technology that will completely change our teaching. Spoiler: they never do, though occasionally we get a nice tool. Effective teaching is about relating to students.
Enterprise tools, teaching tools, and tools for government employees. They all have the same flaws, the person buying them isn't the person using them, so they're optimized for the best "on paper" abilities, rather than caring about the people who are actually going to use them.
I don't envy you. When I hear what teachers have to put up with, it makes me so sad.
@@Keisukiboom!
I'm a teacher as well. Connecting, effectively communicating abstract concepts to children is really the skill we get paid for.
How is AI not an important tech innovation, but GPS is? I think the hype is a little bit justified, to say the least.
@@TheManinBlack9054 I don't see how this relates to any of our comments ^^'
Please have him on again. You guys have a great natural chemistry, and it makes for some really funny observations.
When Adam talks about AI NPCs I can't help but think of the "games" my 5 yo invents. The rules are constantly changing and updated, the goal is unclear, and it ends up going nowhere, without anyone knowing who won.
The bigger issue with AI NPCs, IMO, is that you don't typically want too much creativity from a game AI. You want something that behaves in 'clever' ways relative to the bounds of the intended gameplay experience.
Think about how much the AI, in say, Halo, of Half Life, or FEAR, was praised. But in reality what made these AI seem so convincing was that they had well telegraphed, reliable, behaviors.
You may not know if a combine soldier was going to chuck a grenade through the window, but you leaned pretty quickly that it was a thing the Combine's AI could do.
You knew when you stickied an Elite that he was going to get angry and then charge at you in a Kamikaze rush. Etc . . .
To a degree, you want consistent behavior, you just want very complex and consistent behavior.
Calvinball
Fizzbin
So basically the best games ever invented
Like, anyone who talks about AI generative NPCs should be forced to watch a Let’s Play of Facades in order to show them exactly why that’s a terrible idea.
Adam's description of people saying weird shit to make NPCs say wild nonsense perfectly describes my interactions with Google's Bard.
"Xbox, smoke break?" took no longer to say than "Xbox, pause game" (though what happened to just pressing the pause button?), but it sure was more fun.
I don't trust "listening" devices, but it was fun to do at a friend-group's house full of Xboxes with the voice command stuff not disengaged.
"Xbox, go to the bathroom" would make it go to the homepage/main page.
Didn't we try this with a Twitter robot that had to be deactivated the same day because it learned to be unspeakably racist in a matter of hour?
@@mookinbabysealfurmittens Xbox trying to say something?
@@kevinhengehold4387 After like 24-48 hours, as I recall, they shut it down cos it was already trained by trolls to say awful shit. Pretty hilarious, honestly, cos these tech giants sunk so much into it yet just forgot about 4chan?
I'm so glad someone finally pointed out how EASY it is for people who own AI to just....lie about it.
A disheartening quirk of platforms who've taken measures to at least hide AI stuff that's flooded them that I've seen is that on pixiv, a japanese site for posting art, has the option to turn off the ability to see art tagged as AI. While you might think people would just not tag their AI stuff as such to get past the filter would be the problem, the bigger problem I've seen is that people are just straight up stealing people's actual art and tagging it as AI to minimize the chance they get caught.
The sensible thing would be to create a separate website for aggregating AI content, but of course for many people that would defeat the purpose. (of tricking people)
Literally the only thing Driveshare apps did that made me want to use them over taxis is just not having to make a phone call and knowing how much a ride would cost before you summon the car.
Exactly. Where I am from, the taxi business had a weird aversion against any type of renewal that Uber seemed like a God send just because it was easy to use and you could minimize scams because it was all monitored on your phone.
Both of which are such an easy fix for the taxi company's to make. Hopefully we'll see the return of properly paid taxis in the future.
Yeah, that's the thing: Uber actually did have a product, it was just a quite small app. Calling cabs was terrible, back in the day: you had to call and talk to someone, and then they told you "five to fifteen minutes," and then maybe a cab would show up, or maybe not. And maybe they could find you, and maybe not. And twenty minutes goes by, and you call back, and you get "five to fifteen minutes."
Cab company+uber app would have been a legit product! A money-maker, even! Just, without the labor violations and exploitation, not the kind of money-printing scam machine they wanted.
Where I live all taxi businesses have apps that do all of these things.
@@lady8jane but when Uber came, very few places did
Good to know that Dan has the exact same speaking voice in his videos and in casual conversation.
There are some old stream highlights on his RUclips that sound like he scripted them because of how intelligently he's able to talk about some subjects off the cuff.
@@blueisasomedancer big brain dan
The physical reality of looking at a product in a store also lets you see things about that individual product that you'll never see in a Metaverse. Physical defects like scratches, holes, discolorations, loose joints, smells, everything. There is no way in hell that any Metaverse version of a store will upload full 5d (throw in some "d"s for smell and other stuff) scans of every single product that goes through their warehouse, tagged so that a user can pick out a hyper-specific product out of the stock. For everything where that specific ability to judge a specific instance of a product doesn't matter, web pages are fine.
Web pages have been deliberately broken by our corporate overlords. Specially to waste your time trying to monopolize your attention.
I tried ordering produce through a delivery service once. Never again.
Yep. At that point it's just cheaper to set up a real shop. Why recreate all that stuff digitally? The whole point of digital spaces is to not have specifics. To be interchangeable and non-specific. That's where the savings come in.
The thing is, the people who are coming up with these idiotic ideas don't _know_ what people do in physical stores because they themselves are so far removed from it. Do you really believe someone like Zuckerberg understands that when a person picks up a carton of eggs and opens it that they aren't simply going through the motions of a societal ritual but are checking for broken eggs?
@@IshtarNike I'm almost surprised that they aren't trying to gig-ify the whole endeavor by forcing doordash shoppers to wear a body cam connected to the user's VR headset as a cheap way to "virtualize" a whole store. Now that shopper can have the customer directly in their ear telling them where to go and what to pick up, so convenient!
Wait, hang on, that was a joke Meta, don't actually
I love that in response to AI stealing art to train their algorithms, people invented Glaze to put art through basically a noise filter. Barely changes the original art, but introduces that noise into the datasets the AI trains on. Thereby corrupting the output, as this digital noise compounds to the point the creations are unusable. Effectively poisoning the well. It's fantastic, and I hope AI art programmers never figure out a way around it.
EverywherecI looked, the newest AI is openly described as a "pastiche machine" - I've no idea why any of you seem to think pointing this out breaks some kinda "official narrative", it doesn't lol.
@@WreckageBrother-rd5zf maybe because of all the tech bros swearing up and down it doesn't.
@randomusername1735 It steals art in the same way humans steal art and base their works on it. aka "inspiration"
@@Brandon82967that isnt how ai art works. Ai doesnt have inspiration.
@@alejandrotuazon4831 More precisely, AI doesn't suffer or benefit from the interplay between skill and psychology. For example, the effects depression, schizophrenia, or joy have on artwork.
The mistake a lot of tech bros make is thinking art is all about skill and technique, but disregarding the heart, soul, and all the other messed up crap inside of us that makes art less about the picture we see and more about the story it tells us.
I remember the day FB changed to Meta. It was because of the whistle blower. I worked in FB ads at the time, their whole platform was down for 24 hours right after that hearing. They probably thought the whistle blowers evidence was too substantial. She primarily talked about how Meta uses engagement data to make people sad, and mostly teens. They want people to be sad and angry because it makes people buy things for their ad platform. It's disgusting.
I know this because I remember watching the hearing live, and working in FB ads at the time.
Omigod i want to know more right now. Why did that leak in particular trigger the name change?
@@fangirlfortheages5940 i think it was Zuckerbergs way to smoke and mirrors the whole situation. People seem to forget things when there's diversions over and over again. They changed the whole company direction when they announced their name as Meta. Many celebrities do the diversion after diversion tactic as well lol.
I remember their servers were down all day - atleast 13 hours. I think they were wiping data the whole time to ensure they'll get away with it.
The hearing was very convincing and I think they needed a reason to turn the servers off for a while. The whole meta thing was probably in the works for years and the hearing triggered a perfect time to do it.
I could be wrong, but everything seemed to line up perfectly. The whistleblower must've been too right. It would make sense since facebook could barely keep top talent at their company. Watch the documentary The Social Dillema!
So that if/when the company is sued, the headlines will read "meta sued for social engineering" and not "facebook sued for social engineering"
I’ve never listened to a podcast before, but if Adam and Dan start one together I’ll toss some coin in a tip jar. This video was awesome
YES
"I've never listened to a podcast before" - like, at all? Not just not subscribed, but actually never listened? I'm actually impressed!
I legitimately laughed when they called Zuckerberg a loser for renaming his company meta and now nobody is talking about the metaverse. It's so true tho. A fat L for the robot fratboy
Techbros and renaming companies to the stupidest shit - name a more iconic duo
Sure, must be awful being such a loser-
Must be crying into his billions of dollars
As a DM, I can confirm that players do want to ask nonsensical questions and/or have sex with most NPC's
But overall they are all there to actually engage with you and the story and campaign, of course some people will want to test limits but everyone is on the same page...i.e. we are here to role play and have fun right? If someone comes into the game and starts questioning physics and asking why about everything and running around slapping all the villagers you are quickly going to have the in stocks by the city guard.
Lol it will just turn into the NPC's nervously looking around "hey guys I am really here to talk to you about the dark Mage of Azimoth, not the Minions movie..."
@@killerbee065 A little of that sure. But I've had to deal with a nat 20 seduce lich attempt from a tabaxi bard.
I’ve introduced multiple NPCs that I didn’t know would immediately be latched onto by my buddy, and both times they’ve been taken - one was a married lesbian woman and the other is betrothed to one of the party’s allies lol
“Melons”
Im a CS student with a background in ML and a 3d printing hobby, and hype is so real and theres constantly claims that fundamentally misunderstand the capabilities and practicalities of the technology
Whether your expertise is in ML or in the thing they're trying to shove ML into, it's really jarring to see how little the hype artists understand about either.
The hype is pretty bad because the technology is still not reliable enough for me to trust mission-critical tasks.
The whole impracticality of AI reminds me of how I was browsing someone's art blog on tumblr, went to their "about" page, saw a picture of their business card which had a QR code on it, and read, "scan this QR code with your phone to go to my personal website!". I was like... Are you kidding me? That is so convoluted. Just put a link to your website there. Not to mention, what is someone supposed to do if they're viewing that site on their phone?
...You do know that Lens is a thing, right?
@@Troisnyx still a pain in the ass nonetheless
also i hate this idea so much like i put so much effort into not needing to use my phone for everything and honestly if i came across this i just like
wouldnt
@@Troisnyx Yeah, another app/extension/service to install and harvest your data instead of just being a local camera reader.
My mom tried to convince me to "learn chat GPT" this morning because it would "look good on my resume"
That is the most "well-meaning mom" thing I've ever heard of.
And now he's riding the digital super highway to hell
I totally told my college-aged son the same thing lol.
Well chat gpt is a quite useful tool but you can't trust it. You have to expect that it might tell you nonsensical stuff. I use it to do math stuff for me, so i don't have to waste my time on it, and also to pitch me the idea how to solve something when i get stuck. It usually gives the right direction, but can't properly solve stuff. The gpt3 is mostly useless, but gpt4 is actually an useful tool. Even for advanced math. It can do the easy stuff well, but for more advanced things, it usually takes few tries to get the right direction for something. Its a great tool for doing proofs, because it spits out kinda related ideas and that is often what you are missing.
It can cut down the time to figure out the right approach massively. But still it can't do it fully atm. Maybe gpt5 will actually be able to do most of the stuff correctly. Lets hope
It’s funny that you could probably get a job putting that in your resume, but i wouldnt want to work for anyone that hires me on that
As a fan of both Adam and Dan, I think my heart might explode.
Right. XD It was a pleasant surprise.
Mine has already exploded. I'm literally dying rn
@@MCArt25 RIP ☠
This video is great. I'm a developer myself, and recently I've been floating my resume around. Lots of companies won't even look at my resume unless I include something about GPT and other LLMs. I have used the systems in my work, but they really can't be used to write entire applications without a chaperone and they can't be used for anything that's not surface level. They don't understand performance, idioms, maintainable code, efficiency, etc. The reality is that these tools effectively have just supplanted search engines and other websites like stack overflow. They also help with some of the grunt work but that's about it. And forget writing something new with these systems, they can't really create anything new because they just regurgitate patterns based off of the data that they were trained on. There's no creativity or understanding of the topics that they cover. None of these models even understand what language even is in the first place.
What's really troubling though, is the fact that these systems are effectively black boxes. In any other domain of programming, if you created an app that was a black box then you would be in a lot of trouble. If you can't have any kind of introspection into the workings of the system itself, then there's no way to verify that it's doing exactly what you think it's doing. For all you know the code could be firing missiles or hacking your system etc. In the case of these LLMs, not only is a functionality a black box, the concept itself is also a black box. Even the most intelligent researcher does not fully understand how these systems work completely. As far as their concerned we threw a bunch of linear algebra, probability and calculus together with a ton of data and somehow that produced a system that could replicate patterns.
Shit gets out of hand real fast!
I think it's a bit much to say the concept itself is a black box. After all, you just described the basic principle of what's going on and I'm incredulous that you can make something this complex by just throwing math together and seeing what sticks. It's very similar conceptually to certain fields of statistical linguistics. Not to mention implementing guard rails/directing the output would not be practical if the developers didn't have some idea of what the model was doing.
The system isn't conceptually a black box.
But rather, getting to the change of events behind any one output is a black box because it's so damned convoluted. And therein lies the danger that you end up with this pooping its pants because it suddenly thinks the Brothers Gracchi, the Heimlich Maneuver, Critical Race Theory, and National Economic Planning are all deeply interrelated and the solution to life the universe and everything is 42+Pi.
Edit : Another issue, of course, is epistemology. Namely, as you mentioned, this is being regarded as a stand in for using a search engine. But ChatGPT is no less credulous than google if you feed it bad information.
I've seen at least one and it's so generic as to be useless. Immediately denied the application. Also had "prompt engineer" in skills 😵💫
If that were true, how did it pass the bar exam
@@Brandon82967 just because it can regurgitate information doesn't mean that it actually understands any of that information. There is a reason why it hallucinates so often. I don't care how much data they feed these current models, they will never be able to get rid of the hallucinations with the current generation of AI.
There are many research papers on this subject, though I think the most important one is one that came out around the time that GPT started becoming very popular. I'm not sure if you're familiar with alphago, it's a neural network based on the same technology as these LLMs that was able to beat the world's foremost go player consistently over a series of matches. The paper highlights that the algorithm itself doesn't really understand what go even is on a fundamental level.
Essentially, they found a strategy where you attempt to capture large sections of the board in a way that allows you to take all of the stones of alphago. The model seemingly ignores that you are winning until you finally captured all of its stones by which point the game is already over. One of the main rules of go is that if you surround a section of stones, then those stones are captured. What this means is that the researchers were able to find a strategy that used a very basic concept of the game to beat this algorithm because the neural network did not understand the game. The only thing that the neural network understands is the patterns of play that can arise from playing go, and that's where the important distinction comes from. If you try to use the same strategy against a human being, the human being will immediately counter it.
Man, Dan's point about people treating AI like an oracle was spot on, and is one of the things about LLMs that scares me the most
I'm listening to this conversation while prepping vegetables and saying out loud, "But the problem with that business model is -" and then Adam or Dan finishes the sentence for me! What an amazing conversation. Thank you and I'll see you in Baltimore.
Adam’s statement about video games requiring restraints to function is very poignant as a game dev
Everything has limits
Things are, in the strictest sense of the word, defined by their boundaries. (If you walked into a room where light was evenly and equally projected from all angles and directions, with no shadows or obscurants, you would, quite literally, not be able to see a single thing)
Something most gamers don’t seem to really understand
Heck, without restraints, you character can't even properly walk around, because floors are technically a type of restriant
It’s not just constraints, even if you roleplay and say exactly what the npc wants, someone still has to make that mission, someone still has to make the “correct” parameters
Dan has stared into the Sun long enough to obtain wisdom but somehow kept his eyesight. I am cannot imagine how terrifying it is to know as much as him.
Everything he says sounds like obvious common sense but only after you've heard it, before that you couldn't even imagine. Like the whole "flat earthers are all into QAnon now" thing.
he's not necessarily smart or wise. He's observant, rational and good at critical thinking. These are all skills you can (and should) develop.
AI NPCs would make for an incredible speedrun category where you try to break the AI by yelling gibberish until it completes the final quest for you.
Honestly that sounds just about as horrible as "blue shell%" in Mario Cart lol.
There is no final quest because the quest is also AI generated.
Going further committing cybercrimes and hacking into Activision by walking up to an NPC and speaking enchantment table.
So basically Skyrim
'snot gibberish, 's multilingual cursing. I can call the stupid computer a total arsehole in five languages.
The notions discussed around 37:00 about Tech Billionaires desperately trying to build new Walled Gardens to entice us into (and then, brick up the walls behind us) is... Both resonating and terrifying.
I worked Uber/Lyft for 3 years. I loved the job itself; meet all kinds of people, play my own music and podcasts, I was one of the top 10 rated Uber drivers in my city, it was great.
BUT, after taking out for gas, carwashes, etc, I was making $10/hr plus paltry, infrequent tips. I was also putting 50k miles a year on my car.
The final straw was when I got sideswiped on the highway. His fault 100%, his insurance covered everything. But I was completely unemployed for two weeks while my car was being fixed.
No safety net or job security is, imo, the biggest problem with all the gig economy work.
Dan’s perspective is always a breathe of fresh air.
he's completely uninformed. just look at the citations on his line goes up video.
@@snowballeffect7812 I have. What specifically did you have a problem with?
@@rakino4418 the fact that they were all blogposts and none referred to the available economists who did research into the matter or even the color papers lol. he mentions zero-knowledge proofs zero times lol
@@snowballeffect7812 If I clicked on a video about the interface between people and crypto and the creator started talking about zero knowledge proofs, I'd probably stop watching because it would be readily apparent that the video wasn't ever going to actually get to any sociological subject matter with any kind of depth (unless it was like 8 hours long). Not everything has to be a 3blue1brown math explainer; there's something called scope that good writers pay attention to.
@@tracyh5751 he talked about mining and minting and DAOs but somehow the absolutely amazing notion that you can prove you know information without revealing any information is too much for you? lol.
If he's going to make claims, he needs to cite them. He didn't and he made false ones. His opinions are his own and i generally agree with him, but it was disappointing to see him not do any research and being wrong about what he was talking about. Misinformation is always bad.
I can really just listen to Dan Olson talk forever
Omg omg! I'm so proud of Dan, as a fellow Calgarian it's so cool to see him gain this kind of prowess among the likes of Adam and others. He deserves it! Great work both of you
Hi from Montreal! Us Canucks are taking over the internet haha
Fellow Calgarian in the digital marketing industry and man everything they talk about is the nonsense I have to mentally filter out on a daily basis. Every anti-hype video I watch from one of these 2 is such an exercise in affirmation because I feel my cynicism is justified in so many ways. Sometimes I feel insane because all the rhetoric swirling around on social media is so counter to my gut instinct, but guys like Adam and Dan keep me feeling sane and valid.
Same here, it's awesome having someone as well spoken and observant and just generally engaging as him repping our town especially since a lot of Canadians play their hometowns very close to the chest.
The best justification I've seen for Zuck diving into Meta, was that he desperately wants to establish his own version of the Apple platform. Basically, create a platform (IOS and IOS-based hardware) where he can charge 3rd party developers a cut if they want their app to be included on the platform (the App Store). It's an enviable 'passive income' model, that many businessmen drool over.
I wish there was more discussion on non-creative jobs being overtaken by automation just as predicted. I'm a window cleaner by trade and over the past couple of years it's become very clear that we won't have a call center in the next 3 years. Not only will things be scheduled, invoices sent, and routes planned by the computer but will also have less drive time less drags in the system and more money on the books. I'm personally not a fan of a society willing to slash jobs without compensation outside of the labor market, merely because the incentive of monetary wealth is too damn high in today's world, but that's a whole other discussion. This isn't AI like the language model they've been discussing (for the most part) but multiple algorithms working together to produce better results than humans ever could
I want every job that can be done by a machine to be done by a machine - a good part of what makes me good at my job (word processing) is my ability to make the machine do the repetitive stuff, or boil down a series of steps into a single button-click.
The problem with labor-saving technology is that the benefits don't accrue to labor, in the form of a shorter workday/week/life with better living conditions, but solely to capital, in the form of increased concentration of wealth and power in the hands of those who own the machines, and immiseration of everyone else.
@@dwc1964Hmmm, I agree the benefits will go disproportionately to those who own the capital, but to say it will go _solely_ to those who own the capital is perhaps a bit of a reach. It is possible for everyone to benefit at the same time that inequality gets worse. The rich get a lot richer, the poor see moderate gains.
Ultimately, labour are _also_ consumers. People who currently work in call centres also need to buy food and pay for utilities and services (which probably cost more than they theoretically could with AI reducing inefficiencies). Some portion of those who work in call centres _will_ retrain as something else, another proportion will be retained doing something difficult-to-automate in the call centre, and might get a pay bump if their job is higher skilled. Some amount of the additional wealth generated by AI _will_ be taxed by governments and put back into public goods and welfare.
Fundamentally, it's kinda hard to imagine a scenario where human society gets more efficient at delivering goods and services, and the average persons quality of life does _not_ improve.
@@merrymachiavelli2041 I think the argument is that the machines will take the work and there won't be enough demand of skilled labor and society won't adjust fast enough before we have unprecedented inequality, or worse a mass culling through poverty or civil unrest. You can look at most of the job market post Covid, nobody wanted to work minimum wage being a waitress or cashier when they had been getting more than that from the government in stimulus and yet businesses were starving for employees but refused to pay livable fair wages. You had a few businesses "experiment" by paying low skill jobs $15/hr and suddenly that business boomed.
Shocker, we learned that both business CAN provide fair wages AND the government didn't implode by providing a basic income for everyone.
I agree though, automation and general efficiency increases will reduce the price of everything; so everyone benefits. However, we have to leave behind some draconian concepts that contribute to inequality. The future is going to be more socialized or it's going to be a war from the poor. Looking at history, I don't have a lot of faith our economic systems will catch up or change fast enough.
@@merrymachiavelli2041 that's right now my guy
@@merrymachiavelli2041while I'm not saying that this is exactly what's going to happen, what you're describing isn't really the trend we've seen.
As more and more jobs get shipped overseas or automated, what we've seen is that "excess labor" (i.e. the people who lost their jobs) have been increasingly pushed into the prison-industrial complex. When you turn on employed people into prisoners, you can charge them for their consumption and offset those charges on to taxpayers while suppliers, politicians, owners of private prisons, prison guard unions, etc make money off of every additional prisoner put into the system.
Essentially, anybody who gets put out of work by AI would more likely than not contribute to a similar system we're rather than have their quality of life improved, they will be forced into a less-than-ideal situation where their exploitation can be maximized while all the costs associated with a get pushed onto the remaining high-skill wage laborers. This doesn't necessarily mean everybody made unemployed through automation we'll get thrown in jail, although I'd be willing to bet a pretty sizable chunk of the poorer ones will be, but they will probably be thrown into a similar system where they can't actually enjoy the benefits of automation. Honestly, I think what you'll probably see people who are made unemployed through automation will likely get put into the prison-industrial complex, the military-industrial complex, or some other similar system where they are basically forced to consume products.
The problem is that fundamentally, corporations only have the incentive to extract profit. This can sometimes mean forcing people to consume goods and services through force (like in a prison or through imperialism), and so even though society might be progressing, ultimately they have to expand. Capitalism is based on the assumption of infinite growth, so they either have to destroy themselves the recession or war, or find a new market through imperialism.
It's an inherently self-destructive system that maintains itself through boom-bust cycles, where the average person's quality-of-life only marginally improved compared to those at the top, and suffering is exported to the people farthest at the bottom like in the developing world.
Came here for the topic, met Dan, and left with opened eyes and my feet comfortably on the ground. Thanks Adam, Dan, & their teams for the great conversation! (edited for grammar b/c it bothered me).
About food delivery apps, especially in rural areas: Here's a story I'm still annoyed about months later. I encountered a new nightmare where I used Pizza Hut's own app to order food but their system handed my delivery off to a Doordash driver. I'm certain the workers in the kitchen had my food prepped and in that dasher's hands pretty quickly, but then it lived in their car for over 90 min. I temp checked my food as soon as it was delivered and it was obviously below a safe temp for food that had at this point been out of the oven for almost 2 hours. They also have a call center that calls that don't pick up fast enough at the local store cover over to and they just place your order online for you. But if you have an issue they transfer you to the local store which then covers back over to the call center. So now I'm 2 apps deep and there's no way to get assistance. When they first started doing this they would at least text you a notification that they had given it to a dasher and provide their name, but that has stopped and so there's no accountability for this anonymous third party. I drove the mile and a half down the only highway through my small town to return the cold food for a refund because I wasn't going to make these poor workers have to add a rush remaking my order on top of their already busy plates because their corporate policies are to abuse these delivery apps instead of staffing properly. I left feedback on my experience with their corporate contact and basically got back a boiler plate response that yes they're handing overflow off to third parties due to staffing. I miss the early days of delivery apps when businesses wanted to disallow third party deliveries because they couldn't assure the quality would meet their expectations. Now they don't care because they can run delivery service without employing delivery drivers the same way Uber doesn't run cars and AirBnB doesn't own hotels.
This is such a pure example of the trust thermocline. Companies that automate responding to customer complaints WILL fail, sooner or later, and when it happens they won't even know why.
Part of the problem is that no business worth their salt has ever done delivery to rural areas. It's a bad idea, and in addition to not getting products delivered properly, I bet everyone lost money in the process. Let's have a bike route, ride in safely, sit at a table, say hi to someone in your community, and cut out the middle corporation completely.
Folding Ideas is my absolute favorite RUclips channel!
Me too!
Got a favorite? I never heard of this dude and wanted a good video to start with.
@@Praisethesunson That is a hard one! They are all very varied in topic! I think a fun one to start with if you like gaming would be "Why it's rude to suck at Warcraft". But honestly every single one of them is great lol
@@PraisethesunsonThe flat earth one.
@@PraisethesunsonI’ve been listening to his decentraland video a lot lately, but his nft video and flat earth video are both classics imo. I do also really like his older media analysis stuff, but I think he’s found an amazing niche recently with his anti-grift focus so I want to recommend those first.
I always appreciate Dan's "cynical optimism", I hope for more crossovers with Adam in the future!
He said that and I immediately said to myself "I'm stealing that!" In truth I'll probably credit him, especially since I'm a fan so it'll be easy for me to remember where I got it from.,
Dan Olson's RUclips channel "Folding Ideas" consistently provides thoughtful and important information and analysis.
I love how the conversation on the metaverse slowly pívots into 20 minutes of laughing at Mark Zuckerberg
Man, I could listen to these two talk forever. I want an entire Dan and Adam series now
dan lost a lot of credit with me with his completely uncited nft video. i was frankly surprised no one else called him out on his lack of citations on that one, but i guess no one really cared that much.
The way Dan calls Zuckerberg a loser is the most savage diss I've ever heard
As someone in the pagan community, I've seen a few pagans using AI like ChatGPT or Midjourney as a divinatory tool. So, they're very self-consciously doing the thing Dan says.
It sounds like using a magic 8 ball for similar purpose.
Sounds scary how people are letting corporate parrots worm their way into their spirituality.
@@thegeekclub8810 ah but unfortunately, they always do.
@@arturoaguilar6002 it's truly no different than using chicken bones or tea leaves. It's extrapolating "information" out of stochastic chaos, the point is the human interpretation not the input.
Also Ougi Boards.
Presumably because of my ADHD, a lot of times I will jump from thought A straight to thought J in a few seconds and in order to explain my reasoning I need to slow down and go back and try to construct my thought patterns in a sensible way as if my thought patterns were sensible in the first place, and it's a drag. But sometimes I will get into a conversation with someone else who is able to see my train from thought A to thought J without explanation (usually someone who is also ADHD and familiar with the relevent subjects) and respond with thought M and then the conversation is no longer serially processed but parallel processed and there's this certain energetic vibe to it. And that is the vibe I got from this video. Excellent work. I love it.
13:42 "It's a failure to think through the basics of what your own proposal entails"
That statement applies to so much viral information nowadays. Whether it is hype, a conspiracy theory, a rumor. It's like there is an epidemic of people just escalating and hyping up ideas without thinking about them.
Part of it I think is due to the social norms of the internet. We operate on clickbate, headlines and memes. And all you have to do to perpetuate an idea and advertise it to all of your friends and followers is click a button. So something with no substance can easily become the thing everyone is talking about.
the metaverse hype was legit hilarious. I worked in a call center in 2022 for a financial firm (ik gross, but i gotta eat), and the hype this got from top ppl and the company was hilarious. We're talking articles, meetings, increased investments. Had they never talked to someone about it under 22? It was just shitty, monetized VR chat for thousands lmao.
And that's the secret. Getting everyone to think it's a big new thing worthy of hype. That's how you get rich.
I do not worry too much about AI disrupting creative professions. I worry a lot more about MY profession.
When I entered the workforce, I was able to support myself pretty well, even without a roommate, working as clerical support (word processing, spreadsheet use, filing, etc). But tech completely disrupted that (particularly with the increased use of the Internet, autocorrect and predictive text) and since about Y2k, the only career path open to me has been customer service. It's a soul-crushing job that has taken its toll on my mental health, but I feel like the position I have right now is as good as it gets in that field, and I'm pretty happy right now.
But it's exactly the kind of thing that LLM is designed to do. Most of it is answering questions and following instructions from a particular knowledge base. The only thing the AI couldn't do is the human side, which only becomes critical when a member escalates. So 90% of my job could be done by a computer with a sophisticated enough verbal I/O matrix... what happens when that results in a 90% reduction in the workforce in my field? Sure *some* of us can be reskilled into other jobs, but if this were not an effective way to devalue labor, the ownership class wouldn't be so keen on it, would they?
This is the part the carriage-driver argument never acknowledges. Cars put carriage drivers out of good, skilled work and denoted them to one factory worker with a dozen others waiting behind to replace him
Automation is already killing so many jobs. I know this as I work(ed? - currently not working) in automating people's jobs. Not entirely. But step by step the "workload reducing" scripts I write & end user facing catalogs I cobbled together are setting the stage to cut out jobs... and no one even seems to blink. Cause those people don't really get the tech, just want the workload reduction. The thing: I've long left those companies. Many of my good solutions live on, years later, still having removed that part of someone's job. In a generation or two there will be enough little pieces like that in place to remove people. So no need to worry about language models per se. They're just one out of many pieces. It's the constellation of many small pieces we should freak out over. But each little step is too small to really see. AI won't take your job alone. Together with all the automation steps? Yeah. Get ready to be replaced in the near-ish future.
In a dane world the reduction of work load created by aitomatization would just make work days shorter. We don't need work 8hrs/day and yet some people still work 12hrs. It's ridiculous!
Because I'm a software architect, I normally see immediately through every one of these over-hyped software products. But I always find the sales pitches absolutely shocking and the fact that the "media" seems to accept them with no research at all most of the time and changed the way I think about TV (and online) news.
58:52 I worked for a tech startup that was trying to pass off the work of humans in South Asia as “AI”. A now (or soon-to-be) defunct “shopping app” called Nate. Their “AI” could barely perform its basic functions, so a bunch of humans had to step in and the credit would still go to the shopping robot (that barely worked) and founder of the company (who is an asshat).
If more of that happens, it will be up to consumers to call bullshit and employees to be whistleblowers.
I'm so glad Dan O's crypto video got so huge (>10mil views) cos it's an excellent piece (like everything he does). And his trilogy on "50 Shades"; he absolutely eviscerated it, from wonky fanfic to cobbled mess of a book to film 1 (which mercifully removed Anna's inner monologue - "double crap!") to the baffling sequel films, and so much more... Plus the guest voices reading out bits from the book, and the song. Oh, the song! (I genuinely love it!)
Speaking of messing with the voice commands, I had a houseful of friends who didn't take long to find "other" phrases. "Xbox, go to the bathroom" opened the home page, and "Xbox, smoke break?" made it pause the game. There were plenty of others, but it worked even when I tried it, so it wasn't just specific voice or accent or anything.
P.S. I was shown an advert recently here on RUclips for "Metaverse doing new & great things". "We've got legs!" Not trying to even take advantage of the way people were (reasonably) trashing the lack of legs, just literally faking excitement about having legs. Never mind how useless in the big picture the "feature" is. _There's no other word for it: cringe. It's cringe._
I should've mentioned Dan's channel, Folding Ideas. (I haven't finished the video yet.)
I see what you did there
This needs to be a recurring show, you guys are a good duo.
9:34 "It turns out that if we just create a flat page with a grid..."
What a lot of people tend to forget is that text and grids and lists are all man-made and made for a reason. The goal of technology is efficiency often providing convenience. If you're data gathering, like say, trying to find how much calories is in a serving of cereal, it's more efficient to pull up a list of cereals and its calories per serving than walking into a store, walking down the aisle, scan the shelf and find the cereal you're looking for, picking up the box, finding the nutrition information box and finding the calories per serving in said box. The HTML format was made to be able to provide a simple, condensed sheet of information for people. If you force it beyond that, you're going away from said efficiency and, hence, possible convenience.
Basically, imagine you buying a car. And then someone comes along and goes, "You know what cars lack? Tactile feedback. That's why our cars will force you to pedal a certain speed to maintain a certain higher speed. This technology will revolutionize the car industry!" The pedaling voids the point of the car.
Yeah, and like peak efficiency isn't always desirable -- it's fine to make something just because you think it'd be neat if it existed or because you want to -- but that's never how techbros and hype-runners frame these things. They almost invariably frame things like the metaverse as being innovations in efficiency, or depth of information, or in some way, quantifiably superior to the existing options, when in reality what they're pitching is almost always needlessly complex, less-efficient, and their desire to make it is based on "wouldn't it be neat" or "I could monetize this more easily," but they'll never admit as much.
Like just admit your desire to make a VR metaverse is based on the fact that you read "Snowcrash" and thought it was cool. It'll be way more respectable than the idiot attempts to convince everyone that uncanny VR stores are somehow a better way to commerce that we're actually getting.
@@fanboy50 No offense, but if you're starting off a reply to my comment with "...and like peak efficiency isn't always desirable ", you've missed the point of my comment. Like, I guarantee that if these techbros (well, they're not really tech, they're just bros) successfully make the metaverse for their own personal enjoyment, they'll get bored with it and move on because it doesn't provide them with anything they couldn't already do better, fast and with more enjoyment. As humans, as living creatures, we're always looking for the next more efficient thing. The question is, in what way is something efficient? Power? Speed? Delivery speed? Fun? Less clicks on a webpage? Conveyance of an idea? Etc.
What a great pair, I love their dynamic. Adam is energetic and snappy. Dan is more low key and has a dry wit. I wish they would do a podcast or show together, It would be so good.
"Gameplay only works because of constraints that are placed on the player." I'm saving this quote, Adam
The amount of money Zuck has lit on fire with the Metaverse is fucking depressing when you think about what else could be done with that money.
Also, Zuck was smart once, but power has been scientifically proven to completely break your brain.
Ironically, Meta put out a TON of great AI work during that timeperiod.... and released it to the public for free ;) Including some libraries that are like the backbone of a lot of AI projects (such as xformers). And most indie LLMs these days are now based off of Meta's LLaMA model.
@@karenrobertsdottir4101LIAR
I remember when i was a kid, we had a toy that was very much like ChatGPT. It was called the Magic 8-Ball. Wasn't as high tech, but you asked or questions, and you would get an answer.
To be fair, I think these tools have a lot more utility than a magic 8 ball, or crypto currency. But at the same time, they can do a lot more damage. Not because of 'Rogue AI' or what not. I'm much more worried about people doing dumb things with them because they think they're smarter than they are.
It's a tool, and the good or harm it will do depends on how one uses it. I have a table saw in my garage, for instance, but that saw is not a woodworker. Nor would it know to stop if my thumb was about to strike the blade (Yes, I know saw stop is a really brand if you have thousands of extra dollars XD)
I think the greatest example of Clark's Law and magical thinking at the moment is the belief not that AI will "transform" X Y or whatever, but that AI will "just keep improving." My suspicion is that, while current tech might not be its final form, Machine Learning Algorithms will eventually hit a wall and they repeatedly come up against the problem of conceptualization, because I don't think more powerful computers or larger data sets are going to be enough to get over that hill.
I think you're right. AI is a mockingbird device, it can replicate whatever you feed into it, very impressively so. But mockingbirds haven't turned into Skynet just because they're good at copying things.
It's probably because if you want to use human thinking as a base, you have to understand how humans think, and guess what? Turns out that's nowhere near as easy as it sounds: it goes so far beyond magical thinking that it isn't even funny.
2019: VR will change the world.
2020: Crypto will change the world.
2021: NFTs will change the world.
2022: The metaverse will change the world.
2023: AI will change the world.
2024: Wifi connected buttplugs will change the world.
Happy to tell you that we don’t have to wait until 2024
I now want a video game where you control an AI language model and have to try to get it to do increasingly complicated tasks.
I can't believe I just read an actual good use for an LLM other than a programming assistant.
Build off 'Creatures'.
YES. Fold THIS idea - Adam & Dan hype dehyping the "hype". Hip.
hip
hooray
Hip
hip
Two of my favorite creators having a conversation about an interesting topic, yes please.
I've noticed i have friends that have followed every hype, from MLMs, to crypto, and now AI 😢😂
Honestly yes
I can't wait to see the next crash in whatever the Tech Bros are drooling over.
I despised crypto.I mocked the Metaverse. But I adore AI.
Adam and Dan are on very similar wave lengths but their personalities couldn’t be more different (in a good way!) they have the making of a good podcast duo.
The college in my hometown had "robot food delivery" where it turns out the robots were just remotely driven by some underpaid worker, somewhere, possibly overseas. For me the weird thing was the robots weren't vandalized or covered in stickers
...yet
You're a legend Adam!
Keep spitting these truths to the people that need to hear them.
The duo that DISASSEMBLES grifts
Where can I sign up for the Adam and Dan podcast where they call shenanigans on crazy parts of society each week?
The problem that stuck out to me with the LLM video game NPCs is that you're putting all the onus on the player to tell the game's story.
The NPCs can say anything, but you have to know what to ask it to get it to say something interesting and tell the player a story thats meaningful. Replicating a life-like conversation system in a video game fails to consider that I engage in conversations IRL for a totally different reason than I do it in games.
Dan is the most brilliant individual I have seen on the internet. Thank you so much Adam for bringing him on, and keep up the good work :)
Dan Olsen has such a beautiful way with words. I can listen to him explaining things for hours.
So happy to see Dan on this show!
This is the crossover I had never considered existing. Now that it does I realize how perfect these two are with each other.
"Bobby Kotick is also an evil billionaire but he has the good sense to not go on camera." Unfortunately this prediction did not age well.
When talking about the race to the bottom, it's especially noteable that the big argument for mass produced garbage is that "it's good for the people who can't afford the nicer handmade sruff". Meaning that we're aware the reason people go along with it is because they don't have the money to choose otherwise.
The reason this has people so uneasy with LLMs especially, is because following that thought will turn *human interaction* into a luxury service. And whether or not the machines will live up to the hype, companies have already tipped their hands that this is an outcome they'll willingly push onto us if it becomes cost-effective
Companies won't hire "real" people even for cheap because they'll always seek profits. This means more people will be jobeless which means more people won't be able to afford real human interaction
I think capitalism might not survive automatization. How do you sell anything if nobody has s job?
Nothing has gotten more in the way of progress than money. Either a lack of it, or someone's unlimited supply of it.
You laugh watching Zuck or Elon or other billionaires burning billions, then weep remembering it's societal wealth they skimmed/scammed off all of us.
The first time I interacted with someone in VR, I was like, woah. This could be a game changer. Even with crude avatars, there was a physicality that you don't get from other forms of long-distance communication. Instead of talking to a disembodied voice or a small rectangular video feed, it felt like talking to a person. And this was just with hand and head movements. I saw a future where we got better tracking (face, eyes, fingers, legs), and better headsets (more comfortable, more convenient, cheaper, without requiring expensive hardware and external sensors), and this could become a common way to interact with people long-distance. Not the primary way, but another option, like how video calls are something we save for special calls.
I think that's why Zuckerberg went all in on the concept. He saw that too, also thought it might be the future, and wanted to get in on the ground floor to be the market leader when it exploded.
But in doing so, in tying the concept of social VR to boring office use cases and metaverse nonsense that ranged from embarrassing to scams, he basically ruined the public perception of the entire concept and defused that explosion before it could happen.
I think that's one thing I hate most about this tech hype stuff. There's often something cool at the center of it, like VR and generative AI. But then these tech guys get in there and ruin it by trying to turn it into some unethical profit machine.
Pretty much yeah. Like when I think about ChatGPT and how I interact with it, what it reminds me of is not Data, from Star Trek, but the SHIP'S COMPUTER from Star Trek. The Ship/Station computers are never treated as people or characters, because they are not, but they are capable of all sorts of things due to their enormous databases, including generating images such as Holograms!
And yet this never detracts from the 'human' characters or their own experiences/betterment. To me, that is the desirable outcome of this sort of tool.
It's easy to be cynical though, given how much of an abusive relationship we currently have with technology.
touch grass you be shoked
An aspect of AI hype that I don't see discussed much is that it's a sink for all the GPU compute that's no longer useful for crypto - which is why nVidia is so in on it
I didn't think of that but it makes so much sense. Same with all the VR stuff, I would imagine.
1:06:22 yes. People always wanted an affordable cab with an app. I’m really glad some cab companies are doing this
What a great crossover! Dan brings a lot of calm and collected energy when he speaks, as well as being articulate and witty, so it can be a real treat to hear him speak. You two played off of each other very well! I hope to see you do this again sometime!
The argument that Uber will not switch over to AI drivers so long as there are human drivers desperate enough to work for cheap is more an indictment of the social contract than it is evidence that a robot taxi service is not in our future.
I like that Dan distinguished between the Metaverse and Crypto hype phenomenons and the AI hype boom we're seeing in that the AI hype is actually based on a real-world useful set of tech that people are just blowing up into a new gold rush, because I think Adam has been doing his Adam schtick with AI where he just hyperbolically declares something total BS without leaving room for the rational kernel of counterpoint.
I think the WGA strike guy might be biased when downplaying the significance of LLMs.
@@MorgenWhite I can certainly understand the potential for bias in this regard.
But I've also been playing around quite a bit, prompting various versions of GPT to write stories, and I don't think I really agree. One thing that I've been noticing is that while it can generate convincing sounding content. When I reprompt it on the same kind of story, I get similar superlatives, similar language, similar structures, over and over again.
And not just in a way that rhymes but is almost outright word substitution for a previous generation. I don't write for money, but I do write for fun, and the forms are very very obvious in that regard.
The model is decently reliable at determining a coherent series of sentences. But I think there's a lot more human programmed structure going on under the hood which is obfuscated by the part the LLM can do on its own. Which is hard to really say for sure because the later iterations of ChatGPT are much more mum about the actual goings on inside their software.
But it also makes sense because one of the first user friendly applications of language models was sussing out the most common questions users will ask. Which means OpenAI has a ready made list of what things people are going to ask their language model and can implement guard rails as needed when testing shows the model doesn't give a coherent answer. A much simpler predecessor to these models is how your phones text prediction works.
It's hard to say. The tool is certainly very powerful. But like Adam mentions, very similar AI tools have been implemented all over for decades now. The only thing that makes ChatGPT so startling is that it's an application that is public facing and user friendly.
An example would be Bloomberg launching 'Bloomberg AI'. A Large Language Model that actual guts are just . . . the pre-existing Bloomberg Control Panel . . . They even went so far as to emphasize that Bloomberg AI is trained off of Bloomberg's industry leading data sets . . . which are the real prize of a machine learning model.
Thank god someone pointed this out. The entire discussion, Adam kept trying to group AI in with crypto and the metaverse, and you could tell that Dan was trying to take a much more nuanced approach.
It's also very strange that Adam seems to think that LLMs (large language models) are what the new AI wave is all about, and keeps trying to draw this line of logic that basically says "No, AI can't do [thing X] because AI is just LLMs and [thing X] has nothing to do with language / text." LLMs are just one of many uses of neural nets and (un/semi)supervised learning. The question isn't whether LLMs can do the things that Adam seems to think that people think they can do. The question is whether the TOOLS that built LLMs can do a whole fuckload of other stuff.
Yeah, there's maybe 10-20% of an actual product with current AI, but at least that's way above the 1% of a product that the metaverse is (for a random member of the public, at least), and the ... 0%? of blockchain tech.
Freelance writer chiming in. I understand where you're coming from.
I was mortified by ChatGPT's development, but decided to play around with it a bit and learn to use it at the very least. Yet, no matter how much I tried, I just couldn't get anything useful out of it.
I got to the point where I was still doing all of the research, writing and editing, but trying to pare down information or summarize details to speed up my writing process, using VERY carefully constructed prompts and I realized it just wasn't worth it.
Trying to use it to help me LITERALLY WRITE STUFF was actually just a waste of my time. And that's an LLM's prime directive, really. If they aren't good enough to help with that, what are they good for?
1:09:05 Speaking from experience, Dan just hit the nail on the head. The common advertisement in Uber Eats amd Doordash was "Up to $25/hour part-time!"
That was true, actually! If you managed to spend 50/60 minutes in an hour delivering, you'd make an average of $25 from about 3 deliveries.
But if you get stuck ***anywhere*** in the process, you watch your pay literally disappear with every passing second. I stopped delivering all Wingstop orders because I waited for 2 different orders for 45 minutes each. That turned my Pay-per-hour into $7, less than half of California's current minimum. The gamification of maximizing your time means it can be a slot machine on whether you're making $25/hour or $5.
"Word calculator " is the best description of chat gpt ive heard in a long time.
The joke Adam made about the name of the Lakers Arena has already played out here in Miami. The AAA Arena, the Heat’s stadium, was renamed the FTX arena. The name lasted less than a year, I believe 😭😭😭
Just googled this because I didn't know what AAA or FTX were and it is now called the Kasaya Arena.
Thanks Adam! We love you and appreciate someone willing to call out the BS- if I see another Joe Rogan video w/Elon Musk talking about how AI is scary good! I'm going to freakout- and everytime I get an argument with someone in those comments sections I ordered them to do their own independent research or refer them to your channel 💜
All the big innovations that the internet allowed for came from a time when competition existed in that space. Stagnation is all you can get out of the same 4 or 5 companies that control the infrastructure of the internet now
Even after, half the innovation out there is GNU levels of spite-inventing
Crossover of the century
As a PhD student who does research on experimental artificial intelligence algorithms, it makes me so happy to see other people see through the hype. Like it is SO DIFFICULT to get people in my own field to understand how much of it is a scam
Knowing the hype is fake actually makes me feel worse tho. People aren't going to lose their jobs to machines that do it better; they're going to lose their jobs to machines that do it *worse*
What a lot of tech bros miss when spinning these hype trains for technology is that in a lot of the sci-fi media they're pulling from, technology is so advanced that there are several steps before we get even close, if we even can. I mean, Star Trek's humanity is only able to live in a post-scarcity utopia because they did the scientific equivalent of inventing magic. The foundational principle of multiple technologies in Star Trek is the conversion of matter into energy and vice versa. I don't know if that's something that humans will ever be able to achieve and we're at the very least nowhere close to being able to do that. So until we can get to the underlying principles that make these technologies function, any attempts to just recreate them will be guaranteed to be sub-par to expectations.
Not just converting matter and energy, but the way energy was created was something so near-perfect in it's output that they essentially had perpetual motion. Star Trek tech isn't unfathomably complex stuff we just haven't invented yet, it requires us to break some of the most fundamental concepts of physics