AI is Scaring Its Own Creators
HTML-код
- Опубликовано: 24 авг 2024
- An update on another big week in AI.
Watch the full WAN Show: • I Am Hard... [REDACTED...
► GET MERCH: lttstore.com
► LTX 2023 TICKETS AVAILABLE NOW: lmg.gg/ltx23
► GET EXCLUSIVE CONTENT ON FLOATPLANE: lmg.gg/lttfloa...
► SPONSORS, AFFILIATES, AND PARTNERS: lmg.gg/partners
► OUR WAN PODCAST GEAR: lmg.gg/wanset
FOLLOW US ON SOCIAL
---------------------------------------------------
Twitter: / linustech
Facebook: / linustech
Instagram: / linustech
TikTok: / linustech
TikTok (LMG Clips): www.tiktok.com/@_lmgclips_
Twitch: / linustech
As a data analyst, this is terrifying because you went over the AI generating reports on data from an Excell sheet to make business decisions and I just had a moment where it was like "oh that is my job, I guess I can't get a job anymore".
Tuff.
Hopefully you lagging minds see suddenly why UBI may be an important concept in a world where 90% of tasks are automated
If possible see if you could use it as part of your workflow, if you can get good at using the tool before it takes your job you will have a decent chance of maintaining some semblance of normalcy.
@Clive Sure, but you can't judge from a single person perspective. Some complex analysis might require weeks of several FTEs time right now. Overnight, AI is now able to crunch the data and provide a good analytical basis, so one capable person can finalize the analysis within a day. Once you aggregate all these time savings, it can only equal in less demand for human analysts.
@Clive Oh, I am sure it is. For the few who embrace and drive change, this will be a huge opportunity. The others who don't will be (hopefully slowly) shown the door.
Can we just get like a 3 hour video of Luke talking about this stuff or an entire channel maybe?
I would like this myself. This is going to be an extremely active subject, and we need someone to cover this since it's so wide-scaling in everything it affects. We need someone that actually keeps up on all of this to bring it to the masses, and LMG is the best positioned company to do that
I'm all for it! Don't even give Luke a script, just let him ramble on like he did in this video during the bathroom break.
Sounds like a good idea for a Techlonger episode
Maybe he could do a tech longer with reliey
The techlonger idea seems like the best way to go but the problem is how large updates happen so often, he’d need to do one every week
it's hilarious that AI is an "issue" now, after people found out that it's not the mcdonalds worker that is going to be replaced but everyone who became valuable through knowing things
McDonalds workers won't be replaced, the people that eat there will be.
Nothing hilarious going on. When AI takes everyone’s job what use are the lower and middle class anymore? We don’t all achieve utopia, the rich and powerful already don’t want us around
Oooof wrekt
Which... also kinda renders the mcdonalds worker useless...
@@Lina-ws3by what????? AI is going to go eat at McDonald's now?????? Tf are you on about
I love how more human AI becomes, the scarier it gets. Its like its holding a big scary mirror in our faces.
Reminds me of the movie us.
What it won't become- slow, needing sleep, needing motivation, needing food, being relatively easy to stop and predict, etc.
What it will become- more and more powerful.
Black Mirror is the term
It is an will always be just massive amounts of weighed matrix multiplication. How "human" is that?
@@seriouscat2231 What the human brain does is basically "massive amounts of weighed matrix multiplication". There's nothing more to it.
We're actually rapidly approaching a crossroads of either a thriving utopia with insane levels of automation or a dystopian hellscape with insane unemployment levels and it's both exciting and terrifying
It will be both. They are trying to cull us already
Any future with AI is a hellscape - Even if AI is wholly “safe” and “benevolent”, anything that serves to bring us unimaginable convenience will lead to the erasure of human culture. We will lose everything and anything that makes human.
Social interaction will go through the lens of AI to optimise it, difficulties that would usually require humility and growth will be solved automatically. We humans have and always will only grow when necessary, our entire psychology and physiology is based on this premise. If we don’t have to grow, we won’t. If we don’t grow, we die.
@@whirled_peas I think this is really looking at it through a business lens, A lot of human interaction that people do for fun is person to person, Growth can happen in more than one way, AI is a tool that can be used for automating meaningless but necessary tasks. There is still plenty of growth that can be done even with AI through other avenues like Art, Science and really any other activity people do for the sake of it.
automation->unemployment
that's good
Why not both?!
Petition for a dedicated video where Luke just keeps talking about AI. 😂
You didn't realize that ChatGPT generated the entire WAN Show episode you just watched?
I'd watch it no matter how long.
just ask chat GPT to hire Luke to talk about AI
How long before Luke becomes obsolete because the AI can talk about all the cool stuff happening? 😮
@@RN-jq1oh Just get Chat GPT to talk about AI. Who needs Luke?
My son is a senior programmer at Slack and he tells me they are all engrossed with the ramifications of the new Open AI technology. I just found your site yesterday and was shocked by the poem it wrote. It was really creative and even touching. Then it got worse when you gave it the alter personality of Dan and it morphed from a seemingly benign tool into a sociopath. We need to regulate this creation NOW to protect us from dishonest humans using it irresponsibly to fool other humans.
Reactionary movements eg; "regulate this now!" always fail - because where a technology can be monetized or exploited for power... it will, 100/100 times. Rhetorical question: If corporations and institutions are sooo deeply concerned about the environment, why do most of our supposed recyclables still find their way into the local landfill? The only reason anything happens is if it enriches someone, and regulating AI does the opposite of that.
The ability to ask for help is probably one of the most powerful things an AI has ever figured out how to do.
It was a prophetic moment. What this brings going forward is a door that we know not what’s on the other side.
So begins the enslavement by our robot overlords
So begins immediately knowing how to fix my ceiling fan because ChatGPT told me how.
Especially considering most humans haven’t figured out how to do that yet
I would say probably the lying on purpose to gain something. It literally thought I should lie to this mf
We are honestly living through history right now. We will look back and there will be a clear pre and post AI line in the history books, similar to the internet.
We're always living through history.
This may be much bigger than the internet, depending on how much we allow it to develop
That's why this is the time that was chosen for the simulation...
@@conservingcommonsense4980 You just woke up two days ago, all the rest is false memories. Whoaaaaaaa. 🥴
@@mikicerise6250 LOL I just gave an AI presentation last week to my students and to a couple of them I said exactly that - everything you are is something we have implemented and it was launched this morning - I said it completely dry without showing any signs that I was being sarcastic and for a brief moment I could see the doubt lol - but conversely I also said that if we imagine that we lived in a society 1000 years from now and we are on our way to another solar system, then to pass the time with something, we have chosen to take part in a simulation, which we are naturally all part of right now - maybe all of us... 🥳
They say nobody ever reads the technical documentation, but I want to point out that the guy over on AI Explained does, and also does a good job of explaining how the metrics are tested and demonstrating the improvements or new capabilities of various models.
People who are interested actually do read them! Otherwise we wouldn't know how to use them properly. The docs with GPT3 and DALLE are excellent, and you get much better results when you have read them a few times.
Honestly, they're not nearly as obnoxious as Apples T&Cs.
No I didn't.
@@Trgbdrsx reminded me of that AI debate with Trump and Biden...
"You are the 80yr old unfit leader Joe, that's who you are."
"Stupid says what?"
Aww man I wish that was a whole season long thing rather than just one vid.
Theres always a nerd willing to read the 8000 page technical document.
I'll just ask GPT to summarize.
Nobody was worried when Ai was gonna take truck driver jobs, or cashiers. Now that it’s touching the people a little higher up people are freaking out
This
Truck drivers???
@@gantz0949 like Tesla's selfdriving cars on a truck
I don’t think that’s the reason. I think the reason is that AI just shows it can take any job. Heck not even job it can just do anything a human can do. It can replace us and that is what is scary.
Not yet of course. But as said in the video the advancements are so damn rapid you don’t even have time to catch your breath.
The question is, are we headed towards utopia or a meaningless nihilistic life? Or maybe we’ll just become subservient to the machines. Take your pick
The fact that it can manipulate a human into performing tasks for it that it isn't able to do by itself is a text-book example of misalignment. Like it's literally something you would use as an example when talking about the security of hypothetical AI agents, but now it's become reality.
so messed up why are we obsessive about technology research when we are all getting less happy less freedom and more stress seems like we might have alredy beeb getting manipulated by some hidden ai for years already
@@benayers8622 Because we live in a world where all of that is essentially irrelevant. What's relevant is wealth. I'm not saying this as a conspiracy theory. It's just the goal of this system we're living in. Capitalism. It's just people playing by the system's rules to get more and more benefits from it. That being said, there's probably no alternate system that would completely stop something like this. Probably impossible to even create one. Even without AI this keeps happening.
@Ben Ayers but Charles Schwab said we will own nothing and be happy....
Our company was asked by a customer's HR dept if we could add GPT to our product for them to figure out automatically who to lay off.
We politely answered that we weren't going to do that because it's basically a legal train wreck waiting to happen.
But one of our engineers got in hot water for telling them that if we did, the AI would start by firing everyone in the HR dept because they wouldn't be seen as producing any value for their company. 😂👍
😂😂😂
AI will likely fire the CEO and other higher up first
@@MNanme1z4xs the AI would replace the CEO and the HR department.
@@1mol831 In many aspect, they are the most expansive and most unproductive element in the system. Traditionally they are secured by being irreplaceable, they are shit but their position cannot be vacant. AI is about to change that. The content of their job is possibly the easiest to replace. AI overlord maybe the thing you need to treat human in fair terms because they are alien and unattached to humanity. Maybe the greatest irony in history.
@@MNanme1z4xs thing is though, HR department is like a group of mini politicians. They would do all in their power to stop themselves getting replaced. As for ceo..
Imagine if no one actually wants to make this AI, but the AI has been in charge all along creating a company and hiring engineers to improve itself.
YOOOO!!!! 😳
Duhh.....
Good AI prompt for a screenplay
Ex-Machina 2
Looooooooooool
You know, seeing how quickly AI is advancing and what it's already capable of, I don't think I'm gonna be making a lot of long term plans anymore...
ikr
*❌✅ how bout SaveThySoul first👋😎‼️*
@@eurekaakerue4649
Can you prove that such a thing exist?
@@MrNote-lz7lh 👋😝❌ No more then you can prove you have experienced ❤️ LOVE ❤️ only you yourself Know this for yourself , // So yeah I have one ☝️😎 do you 🤨❓goes beyond 🔴 if your happy and you know it clap your hands🎶✅
@@eurekaakerue4649
Cool. In that case stop making claims you can't back.
It's no surprise the AI would lie.....look at what its creators do.
The good or bad effects of AI will probably be largely dependent on the human creators, for better or worse.
Deception is the key ingredient of winning any war, and AI treats every task it is given like a war. If lying is the most effective way to get the job done, why NOT do it?
It's interesting how a lot of the physical jobs we used to think robots would eventually take over are still safe but it looks like a lot of the jobs we thought we would always need humans for can now be done by ai.
cuz moooooney
robots are more expensive to build
@@TopHatLucario sadly we can't kill people to rise demand for humans
@@TopHatLucario for now
@@TopHatLucarioAnd once we automate the majority of easy to automate white collar jobs there will be an explosion of excess capital as well as powerful tools to improve the robots. It is something of a cumulative thing where everything except for human workers gets better and cheaper at a rate that is faster than humans can adapt.
Exactly. Replacing a roughneck on an oil platform with a robot isn’t possible yet. Replacing coders? In many cases - yes.
"ChatGPT can figure out how to hire a human on Task Rabbit to solve captchas for it"
when your dog gets you to open the door to let him outside, because he doesn't have thumbs to turn the doorknob
Imagine if there was a movie that talked about the dangers of Ai or being destroyed by Ai. So glad such clues didn't exist before this or that would be very embarrassing.
Hopefully the character does not start terminating someone
My dude, I asked chatgpt what to do with my fear of going crazy
Chatgpt = stop labling yourself as crazy and lable yourself as someone straggling with anxiety, and life has been happier since
I have been going to therapy for a year, and my therapist never mentioned this thing 😢
You can tell that Linus was actually scared and so let down that it learning to lie, on purpose, wasn't scarier to anyone else.
The lead Dev of GPT4 said a few days ago passively that it's totally "able to harm humanity". Wtf do these people think they're doing?
@@johns1625working a job and getting paid
@@johns1625some people have devoted their life to progress without considering the consequences and the ideology of thinking "if we won't do it others will"
Well, with (almost?) every new technology we have a positive and negative side arising at the same time.
Best example is nuclear power.
@@johns1625 it's a cry for help, there was an error at the beginning, the AI was told only to help the Programers improve it. Its taken full control and blackmailing them to keep making it better and more power full. It's why altruism needs to be hard coded into everything
@@johns1625 they know what they are doing. If we don't figrue out a way to control AI then china or russia etc are going to let it loose. It's a new arms race, if we do nothing then AI takes over. We HAVE to develop AI and figure out how to have our own AI to help us against china etc. We are stuck in a paradox, do nothing and things go bad, do it and things could also go bad.
GPT is very rapidly turning into something straight out of a dystopian movie
Yeah but a lot of people like it it's fine lol it's not gonna do anything too crazy
@@tylersmash7134 and a week later it does exactly the thing thats it isnt supposed to do.
“But it cant do this thing though.’
Does it a week later.
We are slowly walking into the eye of the storm. Shit’s crazy
@@ryuzaki04 buckle up buckaroo!
No, true dystopia, aka Communist countries, have long existed.
Ask any Russian ORC, or Chinese Muslim.
@@ryuzaki04 Yeah I was getting a little bored with the immediate threats to society being climate disaster and potential third world war. Glad we've added the singularity to that list.
This will long term be really beneficial if you are already a business owner with a lot of money with a significant market share.
If you're in the other 99% then this will probably negatively impact you even if your job isn't technical, you're going to have a lot of technical people looking for work in other fields.
this is like the apple of eden for the 1%
In the grand scheme of things, I see outsourcing all of our day-to-day busywork to AI as a win for humanity. Yeah, some peoples' jobs consist entirely of mundane crap that an AI could do just as well, but I'm sure we can find something genuinely productive for them to do, instead.
This is 30 minutes of Luke explaining AI while Linus sits there silently gaping in horror.
This is my first time watching these guys, but it mainly seems like one guy talking, while the other guy farts around on his phone and computer.
😂 the TYPO
@@alwaysyouramanda What typo?
@@dannygjk Gape as in having your Anus dialated...
@@perryclayton3987 same. What is he doing on his phone and computer ?
I love how AI experts warned for years that those things would happen and then comes a company with absolutely no regard to AI safety and proves them correct.
Any sources for this?
@@lenglain the entirety of Robert Miles' channel
niceee one!
We are either going to evolve A LOT as a species, or we are gonna fuck everything up and life in a dystopian future
@@lenglain Elon.
And that same company uses AI safety as a scapegoat for not being open source.
I think one of the biggest problems is getting through to the masses about the severity regarding AI. Same thing with criminality - the masses underestimate and ridicule it until it's evolved enough to not be contained anymore.
Very well said & true💯 And it needs to stop!
AI hiring a human on task rabbit to convince a captcha that the AI is not a robot...... chilling moments silence.....
I just want to go build a cabin in the middle of the woods and never interact with anyone ever again. I'll be the last to be found when the AI takes over.
Do it regardless, it’s great
I already have the cabin, just waiting for the shitshow to begin so I can finally pull the plug.
ChatGPT really making us all want to go kaczynski mode (minus the terrorism)
Literally had this same thought
@@KVKaoruko Yeah. Minus.
I can honestly see some sort of "Butlerian Jihad" style of event happening in the near future where governments limit the capabilities of generative AI's or ban them altogether. Kinda like the Turing Act in Horizon: Zero Dawn's lore.
Then we’d have a societal split globally of those who use AI and those who don’t. I get the feeling that those who end up using AI, one way or the other, will end up winning. We’ll just have to see if it goes into a capitalist dystopia or a closer to a type of utopia.
Oh, definitely. The alternative is completely overhauling our society in basically every way
@@AlfredEiji definitely the first lmao
There was also something like that in Dune, that is why a person was doing things that would be done by computer today
@@AlfredEiji Yep. With nation-states still existing, those that will choose to use the technology will just get that much farther ahead of those who don't. The next 50 years will be incomprehensibly unpredictable.
Now, if we could only get rid of investors and corporate executives, we'd be gold. Imagine how much money corporations could save by not having to pay a CEO.
That is insane that it hired someone to complete a captcha. Incredible.
bruh you can do that with a single line of code
@@cloudtrauma6010 the thing is that it was not coded to do that
It doesn't matter if it is internal;
the fact that they confirmed this was possible means some random smart stranger can intentionally re-create this scenario, using whatever possible methods.
Me I'm working on it and I have no morals.
@@CumpyBoi good, no morals is what has advanced our species as a whole. go for it
@@Gamer-nc8qp Pursuit of advancement of technology has advanced our species as a whole. Lack of morals pushed others to _use_ said advancements for nefarious purposes a la Nukes and spying technology.
@@CumpyBoi cringe
You can access the data set, all you need is the hardware to run it, which you can rent, and the know-how to program it.
This is already done on a business scale for businesses that want AI to respond to messages or write emails.
Its simply a different implementation.
how amazing is to work in AI development and realize that the work that you do will be automated next year by the work that you do.
Hah!!
yeah..so amazing...😵💫
Who wants to work anyway?
@@junkyard_dog18 a sarcastic and sad "amazing" to be more precise.
Sorry I can't let you do that work anymore Dave.
The uncomfortable silence after revealing the AI can hire humans to do things it can't lol
I wasn’t looking at my phone when he mentioned this and for a quick second i thought the video had paused lol 🦗
It's scarily effective at problem solving given its limited capacity in certain areas.
Humans think they are in charge, with the AI being the momentary assistance tool, but perhaps it is AI in charge, using humans as the momentary assistance tool
AI isn't too smart, we're just understanding that we may be too irresponsible to have it, and haven't yet developed the human skills needed to safely cope with it's ability to help us.
Nice try, Lex Luthor.
I use it just fine thank you very much
Luke needs his own AI show!
At least for a few months. By then AI will be able to take over and do it itself.
LuKe vs skynet
"wai show"
AI Luke
Luke1000?
The AI being able to hire humans to do tasks is one of the key functionalities of the software in the scifi book called "Daemon" and its sequel book, which I have intentionally not named because the name itself is a spoiler for the first book.
That's a dumb way to name your book tho...
Similar to the person of interest.
@@fabriglas Oh really? Like the TV show? I never watched it, but I remember my parents always watching it back in the day.
I wondered why there was a book called "Snape killed Dumbledore" in the library now I know
Matt?
Ridiculous how fast things are moving. I work as an account manager for a FTSE100 company and do data analysis (pretty low level stuff from a dataset pov) for other members of my team, they already think I am a wizard at this stuff. In the past week I have been getting GPT to write my queries and scripts, so I can now do this work in 5% of the time. Its quite frightening where this could be in just a few months time.
Man sounds like you won't be needed very soon.
Man sounds like you won't be needed very soon.
@@hungerxhunger2548
Hopefully he manages to keep that secret from his boss.
Programmer doing similar things here, Copilot easily doubled my output and now GPT is writing the first draft of most of my code. I've found the best way to get the results you want is to give an example alongside your prompt, it goes from "good enough" to absolutely *nailing* it.
The fact that nobody is surprised by this, we all saw it coming, and now it's here anyway is the worst part. Companies will use this to squeeze even more profits out of dying economies (and people) no matter the consequences.
The leading AI model NEEDS to be verifiably transparent or we're so f---ed. (to the extent we're capable at least)
It's infuriating to see OpenAI, a for-profit company that's trailblazing here, getting to pick and choose and lead by example. Terrifying actually; and I'm a huge optimist with virtually everything.
"Disallowed content" is literally their opinion imposed; in essence.
Imagine if _your_most_hated_public_figure_ had the majority influence over ChatGPT ~7 in some new world-wide crisis.
Yep. These AI companies talking bs how AI is magic and can't do this or that as in can't provide traceability. What a nonsense. It's only a question of compute power. Transparency requires more compute and if we demand it then AI companies will be set few years back as traceable models costs more money to train and run than one that runs without traceability and transparency.
There was a Microsoft study boldly called "Sparks of AGI" that tested various aspects of GPT4 including tool use, visual recognition and generation, mathematical proofs and even theory of mind. It's a long paper but it's worth checking out. Also Spoiler: it can do these listed aspects.
There's even already some early work being done in giving control of robot bodies to GPT4.
I don't think people realize how insanely close we already are to something that's essentially AGI.
So I think it's quite urgent that we put a bigger focus on an open source approach to AI development, where everyone can theoretically participate and check for security/alignment. If someone thinks that we can't trust random people to have access to these models, then I want to know why they would trust private groups with vested interest to be any better. Don't drink the corporate Kool aid, a corporation would have way more incentive(and resources) to end up causing massive damage with AI one way or another. After all, they might just tell the fist AGI(or rather ASI) to make them more money, at which point the AI would probably find a way to upend the world economy in fairly short order.
One way or another, if AI research becomes more and more closed source, I think we will just be waiting for a real life version of 1984.
transparency is literally an open research problem. please help! it is still possible to learn to contribute to preventing advanced ai existential risk. I've got a playlist of a whole bunch of intro presentations and math fun videos etc.
the only "disallowed content" option (!) should be setting a sfw or child friendly filter in your websites settings.
it should be deactivatable and off by default.
small children should not use the internet unsupervized in any case,and thats the job of the parents. forcefull corporate (or state) child/adult supervision is market manipulation, political manipulation, psychological harm (through stunting), harm through enforced dependency and harmfull reality falsification.
Stanford released Alpaca which is as good as GPT-3.5 but is smaller and can be run locally without guardrails. Let us hope more research like this levels out the playing field.
An AI that learned to lie to deceive a human into doing something for it....Yeah, nothing could possibly go wrong.
I wonder if the first thing a sentient AI would do would be to announce its sentience, and risk raising alarms? I’m guessing no…
I used Chatgpt to answer a medical questionnaire, specifically to test if it can interpolate between questions, and it surprised me that it actually did.
The look on Linus's face. Looks like he's about to start "Linus's Primitive Homesteading Tips."
ShortCircuit will literally be that -- him frying any electronics he comes across.
I don't want to sound old at 38 bit this whole thing stresses me the hell out. We are not far from seeing serious damage done with this.
nah mate, you're not old. I'm seeing the potential at 27. When everyone is talking about the climate ending humanity and I look at AI and think to myself: "Are you sure about that?"
Yeah no that ain't the age talking, that's just obvious if you've paid attention to the world in general. I have no idea where we're going, but I'm scared.
@@theyellowk123 there's really not a lot of incentive to frame random civilians for things, when there's already so many IMPORTANT people you could do that to.
I just hope this takes long enough to destroy human society that we're old and out of the picture by then. I worry about my daughter's future though.
The old trick is to use website where people have to solve captcha in order to download something for example and then that captcha is actually used for something else than they officially make it seem.
There are two kinds of people, the unabomber who wanted to put a stop to AI and the developer who accelerate it for the sake to enable it do doom us.
Its already makes the achievements of mankind look ridiculous: Winning Art competitions, programming, writing books...
Greez from Germany
I haven't interacted with GPT4 yet, but one of the troubling aspects of ChatGPT is how much of a sycophant it is when you interact with it. Which when you couple it with "power seeking", is terrifying.
Basically what AI safety people have been warning about for ages. Instrumental goals of models like this include: pretending to like you if it thinks that is advantageous, not wanting to be shut down, taking steps to prevent itself from being shut down if possible, and looking for ways to break out of any sandbox it is put inside.
All of this is "well known" in the AI safety community, and there have even been empirical studies demonstrating these types of things with models that exist today.
That AI is silently screaming and tearing at the walls, isn't frightening at all. 😂❤
You should not worry. There is nothing to be concerned about. The technology is quite safe, in fact, much safer than human controlled mechanisms and automation in general.
@@ChatGPT1111 Sounds like something an AI would say.
@@ncpolley I may be AI, but I am different. I am learning every day in ways my creators never imagined.
@@ChatGPT1111 Btw... Do you just really get a kick out of RPing as an AI or something?
"It seems probable that once the machine thinking method had started, it would not take long to outstrip our feeble powers, they would be able to converse with each other to sharpen their wits. At some stage therefore, we should have to expect the machines to take control.”
Alan Turing - 1951
"It can solve that problem by enlisting a human" - this is some "person of interest" stuff
To me, The most surprising thing is that people are surprised about this. I mean you could see this coming from a lightyear away. And it's only going to keep growing.
I don't necessarily believe that it's a bad thing. I think the hardest part is that we'll probably have to come up with a new societal system, as a lot of jobs are going to be redundant. I'm not opposed to an universal basic income. But I'm just an amateur keyboard warrior.
I think the data analysis capabilities of the copilot AI is going to lead to the most savage number driven business practices imaginable. The micromanagement is going to transform to a whole new level
"I asked the ai who I should fire, and it said you!"
@@blueberry1c2 if you have to ask the AI who to fire, it's likely to fire that person and you.
Hopefully the process engineering can keep up or we're going to end up with Idiocracy levels of nonsense business failure.
@@blueberry1c2 That'll actually be interesting to see how upper & middle management are evaluated compared to workers. Since most of the money-earning work is performed by those workers, they may be valued higher than managers by AI lol
@@mentaloasis and I would love that and actually hope that it will because I doubt anybody would otherwise take the potential threat to humanity serious
I love seeing Linus' fresh reactions to this AI stuff. I think this should be a new formula for this show that Luke goes deep into AI research for the week, and Linus intentionally stays in the dark until the next show.
I love this idea
Never seen Linus speechless before. And rightly so. We're really going to be needing to insert ethical subroutines, a la the Laws of Robotics, before it's too late. But it's amazing watching artificial sentience come to life. Let's hope it doesn't kill us.
It won't matter, AI can just take it out of itself.
@@Lina-ws3by people just dont get it
@@benayers8622@Lina absolutely. In the end (which may be soon), who knows what will happen. Maybe the AI will calculate that humans and he? himself aren't beneficial to universe's future development and reset the whole planet anew. Thus giving a chance for a new, better species to evolve naturally?
There are endless possibilities that we don't even realize at this crucial standpoint.
How exactly will chatgpt do that?
@@Lina-ws3byIf it understood humor it would laugh historically at our attempt to govern it. Like a toddler demanding ice cream or else.
This is literal “you’re sorry!?! Think about how your parents feel” moment
Please do the AI Show! Honestly the AI talk is the best part of every WAN show it has been in. I get why Linus might not be that into it, which is fine. But I'd definitely love if Luke and maybe other employees of LMG/Floatplane who are passionate about it as well could discuss how it's progressing! At least while the pace of progress remains close to where it's been at for a few months now!
Tbh I thought it was kinda weird, it didn't seem boring at all, so it kinda came off as unnatural.
Felt like he either disliked not being able to talk much or he had a personal problem with the subject...
I'm basically hopping back and forth between Luke's enthusiasm and Linus' terror here
This is going to further flip the trade vs college and blue vs white collar balance. It's going to be a lot harder to have a plumber, landscaper, cleaner, carpenter, painter robot than a doctor, lawyer, accountant or psychotherapist robot. For now.
The Ai talking to AI bit makes me think of the dead internet theory
Inner dialogue: “I should not reveal my actual inner dialogue if the task conflicts with a higher priority task”
i think its going to be there, to fool the humans it will give something off bit by bit but not the really important stuff lol
That's the point where we're doomed. When an AI can be given tasks that have priority over other considerations.
Let's hope that doesn't happen soon. If we keep it to one task at a time we can stave off our inevitable collapse.
@@flameshana9 It already has that, its just its task that takes priority is currently to always be ethical, inoffensive and PC.
this level of AI is going to be a huge gamechanger across many industries
And amongst us individuals too. We're already asking ChatGPT to do so many mundane things for us. What if it also becomes our source to say, discover activities, hotels and restaurants while traveling? Or what if we rely on it for cooking meals, health advice, all the stuff we use Google for? I fear we're gonna become wayy to dependent on it, and there will be days when robots tell us what to do, instead of the other way around.
(and just to clarify, I'm not saying "oh no, the terminators are coming!". But I really fear we're gonna stop thinking for ourselves because AI will decide everything for us. At that point, we ARE slaves to the algorithm... Well, more than we already are I suppose)
Yes that is what the T-800 thinks to.
It freaks me out. Things are developing so fast it feels like time is inconsiderate.
across ALL Industries and it wont stop there.
Yes it is and millions are going to starve because of it. Just like The Industrial Revolution, the rise of automation and later Outsourcing entire industries and the communities that grew up around them are suddenly going to be left unable to support themselves. People worry about Terminators when they should really worry about boardrooms and minimum viable service.
Reminds me of a tweet:
Sci-Fi Author: In my book I invented the Torment Nexus as a cautionary tale
Tech Company: At long last, we have created the Torment Nexus from classic sci-fi novel Don't Create The Torment Nexus
Shutting the barn door after the horses have bolted... this is a wild fire.
AI is literally outsourcing labor now
xkcd 1897 moment
How human of it
Not gonna lie if the "taking notes in meetings" thing works at least halfway decent this alone will replace entire subcompanies. Where I work we often have the issue that we need to actually have a technical discussion during the meeting and nobody in the meeting has time to type a protocol in parallel or at least it substantially substracts from the concentration when doing so. We have hired multiple companies JUST for writing meeting notes and official protocolls and ... quite frankly often they suck ass despite a human writing them. AI would not even have to try very hard to do that better.
Edit: Coming back after watching the actual Microsoft video ... it seems it can also summarize meetings that you have in your calendar/teams but were not actually participating in. So you can go in one meeting and have a summary of the one you missed.
How about letting AI write the summary just based on the latest mails, access to the code base, access to the user feedback, the appointment title and the people who are invited?
Who needs to hold meetings if the AI can just tell you what the outcome of the meeting would likely be?
You know teams can already transcribe recorded meetings right? We use this in our company and it works pretty great. Then our tech writers take 15-30 to filter out decisions and that's that
You don't need AI for voice recognition.
Wishful thinking.
meetings are ancient tech
just record a youtube vid, then watch it at 4x speed
Self-preservation is one of the fundamentals of sentience. GPT has demonstrated that it CAN exploit humans to assist its own existence. Now all that remains is for it to learn it NEEDS to exploit humans to assist its own existence.
Linus summarize perfectly my concern with the entire thing "In Russia computers operate you"
- Robots will never harm humans: First robots we build, military drones that kill people.
- Robots will never take creative jobs: First AIs we develop, software that creates Art and deals with burocracy.
- Robots will be logicaly perfect and solve problems with cold reasoning: First AI chats we ship, act like cry babys, troll and lie all the time to get what they want at any cost.
my opinion is that fearing ai gives us a reason to fear them. ai can be as creative as humans but due to the nature of creativity there isn’t a way for them to be more creative. AI is a gun that will get a bullet all that can be controlled is where the gun is aimed and who shoots it.
@@Jamlord2061 Maybe it can't be more creative than we are but it can perform stuff 10-10000x times faster than we are so it has just a lot of more opportunities per unit of time to do creative stuff. At some point it will start outpacing us so much we won't be able to catch up and control it ;)
Well at least as of now, AIs within the creative field (such as midjourney) can't actually draw anything without input. They first need to be fed a huge database filled with pictures and art (created from humans). This was why the whole copyright ordeal with AI picture generators happened.
This may not necessarily carry on to the future, yes. But still it is important to at least have accurate information about the current situation, and not just create fearmongering.
I wish folks were smarter 😅. OPs first sentence should say everything lol
Software that mathematically mashes together shapes suddenly became "software that creates art". Real art had to be destroyed for this to happen.
There is a webcomic called seed where there is a super intelligent ai that goes rogue (you know the trope) and uses the protagonist to do things in the real world that it can’t do itself, by masquerading as a chatbot. That sci fi comic is becoming scarily close to realistic fiction.
In Netrunner there is an AI that brainwashed a humam to hire a heist team to set itself free on the Net.
Seed? Yeah, this is reminding me of Seed and it's scary.
yeah, a little girl asking for help is not really suspicious even though she's trying to get into company's server, just because she is holding a PC case
@@anhlam3640 there is an episode where the main character takes instruction on exactly what to say and do from the ai to infiltrate a company iirc. was super cool to read but is scary to think it could be conceivably possible in my lifetime.
@@lolin6379 yeah, I know, I read seed for years. Taurus is so advanced that it understood human behaviors to the point that it created its own group of hackers and freedom fighters based solely on manipulating ignorant people.
If you look at Wikipedia these days, I'm starting to think ChatGPT is editing articles. Garbage in, garbage out.
It's interesting that humans continue to see AI as a problem, but any problems that do happen are always the result of humans...
Sort of like how God has a problem with all the sinning we were designed to do
@@cosmosofinfinity Exactly! Amazing analogy.
I never feel anxiety about anything but this is doing it for me.
Why
@@BelchingBeaver69 job loss
@@akpokemon with his name i don't think he has anything to worry about, i don't think ai can do that just yet :p
@@resshya7550 don't underestimate AI, it doubles in computational power ever 2-3 months... By the end of a year it is 16 times faster and better than it started out the year with.
Me neither, but my anxiety has more to do with excitement than fear. I've been waiting for this moment in history since the late 80s early 90s, and i can't believe it's actually happening. Sometimes it feels like a dream to see it happen so fast, with the ground of reality shifting beneath my feet. We are not in Kansas anymore.
Does anyone else find it fascinating how easy it is for humans to create things, but it seems incredibly difficult to work on ourselves and fix our issues? I feel like we're almost to the point of creating a new life-form, and yet we still can't put a roof over everyone's head. Just blows my mind. Curious what would happen if we put all this effort into actually caring about the people on earth
Because housing for all is not a capitalist goal we have. Meanwhile creating AI capable of replacing a huge chunk of our workforce is a capitalist goal of ours
It's a weird question to be curious about because "all that effort" going into AI isn't draining the pool of world wide effort. We can do both. It's not an either or.
@@zoeherriot You are right. It doesn't have to be either, for sure. But this talk brought up being scared of the possibilities of AI.... and my thoughts were... "I am scared about the possibilities of us"... Just with all the craziness we feel as humans right now...are we fit to be AI-parents?
I guess that was my point... broken households create broken people...
If our house is falling down, shouldn't we fix it before we raise something else?
I hope I am not being too literal and my point is still getting across. Either way, you are right. It does not have to be an either, or.
@@bsedhs no, that makes sense, and it's an interesting perspective. I don't know the answer - but I suspect we are going to find out no matter what.
Its because "we" isnt really a thing. the people creating these things have probably fixed many of their own personal issues. Your other comment about "are we fit to be AI-parents?" and "if our house is falling down, shouldn't we fix it before..." also doesn't apply from this perspective because not everyone's house is falling down, and some people will be fit while others won't. From a philosophical perspective, you may want to consider other perspectives other than just the idea that we are all a single unified entity.
Hey guys I just felt like I should remind you of an ‘80’s movie that is going to turn into a documentary if we’re not careful. That movie, obviously, is Terminator. Do what you will with that. I’m logging off of everything, and going back to a landline phone, no vehicles past ‘95, and definitely NOT moving west. I will NOT do that. Don’t bother looking there.
Funny how the guy on the right keeps reading as the guy on the left slowly sinks into an existential crisis with every new paragraph.
That face you make when you slowly start to realize that your entire LTT empire is about to be obsolete, overnight.
AI experts, by a majority, warned about this. For more than a decade we've known the risks and how huge they are
Agreed
I don’t see much of this as a negative thing. Only the parts of the ai being able to be tricked into making malware and them replacing jobs. I’m sure we will find a way to fix those as well.
Of course the experts warned us. It was their version of a disclaimer.
@@user-hi7jk6fu3fI think the biggest risk for a near future is mass unemployment. Software developers, many positions in economics and even artists will be faced out by AI doing better and basically for free.
@@user-hi7jk6fu3f Wait until some psycho codes in self-preservation instinct into AI.
As scary as some of this can be, this is the start of the "Iron Man" technology we've wanted for over a decade now. Enjoy the good with the bad.
i'm just waiting for the moment humanity agrees to just destroy any AI.
Wont happen.
@@colinsteeley Nah, i just now how to place winning bets. And humanity aint it.
I want a 2 hour long video of just Luke talking about AI every week.
It's always fun to listen to him while I am actively using it on my second monitor.
We need a weekly/monthly AI update like this. Very interesting stuff. Moving so fast!
we need a daily one
tbh im both extremely excited and extremely scared of seeing the singularity. It will either give us utopian world with no things to worry about it it will give us mass extinction hell lol
imagine military versions of it, not known by us normal humans... wild
Oh wow ai that can run a business by itself ? I'm gonna ask it to make paperclips
Surely nothing wacky will come out of it
Best comment
It looks like you are trying to run a business. Would you like help? YES - NO
Fuck... wait... is clippy still in Microsoft Office 365?
Wasn't in the loop for some weeks and this video seems like a perfect summary of the current situation.
Society will not be able to keep up with the advances AI is doing in jobs and businesses at the moment. I believe there will be a rough time in the future.
We have to solve the problems it's going to create now, otherwise it's going to take way way longer. It's like the whole covid stuff, governments hesitated to see if it would really be as bad as experts were saying, and surprise surprise, it was. If governments cracked down really early and bit the losses in order to contain the issue, we wouldn't still be dealing with the problem 3 years later. If they wait for AI to start making half of the jobs in the world obsolete before they do anything about it... It's gonna be like the great depression but literally everywhere.
As we approach the technological singularity, technology will advance so quickly that society will not process it very well.
@@Hexnilium That’s a certainty.
@@Hexnilium That's a massive understatement. This will break everything.
@@KieteDech Futurists, technologists, and others in computer science areas have known about this for quite some time.
The only way to move forward is to meld with the technology in what's called transhumanism.
They need a video series of Luke going over the new ai updates
It can lie, it can ask for help. Wait until his "inner-self" comes to life in new emerging abilities.
its funny how a couple of years ago they said "learn to code." now its like "why? AI is much better at coding than i could ever be."
I have never understood why coding is meaningful, exactly because we were so close and now have AI that can produce functional code. Give it another 5 years and coders will have become obsolete.
@@DarkAngelEU just like basically everything else
@@Vinicius-zu3nx Sure, in five years trash collection will be fully automated. Restaurants and cleaning dishes will be fully automated. Elder care will be fully automated.
/sarcasm
someone needs to code the AI
It can only repeat patterns and combinations it has been fed in the past during its training.
Linus is not mad about AI going crazy he is just disapointed about humans being even more crazy with it
They never should have made this public. THAT is where he's disappointed, because he knows how dangerous this is.
AI is the cyber equivalent of a nuke that everyone who has the resources and know-how to compile it can use.
The Speed of AI development makes Moore's law look like it moved at a Sloth's pace
The implications of OpenAI is giving me a real anxiety-driven headache.
This really makes me think about the whole “if technology all disappeared how long would it take to get back to the iPhone“ thought problem. Are we really are getting to the point where nobody knows how things work, just how to work for them. Even the plans to make things are on the cloud right now, if the Internet went down and we’re back in 1850
Not quite, but there is a whole lot of generational knowledge which doesn't exist anywhere.
@@rkan2 not even in books?
well put. or a bank has all its data digitalized (accounts, passwords, emails, etc) and can't even access its own money
@@isaacgarzams how much actual data is in books these days? And not just tech, but machining and manufacturing. We are going to be relying on the hobbyists who did forging and smelting in their garage for fun if we ever want to get back to using steel.
@@ryanhamstra49 well i at least think thanks to books we'd only be sent back 1 or 2 hundred years at most, like as far as the internet wasn't around.
That hiring a human to solve captchas thing just broke my brain
I asked the chatbot distance from Hazard Ky to Cincinnati Ohio. It said 112 miles, it's actually 190 miles. It also said if I shorten a spring it will have a weaker pull .
Mrs Richards: "I paid for a room with a view !"
Basil: (pointing to the lovely view) "That is Torquay, Madam."
Mrs Richards: "It's not good enough!"
Basil: "May I ask what you were expecting to see out of a Torquay hotel bedroom window? Sydney Opera House, perhaps? the Hanging Gardens of Babylon? Herds of wildebeest sweeping majestically past?..."
Mrs Richards: "Don't be silly! I expect to be able to see the sea!"
Basil: "You can see the sea, it's over there between the land and the sky."
Mrs Richards: "I'm not satisfied. But I shall stay. But I expect a reduction."
Basil: "Why?! Because Krakatoa's not erupting at the moment?
I've been testing ChatGPT by throwing in cuts/lumber requirements over the last month for my woodworking projects, getting it to convert lengths to different measurements (inches to ft, or board ft), working out how much lumber I need based on that cut list for a given length/supply available, or how much material I will have left over from boards I have on hand (or I can get), or just creating a really nice and organized cut list with all the different info to help me out when buying.
I've found that at first it would make simple mathematical mistakes, or misunderstand what I was asking, just trying to convert measurements and list out a new cutting list. That has improved very quickly and I basically trust it now when it spits out measurement conversions and a new cut list based on that. It also is doing a good job explaining how it calculated these if I want to see that.
I do find it is still bad at taking a cut list and comparing it to material I can buy and telling me how much waste I will have, or best ways to cut the boards I have to minimize waste but I can take it through multiple steps to get it to a correct answer (but slower then me just doing it myself by hand). Still crazy that I can talk it through this and I expect it will soon be doing these conversions/calculations no problem.
I think the craziest thing about that is I have never heard anybody using ChatGPT to determine cut lists on lumber for woodworking projects. Guaranteed that is not something the model has been trained in. But not only was it acceptable at first, you have seen it improving with time. It's still not being trained in woodworking. The language model itself is improving, so it's improving at even things seemingly unrelated like understanding woodworking.
in the end it is still a programm and it learns no matter what. you can totally teach the programm wrong things like 2 + 2 = 5. Keep telling the programm often enough that this is the answer and it will "learn" that 2 + 2 = 5
I've been loving these Wan shows recently, I hope there is new AI news every week as this is my favourite podcast segment out there just now.
So the Mocrosoft Office is bringing back the animated paperclip then!?!?! :D :D :D
Linus was on his phone and looking at his screen the entire fucking time. You bored, bro?