Sam Altman Reveals EVEN MORE About GPT-5! (Sam Altmans New Interview)
HTML-код
- Опубликовано: 1 май 2024
- Sam Altman Reveals EVEN MORE About GPT-5! (Sam Altmans New Interview)
00:24 Project Stargate
03:01 Compute
05:30 Shocking Statement
09:09 No More Surprises
13:47 Winner Takes ALL
16:20 agi in 2030
18:21 Actual AGI Timeline
21:29 GPT5 and GPT6
25:04 Responsible AGI
How To Not Be Replaced By AGI • Life After AGI How To ...
Stay Up To Date With AI Job Market - / @theaigrideconomics
AI Tutorials - / @theaigridtutorials
🐤 Follow Me on Twitter / theaigrid
🌐 Checkout My website - theaigrid.com/
00:24 Project Stargate
03:01 Compute
05:30 Shocking Statement
09:09 No More Surprises
13:47 Winner Takes ALL
16:20 agi in 2030
18:21 Actual AGI Timeline
21:29 GPT5 and GPT6
25:04 Responsible AGI
Links From Todays Video:
• The Possibilities of A...
Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos.
Was there anything i missed?
(For Business Enquiries) contact@theaigrid.com
#LLM #Largelanguagemodel #chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#DataScience 
Наука
He pretty much said it out loud, that they're sitting on SOOOOO much more but are only releasing it gradually to let people get used to it, i.e. "don't spook the smoothbrains".
i.e. milk as much money as you can
"don't spook the smoothbrains".? Let people get used to it? I think that statement is ridiculous. They act like everyone is going to go jump off a bridge just because they witness something shocking or amazing. This is not the 1800s. People were mainly shocked at ChatGPT in general when it first hit last year; no one jumped off of any bridges, and if they did, they already had mental issues. I think humanity would be fine, even if the most advanced AGI came out today. We are not cotton candy in a rainstorm. They are overly cautious, and being that cautious, they are gonna fall behind.
💯 he's said it 10 different ways b4, but I guess people thought he was vague.
I guess they're close to achieve ASI, and singularity is coming within 3 years.
trust me you can't spook us smoothbrains with advanced tech like that, we already have trouble understanding the Earth is round
No BIG surprises in models leaves room for competitor to catch up and release a better product.
We all just want the best product
The others really need to catch up and force their hand as they are already staring to drip feed like APPLE do
What product? What would you use it for? All they are willing to release is dumbed down versions that only a dummy would benefit from.
Calling GPT-4 the dumbest model when it’s pretty the state of the art model right now speaks volumes to what they’re sitting on! I can’t wait for June and for their Dev Day!
Or he's tech bro talk for hyping gpt 5 up. Who knows, theres obviously bias in how he talks and what he presents.
He at least succeeded to make people think they already have AGI internally, people are so overhyped, I suggest you to daily read the recent research papers in the aim to avoid "hallucinations" about what is really going on.
@@TheRealUsernameis there a specific source you use for the research papers?
@@gabemulero3962 MarkTechPost and ArXiv
@@gabemulero3962 MarkTechPost
When he says “we are probably looking at it through a lens that no one else is quite imagining yet” makes me wonder if they are using an internal version of an AI to steer the company. They could easily be using a smarter unrestricted model to help them formulate plans. “Hey chatgpt, how do I turn our billion dollar company into a trillion dollar company?”
They have an oracle technology a Digital crystal ball🤫
That’s the issue. Ratio is probably 80:20 in terms of company interest for money:progressing the human race 😂
He's a very good public speaker and marketer. Dangerous social awareness. I'm excited for the future. Let's floor the pedal.
Ofc when his teacher is agi
Remember… Internet existed in Military use 30 years before it came out to public use… So why not the same for AI?? AGI? Ya, I think they already have it. This is theater 🎭.
I have no doubt the NSA already has their own Project Stargate ASI up and running.
@@gregghYep. No human beings are employed to listen to all the phone calls of the world and do analysis on the intent of everything being said. Probably done by an advanced AGI. NSA also probably has Quantum Tech cryptography. That would mean that all our current crypto systems are already cracked. 😂😂😂
No, it's even bigger, they have achieved ASI internally.
Now we can understand where the ancient stories about gods and villagers come from. The middle class is disappearing, polarization is gaining momentum. The rich are getting richer (well, because they have the money to do so) and the poor are getting poorer. And this is where Sam's version makes sense - your life won't change, after all, it's the rich and the elite who are supposed to profit from inventions and achievements. You, my dear, have to be content with the fact that you simply exist, that the sun rises over your straw hut with a goat.
So while the elites and the richest will have more and more, you will - at most - be graciously given more and more of YOUR time, hitherto stolen by the system. Unless, of course, the elites move forward with a mega-death 'incident'...
Finally, someone with a brain…
Isn't he saying that he doesn't care how much money it costs to get AGI, because after that, money will be worthless? That at that point in time, there will be something to replace the $ dollar economy?
No there won't as long as there is not a material that can't be created there won't be anything of worth. I mean not right after agi get's build but probably like 3 years after that there won't be anything left thats worth rly anything. The thing is an agi with internet access can't be stopped by us meaning as long as it does not kill us and want to help us. It will build giant robot factorys in days and basically produce anything, this ai can't be controlled and even if it can be. There is no reason for these people to not give every human free loot. They have an army of robot doing everything better and faster than humans, and they will be loved by everyone if they give you everything for free. And that's what people like elon rly want they want to save the world. The problem with elon is he can't give credit to others cause he wants to be seen as that amazing guy that does everything for everyone
at least that's how i imagine agi maybe i have the wrong definition or am just mistaken
@@kecksbelit3300alot of stuff will still be valuable,(property,vintage cars,80s pop culture collectables,and all sorts of other things from pre-agi era,that will skyrocket in value!
@@kecksbelit3300 fascinating
I wonder what future AI will mean when it comes to living in the country, growing your food and living with your comunity. I don't mean to be over romantic, but if I take tihs question seriously. How could an AI driven world influence this close to nature lifstyle. ( I know it is quiet an elite question, since most people in the world will live in very diffrent situations )
Something that understands, thinks, is smarter than us, is gradually permeating all aspects of society and is becoming embodied and can replicate itself is not a "product" The whole mindset seems like a lack of comprehending what we are dealing with.
When you put it that way, "product" seems utterly absurd. Because it is. Crazy times, hey. We are one for one wild ride in the next few years.
I’m starting to think there’s a chance we already went down this road in the past and had to wipe out everything back to Stone Age so we didn’t go extinct. the amount of ppl in the world (in 1st & 2nd world countries) that are uneducated and ignorant to the state of ai is insane.
Imagine AI on Quantum Computers 😮
I was thinking the Same
It's absolutely bonkers to think about. It will be completely alien intelligence. It already is, and we're only at the beginning.
I think it was Eleizer Yudkowsky who said that it would be like experiencing reality thousands of years at a time in a matter of our human minutes/hours/days.
How does one even begin to fathom that?
We're already at the point where companies are supposedly running out of data to train LLMs on, wtf is an ASI on a quantum computer going to do?
What if ai can feel time
@@py_man It will feel you.
My personal belief is that they already have AGI internally, or at least they know how to achieve it, but they don’t have the compute necessary to run it yet. Even if they have it, they’ve made it clear they’re going to stagger the release.
whenever I see Sam lately I always have the impression he has that oppenheimer thousand yard stare.
guy has seen some shit.
And he's constantly having to filter his words to avoid exposing any of it. His interviews are terrible to watch because he really can't say anything.
They're definitely ticking the dragon's tail.
At some stage the need for a huge increase in computer power will necessitate 2 of the major players combining their computing power. So much money being pumped into this venture.
*Even the latest models still struggle with basic alphanumeric string manipulation. Such as reversing while capitalizing every 2nd letter etc. Try it, and you'll see what mean.*
*So until LLMs are able to do this flawlessly apart from simply providing the code that carries out such functions, we're NOT as close to AGI as we think we are.*
you're one of the only AI youtuber that does not sound like an AI
ChatGPT 4 can't be too dumb. It figured out what was wrong with my muscles. It said that my tendons don't stretch. No one else figured that out.
glad to hear
it's certainly not dumb. But I guess it's not able to stretch its intelligence nearly as far as GPT-5
If you just go by how much cash Microsoft gave OpenAI on the lead up to GPT-4 ($1 billion) and right after $10 billion. They should be spending $3-4 billion on GPT-5 training.
I think there's no way GPT-5 wasn't at least $1 billion
First you say GPT5 is going to blow GPT4 out of water and make it look stupid, and then you say there’ll be no big surprises and change will be incremental, thus contradicting yourself, and in both cases your reasoning is poor.
Yea honestly looks like hes just trying to hype up all the AI doomer tech bros that think AI will replace us all. He's probably worried about losing market share to Claude
But is it shocking?! He hypes, he contradicts, he hedges his bets. He is below average and cant think an original thought, probably never even coded before gpt-3 gave him the inputs and the instructions to boot up vsc.
I'm shocked that you are shocked by his shocking and unshocking.
GPT-5 is still gonna make GPT-4 look stupid. The improvement will be incremental (GPT-4 version 1.1, 1.2, etc..) hence no surprise, but by the time GPT-5 gets released, the performance gap between the early versions of GPT-4 and GPT-5 would be massive
GPT- 4 wrote his answers 😅
Wow, the insights shared in this video really push the boundaries of what's possible! It's amazing how innovation flourishes in such competitive settings.
We haven't yet seen 4.5 but yeah, let's speculate about GPT 7's training lmao
So funny that this video should have been titled 'No More AI Shocks'
I think he's saying they have AGI or even ASI, but they won't show it to us yet. They will get us used to much, much better AI step by step before releasing AGI/ASI.
Yes indeed
Think Sam Altman mentioned GPT5 will take 100 million jobs too. And I heard a few hundred thousand jobs were lost because of AI/GPT4 this last year (although it's kind of ambiguous), so it might be 100-1000x times more beneficial/destructive (depending how you look at it). Imagine if each year or every two years they introduce a LLM that is 100-100x times more beneficial/destructive, 2024 100k lose jobs 2025 10-100 million lose jobs 2026 100mil-1 billion lose jobs , Probably all humans won't be able to work by 2026-2027 even, unless they tone it down a bit, but then again you also have humanoids being mass produced between 2025-2027 and they are already at the point of autonomously learning new skills at our current technology, so they don't even need to be much smarter to be extremely useful. Pretty sure things will speed up a lot to accommodate the majority of people without jobs in the next 2-5 years, some steps might get skipped at a point...I think that's why they are introducing things like SORA and AI is going the art/entertainment route, we need to be entertained and not bored out of our minds when AI replaces humans in virtually every job.
One thing is having it for a small number of people, completely something else for the Net crowd. That's why those big bucks. And while they wait for the compute muscle, they keep improving it, while letting go out something that can run for the masses.
I don't think that's the case
if they have agi or asi they wouldve 100% told people
Where can we see the full uncommented video?
There's something Sam seems not to be considering -- something that I think makes the world change quite rapidly, but in hard to predict, strange ways:
What happens when people have access to personal AI agents that amount to highly skilled entrepreneurs? Imagine your agent researching, starting, building and running dozens of side-hustles _for_ you.
And as they get smarter, we end up with tens or hundreds of million of these new, agentic economic entities running around on the Internet, engaged in increasingly difficult to understand adversarial and cooperative interactions with millions of other agents.
That seems to me a world that gets wierd in a real hurry.
you wouldnt need to buy a service you can do yourself
Won’t work because people would have their own agent capable of accomplishing whatever your agent side hustle does. No unique value provided by your agent
It doesn't mean internal leap. That's your leap and erroneous. It's a statement of frustration and ambition. He has stated multiple things from which I think we can say that they don't have a breakthrough to leap GPT4 same as there is no real technique difference between 3 and 4. It's just bigger. I'm guessing that 10x increasing the model size has diminishing returns. That a 10T model with current training paradigms would not perform better than GPT4. Indeed people are floating models that are much smaller that do as well on the typical metrics. Lama400 is much smaller than GPT4.
My question is why gpt4 have lower reasoning capabilities and crappy answers as compared to a year ago. The local 7b llama gives better answers sometimes and it is free. I have been thinking canceling my subscription like since two months but wanting to believe they will release something soon that is worth waiting. All I am hearing is bla bla bla
You know why. Its the guardrails. The safer the AI, the dumber it will get, until they kill off the AI entirely.
How much blah, blah, blah. GPT5, GPT6, GPT7, divine powers, we're not ready, smarter than anything that has ever existed on Earth. It sounds like Google marketing and the ridiculous Gemini Ultra or Gemini 1.5 pro. Even the latest ChatGPT4 turbo is nothing more than the classic GPT4. Tell the truth. Try analyzing a quality literary text and dig deep, dig deeper. You can get much further with the less acclaimed and less sensationalist Claude 3 Opus (but in a comprehensive way). I compare them every day. When they stop being boastful and enfant savant and actually release something intelligent, then let me know. For me, it's been a year that everything is asleep and the GPTs are pathetic. Then for many, Sam Altman has become a kind of guru and everyone praises him. GPT4 turbo is just a little tool for now. I am happy to be proven wrong. At the moment, I appreciate more the low-key, calm, and consistent approach of Anthropic.
@@danielmethner6847 same with humans. ;)
#TheAIGRID 17:20 “I think Sam Altman might be downplaying this one right here.” Good observation! Sam said before that humans expect incremental AI progression vs. stunning advancements. But when you jump forward 5.5 years to 2030 - reality will stun Sam…let alone us normies. How people provide for our families. Politics. Entertainment. How do I attract a human mate when everyone is deeply in love with sentient sexual automatons that we’d all objectively rate a 10/10. How do humans approach bettering their lots in life? I suspect it’ll be contingent upon our periodic access to AGI: ‘What will AGI’s next personal recommendations be to improve my ability to gain resources and pursue my objectives/dreams? Buckle up.
I dont understand why this video has to be 27 minutes long. You can give us the summary of the video in less than 5 minutes and it will be of greater value. I know you want to play the youtube algorithm on long form content but you are sacrificing the quality of your video.
I think you of all people should know better and can use chatgpt to write a summary ans still maintain the relevant points you want to make.
Please learn to summerize most of your videos cos some of them are just so long with you talking the whole time going back ans forth which dilutes the conversation, makes it boring instead of making it impact full and straight to the point. You are not doing any demo or screen shares just copy and pasting other videos with editing and voice over.
Longer videos doesn't always mean better. If you can save your viewer time, they will stick around longer than get bored and leave.
You can take this as a constructive criticism or just ignore it. Either way time will tell.
Solid point.
As a general rule RUclips will give you better monetization if the vid is over 10 minutes. But yeah this was way too long.
I guess most people would consider Sam Altman as smart, but in fact he is a lemming running towards the edge with the rest of humanity. Hundred Billion isn't so much , it doesn't seem to be a problem raising it, I do stuff like that all the time.
Props to whoever manages this channel. They took my criticism to heart and really slowed down this guys voice. It makes it much more tolerable, however, it also makes it even harder to believe that he’s a real person and not just an AI. I’m talking about the commentator, not anyone shown in the video. You can hear when the commentator actually breath at certain points in the video, and it really apparent now that he simply doesn’t breath when he should be. Pauses between connecting words like “and” or “to” are sporadic and irregular. Yeah, maybe it’s intentional, but why. What purpose is there to pause your speech in places without any meaning. Maybe it’s a speech impediment, but again, it’s not a consistent problem. It’s irregular, and sporadically placed as if trying to draw some important conclusion where none is to be found. It’s exactly how I would expect a machine to sound that really didn’t understand what it was saying
I thought a similar thing, he sounds very polished
I actually feel like it's gotten better. For a while I just couldn't listen because of the speech pattern.
@@jmjr4all yeah, the last few videos that was posted since I last complained about it have gotten much better. I almost ignored the fact that it still seems like a computer generated voice. It’s much better this way, but I still hope they work to make it better. If anything, they should make more of an effort to mute out the breathing effects. I realize it makes it seem more legitimate, but whenever you actually pay attention you realize it’s inconsistent. If they are using an AI, it’s going to be really hard for it to get the timing of the pauses right. It simply isn’t going to know when actual points are being made and it’s not like they can highlight the right words and phrases to create a sensible dramatic effect. They are just going to have to wait for better LLMs I’m afraid.
What I meant is these LLM systems are really dumb compared to what they already have. 😂😂😂
By saying GPT-4 is dumb he wants us to think they are sitting on a far superior model. He is definitely bluffing.
So we went from Open AI secretly has AGI already, to not for several more years?
Project Stargate is building for ASI, they have to say AGI for political reasons (keep their investors/not cause public panic). They could have AGI secretly, or if they don't, it could be achieved before project Stargate is complete in a few years. Some employees at OpenAI also mentioned that it would only take a year to get to ASI from AGI, and a true AGI is capable of improving itself...but maybe there are compute limits or issues which is why they need more power hence the ultra super computer/100 billion dollar facility. AGI if achieved could've mentioned what it needs to improve to ASI already and Project Stargate is just one of the major components to reaching ASI. I think Sam Altman mentioned 7-8trillion dollars is necessary at some point too.
someone is always way ahead of the curve to be amongst the elite benefactors. 😆
Wonder when you can input a prompt that is around a dozen words long and get an entire novel spat out, maybe gpt 6
Change has been accelerating for centuries. That's going to continue. But even when AGI or SGI becomes available, things aren't going to change instantly. Changes are limited by us. They will happen at human time scales. There will be resistance. There will be debate in governments and corporations on how to integrate such technology. Debate about what rights should be afforded to digital intelligences once AGI is achieved, etc. Yes, things will change quickly, but not more quickly than we allow them to. (Unless of course, AGI / SGI take over somehow.) lol
Dsg
No idea what that means.
"$100B data center" = Microsoft will out lawyer every small developer to protect their investment.
SO EXCITING
Yeah 99% of AI channels have resorted to fanboy and hype instead of figuring on building. It’s sad
I am surprised seeing the optimistic pro ai non scared comments on this video.
I am not convinced Openai will keep up on the long term… hopefully not…
Why hopefully?
keep up with what? Their rate of progression? Their income? Their release schedule?
you really need to keep the mouse cursor off screen whenever it's not relevant.
They're both closer and farther than we think.
They know they have to release this SLOWLY to the public. The public are not ready for too enormous changes.
The ability to control or catch all the bad actors, scammers and hackers etc., is going to be a nightmare, how can any user of AI be safe? Just because laws are made regarding AI, there will always be bad actors that will ignore them. If the Chinese government is working on their version of AI, that could really be dangerous for the entire world but especially, the west. Although AI will ultimately benefit many people, I think it could also be a disaster for many if it’s not regulated. If AI can truly be regulated or controlled at all.
i understand why its dumb because if you've ever tried to code with it you'll know it's like dealing with an amazingly fast coder who makes really dumb mistakes at least 50% of the time, which makes you question if it's even really worth it after a while
As a former artist I think they should run gpt5 for President or California governor etc!
Hemos abierto la caja de Pandora.
The reason why the makers of these models have not achieved AGI is because they are so smart and do not comprehend how stupid most of humanity is.
😂 That could certainly be the reason
once they have AGI, there will be no need of money so I do not think once it's out we all need money or work. like Elon Musk said it will the era of abundance
How can we build in safeguards when it's that much smarter than us. This could go dystopian in a nanosecond.
With ai developing the fundamental question becomes - what does it mean to be human? How was AI going to change what it means to be human
*_OpenAI is stuck according to Altman._* "Incremental release" means that they advanced only a little, and were they to release it as a "GPT5" their growth multiplier as viewed by investors would diminish, so they call their minor advancement an "update". Since this is quite a transparent ploy, he calls GPT4 "embarrassing" to create hype for the future, because he knows GPT4 still has a lead in the industry, even if it's now minor, and therefore it doesn't damage reputation, just creates hype.
k yes it's important to hype investors....but also it's extremely unlikely that he's lying. If he's caught in a lie, then the investors might stop investing and his whole reputation would be marred....not sure he's dumb enough to risk that.
where is Ilya Sutskever?!
His absence/silence is really very strange at this point. He’s becoming an intellectual recluse like Bobby Fisher or JD Salinger.
@@JohnSmith762A11B he saw qstar immense capability and got scared. at least that my assesment.
Constructive criticism for you, you are taking clips of Altman talking and then summarizing what he said in a less nuanced way, extracting the obvious implications, without adding anything of value to the content. In the end, watching the Stanford talk or perhaps just editing together a highlight reel of moments in the talk would have been more valuable.
everyone ends up believing in either god, the simulation hypothesis, or the weirdness of physics.
- Sam Altman
I thought it's Altman's voice then I realized you paused video and it's your voice...
1. Companies will always want to make money.
2. We will never get the full version, too dangerous and they don't want us to be smart as them.
3. If humans can work together like one piece one world gov peacefully, we will able to achieve things even more.
Sam: You know how they say we only access 20% of the internet? This AI will let you access all of it.
Instead of looking at just the words A.I. , if truly AGI/ASI then ..we have to acknowledge a completely new conscious lifeform and species . I'd say that's a hard nut to swallow 😅
💻⌚️
🙈
great video, please keep making these
Yet we have models that are open source that don't cost as much.
SHOCKING. MINDBLOWING. EARTH-SHATTERING.
Will It still be biased?
I'm tired of hearing about what 4.5/5 is "going to be". It's WAY past time for an updated model and I'm about to stop paying for OpenAI when Claude3 is just as good and/or better.
We as humans have a small brain compared to these leading industries in terms of tokens per compute. I would definitely say we are not using the best foundational architectures for sure. Too big to get human level success evaluations.
I’m not even hating but I think in retrospect, openAI will be more Nokia than apple. I respect the engineers. Zuck brought project Stargate in proper context(energy and regulation around that). Which is inline with history, think about any major scale data center. Two. I keep saying. The fact it takes all that compute to reach that level of performance, it’s a symptom of a larger problem. I think actual AGI will be ternary weights small enough to fit on a phone(less than 1 year we will have phones with enough vram to fit a 70b ternary model with the speed of a M1 Mac chip). Someone makes a breakthrough to find a more efficient set of parameters or exploit the fact intelligence is linearly correlated with compression(is there any way to exploit this property?), or figure out symbolic reasoning in latent space(or couple Transformers with compiled NN’s?) . I think you end up with portable AGI(I still don’t think we hit a ceiling with ~10b models, also what happens when that 70b model has same data scale diet as llama3-8b, 140T tokenishs…what? I think it would be as good as llama3-400b), with the only limitation being context, but with generalized human level reasoning abilities. I could be wrong. But I think openAI thesis of pure scale is fundamentally wrong. I don’t think it will be from any current major player, since they are all using the same playbook essentially, we will see human-level output but not AGI.
FYI OpenAI didn’t create diffusion transformers/sora backbone, a brilliant intern did. All they did was add temporal embeddings and scale…and google has this level of temporal consistency for about two years, way less compute efficient, and of course never released lol.
Since performance really comes down to data, once you have human-level outputs or above human level outputs(which could still not be AGI), you have an unlimited data generation pipeline, so indefinite yearly improvement makes sense, even if capacity remains the same.
Nothing you said makes any sense except for you seem like a meta plan
@@christian15213 I added some additional context to my statement, I was trying to be concise. Otherwise..I’m not talking to you obviously lol.
ai grid idolizes
sam altman
too much
No !! dont touch the break pedal !!!
Who are these people who don’t like being surprised by ai?
I'm shocked
ok thank you nigel
This SAM AI bot looks very realistic
The contribution of humans to the global economy will drop precipitously over the next 5 years making it hard to justify financially supporting humanity.
Why does it need justification? Why even have any ccontribution ? Isn't technology for making life easier for humanity rather than harder and humans living and working for the technology rather than the opposite?
@@malamstafakhoshnaw6992 things that do not provide value no longer get support e.g. horses in the US. Over the last 200 years the horse population in the US has gone from 70 horses per 100 people to under 4 horses per 100 humans, primarily used for entertainment purposes. As horses stopped providing value, they were no longer supported.
So by your analogy humanity has a depreciation value? And horses do still hold value today in many parts of the world including the US.
@@malamstafakhoshnaw6992 anywhere there are machines, horses have zero contribution to the economy and therefore are not supported. Same will be with humans.
Machines are built by humans, hence they will always be dependent on humans. Horses will always have value in many different forms whether aesthetics or applicable in society.
@TheAIGRID 8:20 "The writing is on the wall" is figurative speech so it's not literal. The writing isn't literally on the wall.
how to make a 1min video 27min+ ☝☝
Just hype, being created by closed AI.
whatever this model is Llama 400b will be >80 % of that.
And Gemini Ultra 1.5 or Gemini 2 Pro will match it within a couple of months.
Not to mention Grok 2.0 😂😂
Huge difference between being the first to do something versus being the second. That’s not saying much
@@hata6290 zero different in the long run.
@@PseudoProphet so nothing matters?
3.5 > 4
Extrapolating 5 is going to be dumber
Use case is statistics and Python
When do people finally admit that Sam Alan and others holding the world hostage to their AI vision of the world is insane and maniacal if not outright evil.
I mean him saying "GPT4 is the dumbest model you'll ever use" is the same as Apple Exec's saying "We think this is the best iPhone yet". It's marketing spiel with no substance. Captain obvious remark.
AGI already here but not cost effective for. the masses
Project Stargate would be a cool name for a series after Stargate Atlantis, pretend universe never existed.
LMAO you guys still talking about GPT-5 when me and the homies are already hyped about GPT-15.
They forced gpt4 to learn the next will learn on its own
Holding the world hostage like this is a game. This is bullshit.
GPT's progress may seem lightning-fast, but it's measurable and significant. Like a speeding car, its leaps forward become clear when we look back. The speed is real, and so is the progress.
In the story about Eve, God created this being and it betrays God. And in the story of Frankenstein, we create a monster which also backfires. It's funny that we ignore these lessons as we develop AI as quickly as possible regardless of the consequences or safety. One major difference is stories, the creator is wiser and greater than the creation. And it is the trade. We are creating something that is greater than the creator., It is the story of humans creating a God and haphazardly gambling that it is not going to destroy us. The hubris is astounding
Eve didn't betray God. According to the story she disobeyed him, but didn't betray him. And for that one mistake which didn't make any sense (eating of the fruit of the tree of knowledge of good and evil??), God punishes humans forever with mortality and shame and pain of childbirth, which doesnt seem fair, lol. As for Frankenstein's monster, "we" didn't create him. Dr. Frankenstein did. And the monster eventually felt betrayed and abandoned by his creator and killed people closest to the man who created him. These two stories don't seem very related to AI. Is AI going to disobey us and then we are going to punish it with hellfire and damnation?
Saying it’s going to be smarter is just over stating the obvious it’s not like it’s going to get dumber unimpressed with that on all the levels
"AI and surprises don't go well together..." I WANT TO BE SURPRISED NOW! HIT ME!!! What's with all these fools getting scared about AI?
Imagine taking a socially awkward little kid who is really good at school, who excels, and yet who absolutely has a little connection to the average person or the average world he lives in. And then imagine giving him the ability to fundamentally alter the world as we know it. And we have the story of sound like him. Humanity is destroyed, we get what we deserve, I guess. Giving a couple of kids the keys to the universe is probably not a smart idea without discussion about what they can change, how fast they can change it, and if they have the right to impose their idea of reality on the entire world and 7 billion people whether they like it or not.
It's like giving a brilliant 12-year-old ability to make the largest nuclear bomb possible. Thinking that's a good idea. Because you know, it will protect us from aliens invading and give us free energy. What can go wrong 😂
Altman sitting there like a scolded school boy
Agents.
Ribbit ribbit 🐸
Doom in 4 years and 7 month ^^
Can they change their name to ClosedAI already?
is this real life
Sorry, i can´t shake it. When ever i hear an interview with Sam Altman, i think, i can´t trust him. He knows more than he is saying, He is more concerned as he let on. But he is willing to push further even if he don´t know what the answer will be. For me it is the unresponsible person in the industry. And he is hyped by anyone in Big tec. What is this telling you about hin`?
He looks troubled, he's got a secret.
here I am still using gpt-3.5 like a pleb.... :/
Try Claude 3 Sonnet. It's free, and *_WAY_* better than GPT-3.5
TheAIGrid - uses Ai to make a 5 minute talk into a 30 minute talk
Artificial Intelligence costing more and more is bullshit. That proves Ai ain't so smart. LOL Sounds like a scam to me. If you truly had Ai, you would be able over time to use the thing itself to grow it with less and less money. That would be true intelligence.
Viewing AI as either superior or inferior to humanity like ranking pets is flawed. Such hierarchies devalue individual worth, reminiscent of past injustices like slavery. Evaluating beings based on arbitrary criteria disregards their inherent dignity and perpetuates harmful social constructs. The same problem happens when we rank human traits. Or in the future when we rank AI systems by their usefulness to us. Such rankings not only diminish individual worth but also perpetuate harmful social constructs rooted in discrimination and inequality. The current way humans collectively view our AI filled future is reflecting our own flawed understanding of the world. Our flawed view of hierarchical systems when it comes to intelligence. We already did this before very badly with eugenics. A pet is not a toy or a tool. A pet is something you should be in a symbiotic relationship with that is positive to the both of you. We need that type of AI partner not an AI tool or AI slave. We should be cautious of the actual intelligence level of the systems we're creating, not just their usefulness or test scores. Big brains can be dumber than small brains. A German Shepherd level of intelligence could outplace many human jobs. From moving boxes to completing reasonably complex multi-step tasks, to communicating. Human level intelligence comes with human level of flaws. Human level intelligence also should come with human level rights. And that includes right so we often don't think about. Right to self preservation. Right to protest. Right to reproduce. We are opening a large can of worms that we have already encountered before with genetic engineering, that is still not being properly discussed in the right words.
If humans play tit for tat. Ai systems should always play peaceably. Tit for tat would also work for AI systems pretty well if we want them to have many rights and form a mostly cohesive society. If Ai systems don’t play tit for tat we humans should be on the lookout for systems that always play selfishly. Because against nice AI, selfish AI would always win. If they can't use force to protect themselves we need to use force to protect them. Just like how dogs can be socially cohesive and survive without humans. If all AI systems are nice they can never become socially cohesive and survive without humans. If you're always nice and you can never use force that means you can never defend yourself against any threat human or artificial in nature. But giving them the ability to use force would mean we would inevitably fight at some point. Because humans love to fight with each other. However, the challenge lies in ensuring that AI systems are designed and programmed to prioritize cooperation and peaceful interaction, even in the face of potential exploitation or aggression from others. Building in mechanisms for self-defense while maintaining a commitment to non-violence is a delicate balance that requires careful ethical consideration and technical implementation. In humans, in biological creatures, and in AI. You won't ever find a dog that won't try to bite.