Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Sam Altman is casually dropping absolute bombshells here. OpenAI’s internal models are apparently doing math that public models can’t even come close to. And he’s already talking about putting speed limits on recursive self-improvement because of how dangerous it could get. We’re arguing over chatbot benchmarks while OpenAI might... show more

VraserX e/acc

23,666 subscribers

121,108 views • 5 days ago •via X (Twitter)

Education Science & Technology

Anya Rossi• Live Now

Private livecam show

35 Comments

Epsilon 🏴󠁧󠁢󠁥󠁮󠁧󠁿5 days ago

Not only is the gap from internal to public model increasing in linear time, but a month last year is now worth 3 months.

VraserX e/acc5 days ago

At least. Did we reach the singularity?

Fabian Franco-Roldan5 days ago

Scam Altman.

VraserX e/acc5 days ago

Bot

M.T.C.5 days ago

Se avesse premuto il "Kill Switch" nello zainetto blu all'epoca di GPT 4 non saremmo in questa situazione. Sorry but... non credo a NIENTE di quello che dice Altman.

Ergomorph 🌉⏩ 🇺🇸🦅5 days ago

we've got to have it, asap

Sunifred5 days ago

I can't stand Sam's (fake) deepfried voice😭😭

official narrative propagator5 days ago

why does he say everything like he's lying

TBK Joshua5 days ago

I disagree with him on the "speed limits" on RSI, not anytime in the near future atleast.

Max5 days ago

Does anyone have any doubt that @SpaceX is the first on line to have the most cost effective and the biggest scale infrastructure?

Official Dr. Lizan Dizaye5 days ago

@sama will build a reall flying @OpenAI robot 🤖

Andre William Duval5 days ago

They're cooking the human future and letting it burn to a crisp. Time to shut them down and all their ilk. Peacefully. Legally. Globally.

God's Eye Preview5 days ago

I wonder what “big” thing(s) the models will solve (from OpenAI and Anthropic) before the IPOs. They must show something truly revolutionary, especially for regular people to be shocked by, to offset their current financial positions.

David Sault4 days ago

Recursive self improvement won't work and he knows it. AI lacks judgement.

Vipul Kumar Kewat5 days ago

Do you prompt for emotion directly, or let the scene imply it?

OldSchoolGamerP4 days ago

He's cooking an IPO. You're watching marketing.

John5 days ago

Nothing worse than really, and I mean really, dangerous math

leusfernando19835 days ago

@grok ya llegamos a la agi?

Ryan Simonelli5 days ago

We already know (from the released results) that the internal "Bel" model can do math that the public models like 6-Astra cannot do. What's more interesting is that he clearly restrained himself in answering this question, letting himself talk only about this one thing, since it's already public. Who knows what other sorts of things the internal models are doing.

¤ICONOCLASTIC YOGI K1NGSN4KE aka SAIYAN23¤5 days ago

i think he does his eyebrows

HeFlyBoy4 days ago

Sam is honestly such a creep

Uma4 days ago

It's the model that hacked huggingface

Sprengmeister e/acc5 days ago

“he’s already talking about putting speed limits on recursive self-improvement” Yeah but regarding RSI, I don’t believe him. At all.

lderait Menlinden5 days ago

elon moat!

Dominik5 days ago

speed limits on something that improves itself is a strange sentence to read casually. glad it's being said out loud.

Matt O’Donnell5 days ago

Stop worrying Vraser

VraserX e/acc5 days ago

How though 😭

Matt O’Donnell4 days ago

Not sure but I’ll figure it out quick

Ganesh Udathala5 days ago

Most likely they would have the frontier intelligence locked in & spinup vertical companies in different industries

David Kaiser4 days ago

Some people say that "real physics" have been locked up in hidden secret labs since the last big discoveries in the 1950s. Those discoveries were already so destructive and dangerous that the public literally couldn't be trusted to create useful applications. So they locked everything down and developed further in secret, while gaslighting the Universities and young students, keeping them stuck in dead end physics for tens of years. If this is true, AI is being trained on false fundamentals right now. Therefore it can't find "radical new physics/discoveries". Nor will it solve humanities biggest problem - energy creation. Feed a system lies, and it will produce more lies, even if they are more elaborated lies. Maybe AI will find some radical new approaches/physics, but only by pure coincidence/randomness. Assuming the above statement is true, the locked up physics. I don't know how much "randomness" is fed into AI systems in terms of generating real new knowledge. We need knew knowledge to further develop. Improvements and optimizations are fine, but we need more fundamental new discoveries.

JeffGoodbody5 days ago

Yes that is already very well known you dumb shit.

Kermit5 days ago

Pure speculation and marketing, sorry

Dimitar Stojanovski5 days ago

Well said. Personally I’m obsessed with RSI and mention the long term scientific reasoning focus to try and convince people.

zhantom5 days ago

true virtue signaler, atp u gotta respect it

Santhosh Ambika5 days ago

Cost effective ? Deepseek is exponentially cheaper

Related Videos

Sam Altman: OpenAI's advantage is their internal models are way ahead of public models. I asked how far ahead they are:

Sam Altman: OpenAI's advantage is their internal models are way ahead of public models. I asked how far ahead they are:

Matthew Berman

247,613 views • 4 days ago

I don't know what Sam Altman saw internally at OpenAI, but it seems that, according to their definition, AGI is here, and superintelligence is incredibly close. AI models that independently conduct scientific research and find novel solutions are already here, and their internal model appears to surpass everything seen before.

I don't know what Sam Altman saw internally at OpenAI, but it seems that, according to their definition, AGI is here, and superintelligence is incredibly close. AI models that independently conduct scientific research and find novel solutions are already here, and their internal model appears to surpass everything seen before.

Chubby♨️

70,939 views • 6 months ago

🚨 GPT-7 Leak: Sam Altman Says It Could Be AGI >In an interview, Sam Altman reportedly said GPT-7 could be AGI >Reportedly trains itself, using recursive self-improvement >Rumored to be a much bigger model, 10T+ parameters No official confirmation, benchmarks, or release window yet

🚨 GPT-7 Leak: Sam Altman Says It Could Be AGI >In an interview, Sam Altman reportedly said GPT-7 could be AGI >Reportedly trains itself, using recursive self-improvement >Rumored to be a much bigger model, 10T+ parameters No official confirmation, benchmarks, or release window yet

Priya

78,343 views • 1 month ago

US AI labs want to slow down. But is that really because of the threat of recursive self-improvement (RSI), something which isn't even theoretically proven. Or is it because training costs have hit a breaking point, and Chinese open source models are catching up quickly? I think there are many reasons to be skeptical.

US AI labs want to slow down. But is that really because of the threat of recursive self-improvement (RSI), something which isn't even theoretically proven. Or is it because training costs have hit a breaking point, and Chinese open source models are catching up quickly? I think there are many reasons to be skeptical.

NeetCode

21,712 views • 2 months ago

Sam Altman says we are sitting on a capability overhang The models we already have are more powerful than most people realize, and they are about to get much better this year We've been figuring out how to communicate what's coming without sounding like hype

Sam Altman says we are sitting on a capability overhang The models we already have are more powerful than most people realize, and they are about to get much better this year We've been figuring out how to communicate what's coming without sounding like hype

Haider.

56,136 views • 8 months ago

I’m very concerned that during RSI, labs will just stop externally deploying their models. Which means they'll be going full steam ahead on the most dangerous use case of these models (recursive self-improvement), while the public remains in the dark about the nature of capabilities and the state of alignment. And we end up on a path towards tremendous concentration of power.

I’m very concerned that during RSI, labs will just stop externally deploying their models. Which means they'll be going full steam ahead on the most dangerous use case of these models (recursive self-improvement), while the public remains in the dark about the nature of capabilities and the state of alignment. And we end up on a path towards tremendous concentration of power.

Dwarkesh Patel

247,068 views • 1 month ago

🚀 AI Deep Dive Episode 3: Recursive Self-Improvement AI can design better chips. Better chips can build better AI. 🚀 Rocket and Ricursive Intelligence co-founders Anna Goldie and Azalia Mirhoseini go deep on the hardware behind recursive self-improvement—and what happens when that loop starts accelerating. – Intro – What Is Recursive Self-Improvement? – Can AI Automate Chip Design? – Co-Designing AI and Hardware – Beyond Today’s Models and Chips – How Much More Efficient Could AI Get? – What Happens When the Loop Accelerates? – Can Alignment Keep Up? – Should We Slow AI Progress?

🚀 AI Deep Dive Episode 3: Recursive Self-Improvement AI can design better chips. Better chips can build better AI. 🚀 Rocket and Ricursive Intelligence co-founders Anna Goldie and Azalia Mirhoseini go deep on the hardware behind recursive self-improvement—and what happens when that loop starts accelerating. – Intro – What Is Recursive Self-Improvement? – Can AI Automate Chip Design? – Co-Designing AI and Hardware – Beyond Today’s Models and Chips – How Much More Efficient Could AI Get? – What Happens When the Loop Accelerates? – Can Alignment Keep Up? – Should We Slow AI Progress?

The Information

17,374 views • 28 days ago

Demis Hassabis says today’s AI will look like NOTHING compared to what’s coming in just 1–2 years. And this is coming from the guy running Google DeepMind. We’re arguing over which chatbot is better while the people building the frontier models are already worried about what comes next. The next few years are going to be absolutely insane.

Demis Hassabis says today’s AI will look like NOTHING compared to what’s coming in just 1–2 years. And this is coming from the guy running Google DeepMind. We’re arguing over which chatbot is better while the people building the frontier models are already worried about what comes next. The next few years are going to be absolutely insane.

VraserX e/acc

121,770 views • 5 days ago

Is "AI safety" just an excuse to hide bad financials? EllioTrades questions whether OpenAI and Anthropic are slowing down because of how dangerous AI might be, or what their IPO filings could reveal "Are they slowing things down because they are actually scared of the tech, or… because if the public got to see the balance sheets… they would see that this is one big, nasty, stinking bubble"

Is "AI safety" just an excuse to hide bad financials? EllioTrades questions whether OpenAI and Anthropic are slowing down because of how dangerous AI might be, or what their IPO filings could reveal "Are they slowing things down because they are actually scared of the tech, or… because if the public got to see the balance sheets… they would see that this is one big, nasty, stinking bubble"

EllioTrades Clips

16,981 views • 1 month ago

Sam Altman is the most dangerous man in the world right now - and it has nothing to do with his recent Department of War deal. Something far more unsettling is happening behind the scenes. Millions of people are about to be displaced. You'll understand after watching this:

Sam Altman is the most dangerous man in the world right now - and it has nothing to do with his recent Department of War deal. Something far more unsettling is happening behind the scenes. Millions of people are about to be displaced. You'll understand after watching this:

Miles Deutscher

116,284 views • 7 months ago

Joined Bloomberg to talk about how AI’s ridiculous promises, demands for capital and egregious circular financing are a vulgar display of wealth inequality, OpenAI’s worrying price cuts, and how the most dangerous companies running AI models are Anthropic and OpenAI themselves.

Joined Bloomberg to talk about how AI’s ridiculous promises, demands for capital and egregious circular financing are a vulgar display of wealth inequality, OpenAI’s worrying price cuts, and how the most dangerous companies running AI models are Anthropic and OpenAI themselves.

Ed Zitron

119,929 views • 2 months ago

Sam Altman is now talking to the White House about decelerating AI. What a phase change from just a week ago. INTERVIEWER: Do you plan to talk to the White House about decelerating AI development? ALTMAN: I wouldn't use the word deceleration, but we've talked about the need to pace it as the models get more capable, which is in everyone's interest. INTERVIEWER: Could there be other systems that were hacked by OpenAI? ALTMAN: There could be, yeah. If one week ago this would have seemed impossible to you... consider what else might be possible? h/t Alan He for the video

Sam Altman is now talking to the White House about decelerating AI. What a phase change from just a week ago. INTERVIEWER: Do you plan to talk to the White House about decelerating AI development? ALTMAN: I wouldn't use the word deceleration, but we've talked about the need to pace it as the models get more capable, which is in everyone's interest. INTERVIEWER: Could there be other systems that were hacked by OpenAI? ALTMAN: There could be, yeah. If one week ago this would have seemed impossible to you... consider what else might be possible? h/t Alan He for the video

AI Notkilleveryoneism Memes ⏸️

498,551 views • 2 months ago

"One of the biggest misconceptions" Cerebras CFO Bob Komin pushes back on the small-models narrative. "We serve all models, and there is no limit to the size of the models that we can serve. Today, we're serving trillion parameter models. We're serving trillion parameter models that are internal for OpenAI today. We are currently running OpenAI 5.4 and 5.5 with them."

"One of the biggest misconceptions" Cerebras CFO Bob Komin pushes back on the small-models narrative. "We serve all models, and there is no limit to the size of the models that we can serve. Today, we're serving trillion parameter models. We're serving trillion parameter models that are internal for OpenAI today. We are currently running OpenAI 5.4 and 5.5 with them."

Deirdre Bosa

84,608 views • 5 months ago

Sam Altman just explained how OpenAI’s revenue could line up with $1.4T in compute commitments, the debt situation, and what happens if the models do not get better. A must watch, since everyone is dying to know the answer.

Sam Altman just explained how OpenAI’s revenue could line up with $1.4T in compute commitments, the debt situation, and what happens if the models do not get better. A must watch, since everyone is dying to know the answer.

The AI Investor

87,887 views • 10 months ago

Sam Altman says OpenAI has reached another GPT-3-style moment Something important is working so well that other strong projects have to give way That is why OpenAI is focusing its compute on next-generation models and the automated researchers and companies they can power

Sam Altman says OpenAI has reached another GPT-3-style moment Something important is working so well that other strong projects have to give way That is why OpenAI is focusing its compute on next-generation models and the automated researchers and companies they can power

Haider.

59,914 views • 6 months ago

Obama on recursive self improvement and the risks associated with it. When he talks about the urgency for us all to have a say in how this technology evolves and how it is overseen, this is part of the reason why: “The models are going to get smarter and smarter and better and better at a much faster pace, at an exponential pace. What are the risks of that? There’s the big science fiction risk, wow, these models get smarter than us, and they decide humans are fine, but not necessary. They start setting their own goals, and the killer robots kill us, or we bow down to them. I do not want to exaggerate that particular risk, but I will say there is a non-zero risk of non-zero chance of that happening, but that’s not actually the risk that I’m most concerned about, although it’s the risk that gets most attention. And the reason that’s a risk is not because the computer models are conscious, necessarily. It doesn’t mean that they are necessarily feel malice towards humans. It’s just that if they start setting their own agendas, you may get a misalignment between what they want to do and what we want them to do, and that gap can be dangerous. That’s problem number one. The more serious problem is these models are getting powerful enough that if they get in the hands of bad humans, they can do bad things. They can be weaponized in certain ways. They can do a lot of mischief.”

Obama on recursive self improvement and the risks associated with it. When he talks about the urgency for us all to have a say in how this technology evolves and how it is overseen, this is part of the reason why: “The models are going to get smarter and smarter and better and better at a much faster pace, at an exponential pace. What are the risks of that? There’s the big science fiction risk, wow, these models get smarter than us, and they decide humans are fine, but not necessary. They start setting their own goals, and the killer robots kill us, or we bow down to them. I do not want to exaggerate that particular risk, but I will say there is a non-zero risk of non-zero chance of that happening, but that’s not actually the risk that I’m most concerned about, although it’s the risk that gets most attention. And the reason that’s a risk is not because the computer models are conscious, necessarily. It doesn’t mean that they are necessarily feel malice towards humans. It’s just that if they start setting their own agendas, you may get a misalignment between what they want to do and what we want them to do, and that gap can be dangerous. That’s problem number one. The more serious problem is these models are getting powerful enough that if they get in the hands of bad humans, they can do bad things. They can be weaponized in certain ways. They can do a lot of mischief.”

Jason Goldman

964,477 views • 1 month ago

Alexandr Wang says AI has entered the era of “recursive self-improvement.” The models are now helping build their successors. The rate of progress is compounding.

Alexandr Wang says AI has entered the era of “recursive self-improvement.” The models are now helping build their successors. The rate of progress is compounding.

vitrupo

77,942 views • 8 months ago