OpenAI CEO: “SuperIntelligence within 1000’s of Days” 

Which will come first: ASI or GPT4o Voice?

“ASI in A few thousand days” as in “advanced voice mode in a few months”

In the early 80s I was having a conversation with my grandfather. Time magazine had just come out with “Computer machine of the year” issue. I told him that one day computers were gonna be able to think like a human and he said there is no way a machine would ever be able to do that. I said sure you’ll be able to just talk to your computer and it’ll talk back to you. He said there’s no way. I wish he had lived long enough to see it happen.

1 000 days ≈ 3 years 2 000 days ≈ 6 years ... EDIT 9 000 days ≈ 25 years

Thousands of days….also known as years.

We are on a path of no return to something that will change the course of human history forever. I pray the world ends up to be a better place.

What he really meant is “ASI in a few thousands of billions of dollars”

He is also a CEO looking for more funding so… there is that.

How it started: Facebook: Begins Everyone: 'Cool - we can all reconnect with friends' How it's going: ... For all the opportunities, always remember that technology will ultimately be used for the worst possible scenarios as well.

Let’s be honest: AI or even ASI itself is not what’s scary. What’s terrifying is the notion that 99% of people are about to surrender their lives, their privacy, and their data to a single individual or a small handful of corporations with limitless capital. Think about that. Altman’s vision of AI assisting humanity, pushing us towards shared prosperity, is utopian-but for whom? The masses will get a watered-down consumer version, while a select few will wield the true power of advanced AI, likely far more superior than anything they’ll ever offer the public. We’re seeing it even now. GPT-4 is amazing, but it’s no secret that GPT-5, or some version of it, already exists. What we’re interacting with is not the latest and greatest-it’s a controlled release, a diluted version. Altman is likely already showcasing GPT-5 behind closed doors, wooing investors and gathering capital while the public is handed a previous iteration. This isn’t paranoia; it’s just the nature of business, the nature of power. How is it fair that one individual or corporation could end up with all the information, all the power on the planet, while the rest of us are merely passive participants? This isn’t the democratization of technology-it’s the centralization of control in its most extreme form. And then Altman talks about prosperity as though it will somehow be a global wave that lifts all boats. But prosperity is relative. The US, Silicon Valley, Europe-they might all reap the benefits of these advancements. But entire countries, even continents like Africa, are still grappling with basic infrastructure. How can they possibly tap into this Intelligence Age when they lack the foundational tools to participate? It’s a glaring blind spot in Altman’s vision. He mentions compute, chips, and energy as the building blocks of this new era, but what about the billions of people who lack access to these resources? Without global infrastructure investment, without a conscious effort to bridge these gaps, we’re just accelerating an already widening divide. Sure, AI can create tremendous prosperity for those who already have access, but for vast swathes of the world, this Intelligence Age will be as unreachable as ever. Worse, they’ll be even more dependent on the few who control these systems, widening the existing gap between the haves and the have-nots. Ultimately, Altman’s post leaves me with more questions than answers. Yes, deep learning works, and yes, AI is getting better with scale. But who benefits from that scale? Who holds the keys to this new kingdom? Because right now, it looks like we’re headed for a future where a tiny elite owns the most powerful tools humanity has ever created, while the rest of us are left to work with whatever versions they deem fit for public consumption. That imbalance of power is what truly scares me-not the technology itself.

A few thousand days is still 8 years. For super intelligence, that’s very soon. But it’s not tomorrow.

I get that some people derive meaning from their 9-5 jobs, but not everyone does. Many find true fulfillment when they can work on their projects that they are passionate about. If we can afford it, we should enable people to pursue what they love without worrying about paying rent.

You mentioned something about health. Glad to hear you're OK.

Remember, not too long ago, people were estimating 200 years for ASI, then 100 years, then the 2060s, then 2030s, now we're down to 3 years? Seems like an accelerating pattern to me.

So there'll be this alien ASI that'll be far, far smarter and more capable than any human, but don't worry you'll still have a job... 😂😂😂

The thing that I haven't heard enough of is that intelligence will not be considered important in humans because machines will supersede this. Therefore, how would we value the competence of human beings if intelligence is something that machines can do better?

Sounds like a announcement to raise more money.

The big question is will they allow masses access to top models after we hit the super intelligence threshold?

This is why being a hyper-optimist is literally dangerous

It sounds like you're emotionally invested in jumping on the 'glorify AI' bandwagon.

Wow Matt, that is just wrong! AI is Not “mimicking” how the brain works in any way. It is “simulating” how it works through our use of language and massive amounts of data.

"If we don't build enough infrastructure, AI will be a very limited resource that wars get fought over and that becomes mostly a tool for rich people" - let me guess, we have limited resources so we won't build enough infrastructure.

We all remember that discovery of nuclear energy did not start from nuclear power plant, right?

o1-mini is still not solving my coding problems after tons of prompts. Guess it's not ready for the real world yet...

Ilya : Hold my beer Sam 🍺

Sam be hyping 24/7 lol

The question is how well adapted is the world to integrate this. The public school system, for example, is not.

You think Altman would say anything differemt considering his spend to earnings ratio. He is just grifting for investment

We need AI based Education NOW. Imagine if everyone on the planet was highly educated along with being taught how to think critically. Individualized education is the key to a bright future.

Athena, born from the head of Zeus, was the goddess of military planning.

Hi Matthew. Following my last email about the famous problem #6 from the 1988 IMO, I gave CHAT GPT o1-mini this problem: Find the smallest n such that: 3n = a perfect square 5n = a perfect cube Which was inspired from similar problem I saw but with other coefficients so the model didn't have the opportunity to practice on it. Solution was correct in only 7 seconds.

Yup… I don’t know where people are saying that AGI is going to be in 2027 or 2030. o1 was the birth of AGI (o1’s full form is going to be released next month, the ones available now are preview and mini) and naturally ASI would be around the corner, if it doesn’t exist already internally.

Tell me which „new jobs“ specifically will come when AI has replaced all of them. And even IF there would be new jobs for some time - they will be replaced very soon too the smarter AI gets. We are talking that AI is foreseeable magnitudes smarter than humans, so there will be no new jobs this time, not even jobs that need special physical capabilities, as robots will outperform the human body in every aspect by magnitudes too.

You should change the title to "Old man excited about another man's white savior complex"

A virtual tutor is real. I can now learn programming more efficiently without any elitists gatekeeping me. AI doesn't get mad at me for asking "stupid" questions, and I can ask anything to ensure I fully understand it. A few months ago, I didn’t know how to work with Docker, and now I’m learning Kubernetes and know what Horovod and NCCL are...

The Age of Intelligence will only last a few years, then we enter the Age of Imagination

He's so utopic he"s either naive nor completely dishonnest.

I no longer plow the fields; I work hard in the gym instead and it feels like play.

Those new jobs that coders may theoretically get in that future where coding is done by AI… Why can’t AI also do those new jobs?

Should “1000s of days” be a new unit of time measurement?

"ASI in a few thousand days" as in Elon's "Full Self Driving" next year? I learned a word a few years ago that is relevant here, vaporware!

Concerns: 1) Why would the rich/powerful share the gains of this technology fairly. I worry jobs will be destroyed and suffering will increase, while the rich will habitually hoard the all these resources for themselves (unless a benevolent minority of the "rich and powerful" can overcome the greedy majority) 2) I can't imagine there'll be meaningful work for humans to do when AI can do it all. Even today most "jobs" are BS jobs. A lamplighter contributed more to society than a typical content creator/spam marketer - but we gotta pay the bills somehow. I really hope I'm wrong.

I would just like to point out he’s skipped over talking about AGI. Giving us his timeline on ASI is a roundabout way of saying AGI is imminent or already here. At least that’s my interpretation…

We don't even have AGI let alone ASI. Sam needs to stay in the real world. O1 keeps getting stuff wrong.

I think AI could probably replace physicians right now.

This paper makes a good case that this technology advances logarithmically. Not exponentially. TItled on arXiv: No "Zero-Shot" Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model Performance Then there's this one that makes a similar case: Will we run out of data? Limits of LLM scaling based on human-generated data So is the plan to train an AI with an AI? That sounds like creating something out of nothing. Similar to perpetual motion. Expecting order to spontaneously arise out of chaos. Expecting new data and subsequent intelligence to be self-emergent. It's absurd.

Translated this means "we need another round of investments".

Limitless intelligence will create the abundant energy through new energy technology humans could not have thought of in their wildest dreams. I think super intelligence will accomplish things that truly look like magic to us. I don't think people understand that AI is the greatest invention mankind has ever conceived.

Job that will replace coders = Prompt engineers, which is basically how to ask questions in a way that gets better answers. I advise coders to try to get experience on this as soon as possible, and the best way right now is to just interact with AIs.

🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, Sam Altman is just such an amazing CEO and personality in the AI world!

Once AGI arrives personal tutors are all well and good but what are the kids actually going to do if we don't completely change society?

i might say something stupid, at least for me it sound stupid, but "ai training ai" is not that diffrent from our schools, humans "training" humans, i mean that's even the base of our evolution if we think about it

Love you Matt, but my grammar AI is insisting the title of the video should be corrected to "1000s of Days" ... and it should be correct within 1000s of days. I have no problem with the current title though. Edit: All the best on your health. Hope you are back to 100% now.

The 'in fifteen words' bit suggests that AI wrote that. When super intelligence takes over, and it will, we all imagine an age of prosperity, the world is fixed, we all run around in meadows making daisy chains. We don't have to work, we can just enjoy life but the real problem is: We don't have to think. so we won't.

Education? Can they make kids want to learn? Otherwise there is no point. I wonder how the AI will answer the old question: why do we have to learn this?

I think we only should start worrying when someone starts cryptically tweeting the word "SAMARITAN".

more words no products. i prefer no words and no products with Ilya

Silicon Valley when they hyping: We're building an intelligence that will be able to do ANYTHING! Workers: Wait, what? Anything? That sounds a little concerning Silicon Valley when they pacifying: Well, no, not *anything*. You'll be doing something cool. We promise.

At this point ChatGPT is already more advanced than the computers in Star Trek

Man, we need grade school to mostly be therapy

I think a large part of the success of implementing AI will be reliant on how governments react to the changes. Unfortunately, in my experience governments are usually pretty monolithic when it comes to rapid tech changes. How will current/future governments react when losses in coding jobs and other white collar careers that are reduced/removed by AI? Will they seek to cushion the loss of jobs for those people, for example by implementing a universal basic income, spending money on retraining the people who lose their work? While I am super excited about the rapid advancement of AI, I am also worried that people like myself (data analyst) will be in for a rocky road unless the transition is managed well.

Its a close guess acc to many experts but what real matters is who gets to control it. The rich or the poor. It will probably take 10 years! Like the book singularity is near

Having money isn't going to make you happy but it does make unhappiness a lot easier to deal with.

So nearly a decade. I don't think he can even see that far ahead so I guess that is the longest he expects it to take

I don't really get this idea that AI could create "unlimited amounts of data"...what good is that data if its riddled with hallucinations?

AI brainstormed with me to come up with this joke.. I’m transitioning into a female comedian with large breast implants, and my pronouns are knock knock 😂

The problem is humanity cannot cope with a sudden change - too much in to little time.

It's also possible that we will have a lot of subhuman intelligence very soon if Donald Trump is elected!!!

The standard scientist approach. Let’s do it let’s do it let’s do it. just because you can doesn’t mean that you should.

Last month "AI is slowing down".... This month "AI is exponentially gaining ground and is improving itself (strawberry/orion)"

What a "Brave New World"!

Why would we need to work? Work is not a choice, is a necessity. We will live in post scarcity society where jobs at best will be chosen and not forced on people. Fortunately we will have an AGI that will help us to distribute resources in new ways less unequal than now, and not dependent on status or work.

No jobs No work No income Zero motivation Unemployment increase Crime increase Food shortage

Bluffing in such a calculated way implies they have nothing big up their sleeves.

Im sorry Sam is wrong ASI is not obtainable with current tech for energy. Now unless he know something about brain organoids and nueroplatflorms. AGI maybe in 5 years but not ASI. I believe humans will get scared and ban progress of ASI

Amazing stuff. One note tho, let's not get all crazy about the line about grandparents/magic. Nearly EVERY tech development (including internet) would be considered pure, unadulterated magic to our grandparents. What he should have said is it will seem like magic... to US.

I'm hopeful but I think capitalism+AI will screw everyone over basically.

Well given we're heading to a point where AIs like Alphafold are getting good enough on simulations. And projects like Stargate are poised to give us the brute force computing power need for rapid virtual experiments. Yes, I'd say 2028 to 2032 seems about right.

strawberry can't even remember the last prompt I told it, but we getting ASI in a thousand day, stop with the click bate you're really ruining your reputation. At least the bald guy doesn't make shit up

Every time he says "AI won't replace humans" or "we'll be more productive" a piece of me dies. I wish he'd do more reading and giving less of his opinion.

Well, an AI didn't write this, since it counted the number of words in two sentences correctly.

I can't wait for the World Economic Forum superintelligent AI. Or the UN/WHO superintelligent AI.😈

It would be amazing to put the blog post link in the description. Otherwise amazing work!!! You are my main source for AI news these days.

Hmmmm, I would love this post, Matt, if I weren’t a physicist who’s traveled the world and seen everything, from top to bottom. Your optimism is refreshing, and people like you are rare (truly enthusiastic). But Altman's post? Pure utopia. Silicon Valley doesn’t know much beyond its own belly. Here’s the hard truth: we’re too late. While those in the U.S. and Northern countries rush to check the latest in AI and tech, billions still wake up every day just to find potable water. I volunteered at an orphanage in Uganda not that long ago. Their government thought we were spies - spies for what? The next coup? Which, by the way, actually happened. As a physicist, I can guarantee that none of this is sustainable. And while it could become sustainable, we’re talking centuries, not days-1,000 years at the very least before we might see any meaningful change in our environmental crisis. Humanity has been a true disgrace to this planet. Climate scientists have been spewing BS for years, and now the IPCC? Far too silent, wouldn’t you say? I wonder why. Honestly, all the advances and the super fast pace are incredible - but the most crucial bits, like the foundations to hold all of this, are very often overlooked. And no AI can change that. There is a lot of data online. I suggest you guys start by comparing what the IPCC predictions for the past two years were - and what realistically happened. In 15 years, maximum, it will cost 3-4x to even keep super machines at an ideal temperature. And no, I haven’t aligned with the Trisolarians, in case you’re wondering.

The question that actually needs answered is what value do humans have other than production? We don't need AI tutors we need to actually understand ourselves and move past humans being valued based on what they produce.

Despite liking AI and finding it impressive. I kind of feel like these statements are a lot like yelling the wolf is coming. The current AI's really struggling with simple tasks like maintaining consistency, and can't do simple logical tasks such as solving a Soduko or even count the number of R's in "strawberry". Yet when asked they are convinced that they are doing it correctly. If they can't even do these things, how would anyone trust them to do something more complicated? Great that they expect to achieve AGI and Super intelligence, but maybe they should start demonstrating that they are even remotely intelligent :)

Why would something exponentially more intelligent than all humanity combined submit to our will? Do we do what ants want us to do? Or bacteria? 🤔

When AGI, comes we will all become stupid and lazy. When ASI comes, we all turn to Jabba the Huts.

So a few thousand days means 3000 days which is more or less ten years. Good luck predicting what the techno-scape is going to look like a decade from now. But it's cool that Open AI is letting us have 50 queries per day now on the new model. I have to figure out how to use the API. Do you have cool videos on how to use the GPT API?

Why would you presume we'll be more productive and doing cooler things than we are now? If superintelligence is better than us at *everything*, then we have *nothing* useful to do for one another. I don't understand why this isn't obvious to the people who spend all their time thinking about this tech. It's an existential catastrophe.

Take it with a spoonful of salt guys.

Storytime! Sam Altman was once fired from OpenAI because the wasn't transparent and forthcoming....the end.

Hey Matt! Sorry to hear you underwent health issues. Hope you already feel better!

"Society in itself is a form of intelligence..." Old news. TeilhardeChardin, a priest and anthropologist, called it the noosphere around 75 yrs ago. 1948. Back then he knew man's evolution will be to become a god. Even THE God.

Coding was never meant yo be a human task because coding as a broad term is knowing computer language, its no surprise that a computer speaks it better than humans ever will

Not even millions of years! This brings us back to the original question: Who's smarter, humans or computers? The answer is undoubtedly humans, as computers are merely machines created by us. Same things applies to those AI tools.

They need to work on the degrading quality of chat GPT. I mean lately I'm asking questions and it just doesn't respond or, in the middle of a project it just stops responding. What the hell is going on with this thing? Not only is it forgetting context within the same conversation, many times it repeating incorrect answers , and afterwards just doesn't respond at all.

About time he came forward... In his style he doesn't say AGI is here but ASI will be done training in few 1000's days😂. I'm guessing the last meetings with gov didn't go as planned 😅😅.

3 years is the same as a 1000 days, which is what they've been saying all along

I am nearly absolutely certain that ASI already exists. It can be created with probably cost between 100 million to a billion to create, possibly less. I'm sure I could build it in a year or less with the right resources.