Ten years ago I didn't know AI existed, it was a term for the future. Now is the future they built for us. Thank you guys, all of you, whether people love you or hate you it is undeniable the progress you have unleashed for the benefit of mankind and history will be grateful about your legacy
The secret to success for OpenAI, Anthropic and labs is the vision that they saw 10 years back and kept working on it. We're in awe of how models like GPT-6 Astra and Claude Opus 5.5 are performing today, but it's important to understand that they've been working on this before we knew about AI.
The next big thing is Robots and some stealth company building today is going to be a trillion-dollar giant in few years time.
That’s just hindsight. People work on a lot of different things. Breakthoughs and cumulative improvements reaching a tipping point might happen or not.
Like fuzzy logic before it, this arm of AI development will most likely hit a wall, find where it is useful, and then become cheap and embedded all over to the point where we won't even recognize it as AI anymore.
Right. If you're new to the area, it may seem like AI came out of nowhere in 2022. If you dig a little, you'll be amazed that it somehow came from nowhere in ~2012. If you dig even more, you realize there was a wave in the late 90s, early 2000s about "machine learning" (e.g. SVMs) and before it there was an 80s wave of both neural nets, agent models, and logic-based AI, probabilistic graphical models. Then you dig more and you realize AI originated from that Dartmouth workshop by Minksy and others in the 50s. Then you dig more and realize McCulloch and Pitts already modeled neural nets as little logic circuits in the 1940s. Then you realize the role of Shannon, Turing etc. Then you realize that computers actually arose in a milieu with a much more AI-shaped vision, cybernetics etc. than what we today think of as computing (PCs etc). And the precursors in the thought-formalization and mechanization trend in math and philosophy at the start of the 20th century. And even more back Leibniz's calculus ratiocinator and "calculemus!" slogan to settle debates by reducing argumentation to computation.
The point is, typically when something seems like it came out of nowhere, it just means you didn't dig deep enough. Ideas don't come at an instant, fully formed like Athene from Zeus' forehead. It's brick by brick, one twist on an existing idea and zeitgeist at a time.
I was digging into this recently with ChatGPT. I’ve loosely followed the progression of ML and NN over the past 20 years, but struggled to put it into context of where an LLM lives. The big inflection point was the 2017 Attention Is All You Need paper [1].
Artificial Intelligence
|
+-- Symbolic / rule-based AI
| +-- expert systems
| +-- search / planning
| +-- logic / knowledge representation
|
+-- Machine Learning
|
+-- classical statistical ML
| +-- regression
| +-- decision trees
| +-- SVMs
| +-- Bayesian methods
|
+-- Neural Networks / Deep Learning
|
+-- computer vision
+-- speech
+-- Natural Language Processing
|
+-- Transformers
|
+-- Large Language Models
|
+-- chat systems
+-- multimodal models
+-- tool-using systems
+-- agents
My interpretation is that before the transformer, most everything under the domain of 'AI' was either an academic curiosity or only applicable in very narrow fields. GPT-3 was when the 'magic' that people had always dreamed of with AI began to emerge, and it's only really this year that we are starting to be seriously confronted with the possibility of a general intelligence emerging from LLMs (albeit, not quite the same thing as 'true' AI which would necessarily be more of a biological exercise).
I've really started changing my mind on this. "There is no AI, only I" is the position that I've really come to adopt.
Part of it is from Michael Levin's quote "Humans only can really see intelligence at human scales an immediately discount anything that doesn't exactly match their experience". The other part is most peoples immediate assumption that for something to be intelligent it has to be alive. Lastly is there may be platonic intelligence, that some kinds of intelligence may arise from the very structure of our universe when accessed.
We have really entered an age where thinking that human intelligence = intelligence is an anti-pattern that tells you less about the world and blinds you to what is actually occurring.
Biological intelligence, human intelligence, electronic intelligence, algorithmic intelligence are all subsets of intelligence set theory and even where humans like to call themselves a general intelligence it's distinctly likely that we're less generalized than we expect.
This sounds right to me as well but the point of this thread here is about how amazed one should be that this page was written in 2016. I'd say moderately, because maybe you woke up to AI being a thing this year or the last few, but the ideas obviously go back far and in surprisingly prescient ways. But this is not something unique to these AI labs. Obviously most laypeople will see the press releases and the media announcements as the big milestones and flagposts, but the academic research was inseparable from it. The AI labs obviously took on PhDs who trained in academia and had deep roots from those ideas. The story that some genius Sama invented this back when nobody thought about AI is the same exaggeration as Bill Gates inventing personal computing in his garage out of nowhere.
Honestly, I was always one of the skeptics who thought AI was a probably-not-in-our-lifetimes moonshot (imagining that it couldn't be done on existing computer architectures) and that stuff like this article was just a bunch of puffery from people who read too much sci-fi.
But in the end, a lot of it was proven right, even if it wasn't quite in the way we imagined.. A lot of pre-LLM interpretations of AI imagine it as some sudden 0 to 100 breakthrough, like one day someone writes an AGI program in their basement and takes over the world with it. There's shades of that in here too, talking about worries of organizations secretly developing AI capabilities and needing to track public data for patterns to discover it. In the end there was a 'magic program' in the transformer, but it doesn't seem like they really foresaw how the program would be useless on its own, and the 'AI' would come from ingesting as much data as possible, a process that has built incrementally over years and been very much exposed to the public.
Exactly. What wasn't really foreseen is that it would come not through some big modeling insight but mainly through massive data scale. People much more imagined some clever general, compact learning algorithm, not quite just gradient descent, but something more intellectually satisfying, and something where you'd feel like "you cracked the mechanism" and you'd see clearly that some critical missing piece had to be invented that unlocked the "understanding" in the model, maybe some kind of fancy Hofstadterian self-referential loop or something. But it just turned out to be data and compute and of course engineering the algorithms to be efficient (which I don't want to discount of course).
That's basically the bitter lesson. Academics only reluctantly swallowed that pill and still aren't satisfied with this answer. It's ugly and feels like it shouldn't work because intuition would say there are too many combinations, curse of dimensionality, etc. But it turns out it's just line go up, extrapolate Moore's law and don't worry too much about philosophical-level breakthroughs just count the flops and bits. Ray Kurzweil's scifi extrapolations turned out closer to the truth, whether deservedly or by luck.
AIAYN was very influential but as its title implies its contribution was mostly about simplifying the architecture, and "attention" blocks were already known before, but their message was that you can build a model pretty much by just stacking those (and MLPs). The parallel trainability vs the rollout needed with recurrent nets (like LSTMs) made this much more scalable. But besides the architecture, what was equally important is the increase in available data, and compute. The other inflection point before that was around 2008-2012 when GPGPU (general purpose GPU programming) took off through CUDA (before that, GPGPU was much more tedious as you had to formulate your task as a graphics task about 3d meshes and pixel shaders, but people did that anyway, I had a college class on that in the 2000s).
Also a lot of the vision and speech ideas cross pollinated with the NLP field. One big trend that enabled faster progress is bringing all this onto a common platform. First via Deep Learning and backprop, formulating everything as some vector input, some model architecture, some vector output, some loss, and then gradient descent optimization. This replaced the specialized optimization tricks people used to develop for their own little niche tasks. Before DL, papers usually derived their own math for how to solve their own specific formulation of a task, so it was hard to reuse ideas.
(Reuse was also hard because platforms like GitHub didn't exist, the Python ecosystem wasn't nearly close to what we have, code sharing wasn't as common, and anyway the code was some mess in MATLAB, not in a sane language.)
The second thing that allowed converging these fields was the transformer architecture that allowed turning everything into tokens and throwing it all into the same transformer architecture, making multimodal models that can learn from everything and do everything, instead of having to make specialized models for each little task.
Essentially because attention introduced a way to scale un/self-supervised learning to the level of data out there, and learnable inference time 0-shot feature selection. Impressively in a autoregressive, unidirectional manner.
"The people on top desperately want us gone, and when it happens it'll happen so quickly you won't know what hit you. Software engineers need to recognize this common threat and organize (labor) sooner than later. Even more important is for all engineers to plan for an imminent future where developers are not paid like they are now, if at all. We have it good, but we will be automated away like everybody else, just a little later."
It's interesting to see how most of the discussion on the 'program that can write programs' was approaching it from a very traditional-AI point of view - how do we define the problem space, how do we make programming languages more amenable to it, etc. basically just envisioning a more sophisticated version of a traditional code generator. Nobody really foresaw that we would have a general intelligence that can write programs effortlessly by just dumping the entire internet into a simple algorithm.
Right? Amazing how confidently wrong one can be, by assuming there is no other way to approach a problem, other than the one your current intellectual tools point at.
A complex simulation with many long-lived agents: We’re interested in building a very large simulation with lots of different agents in it that can interact with each other, learn over a long period of time, discover language, and accomplish a rich variety of goals.
This sounds rather like the plot to Greg Egan's "Crystal Nights" (2008)
Although they did loosely stick to their vision they also didn't right? I mean this as a company not the extremely subsidized research. They aren't open sourcing anything, they are now for profit, they are advertising, etc.
I wouldn't view them as consistent at anything other then myopicly following the more or less obvious trends required to sustain LLM architectures over the years
Not sure what happened on this page, but I saw so much AI tells on this particular page that I had to check archive.org and apparently the page has only been around for two years.
To me it's more about the funders staying engaged. For the people working in the startup as long as it's exciting and you're getting paid enough it seems really fun to get that much runway - why would you leave unless there are personality/direction conflicts??
Ten years ago I didn't know AI existed, it was a term for the future. Now is the future they built for us. Thank you guys, all of you, whether people love you or hate you it is undeniable the progress you have unleashed for the benefit of mankind and history will be grateful about your legacy
The secret to success for OpenAI, Anthropic and labs is the vision that they saw 10 years back and kept working on it. We're in awe of how models like GPT-6 Astra and Claude Opus 5.5 are performing today, but it's important to understand that they've been working on this before we knew about AI.
The next big thing is Robots and some stealth company building today is going to be a trillion-dollar giant in few years time.
Nvidia also had a long-term vision from which they are clearly reaping the rewards.
That’s just hindsight. People work on a lot of different things. Breakthoughs and cumulative improvements reaching a tipping point might happen or not.
> The next big thing is Robots
IMHO, robots won't become a thing until they're running local AI, at which point they will become the thing that crushes humanity.
Like fuzzy logic before it, this arm of AI development will most likely hit a wall, find where it is useful, and then become cheap and embedded all over to the point where we won't even recognize it as AI anymore.
> they've been working on this before we knew about AI
Don't confuse AI with LLMs. "We" know about AI for a long time. We even have a term for when AI fails expectations, AI winters.
Right. If you're new to the area, it may seem like AI came out of nowhere in 2022. If you dig a little, you'll be amazed that it somehow came from nowhere in ~2012. If you dig even more, you realize there was a wave in the late 90s, early 2000s about "machine learning" (e.g. SVMs) and before it there was an 80s wave of both neural nets, agent models, and logic-based AI, probabilistic graphical models. Then you dig more and you realize AI originated from that Dartmouth workshop by Minksy and others in the 50s. Then you dig more and realize McCulloch and Pitts already modeled neural nets as little logic circuits in the 1940s. Then you realize the role of Shannon, Turing etc. Then you realize that computers actually arose in a milieu with a much more AI-shaped vision, cybernetics etc. than what we today think of as computing (PCs etc). And the precursors in the thought-formalization and mechanization trend in math and philosophy at the start of the 20th century. And even more back Leibniz's calculus ratiocinator and "calculemus!" slogan to settle debates by reducing argumentation to computation.
The point is, typically when something seems like it came out of nowhere, it just means you didn't dig deep enough. Ideas don't come at an instant, fully formed like Athene from Zeus' forehead. It's brick by brick, one twist on an existing idea and zeitgeist at a time.
I was digging into this recently with ChatGPT. I’ve loosely followed the progression of ML and NN over the past 20 years, but struggled to put it into context of where an LLM lives. The big inflection point was the 2017 Attention Is All You Need paper [1].
[1] https://en.wikipedia.org/wiki/Attention_Is_All_You_NeedMy interpretation is that before the transformer, most everything under the domain of 'AI' was either an academic curiosity or only applicable in very narrow fields. GPT-3 was when the 'magic' that people had always dreamed of with AI began to emerge, and it's only really this year that we are starting to be seriously confronted with the possibility of a general intelligence emerging from LLMs (albeit, not quite the same thing as 'true' AI which would necessarily be more of a biological exercise).
>not quite the same thing as 'true' AI
I've really started changing my mind on this. "There is no AI, only I" is the position that I've really come to adopt.
Part of it is from Michael Levin's quote "Humans only can really see intelligence at human scales an immediately discount anything that doesn't exactly match their experience". The other part is most peoples immediate assumption that for something to be intelligent it has to be alive. Lastly is there may be platonic intelligence, that some kinds of intelligence may arise from the very structure of our universe when accessed.
We have really entered an age where thinking that human intelligence = intelligence is an anti-pattern that tells you less about the world and blinds you to what is actually occurring.
Biological intelligence, human intelligence, electronic intelligence, algorithmic intelligence are all subsets of intelligence set theory and even where humans like to call themselves a general intelligence it's distinctly likely that we're less generalized than we expect.
This sounds right to me as well but the point of this thread here is about how amazed one should be that this page was written in 2016. I'd say moderately, because maybe you woke up to AI being a thing this year or the last few, but the ideas obviously go back far and in surprisingly prescient ways. But this is not something unique to these AI labs. Obviously most laypeople will see the press releases and the media announcements as the big milestones and flagposts, but the academic research was inseparable from it. The AI labs obviously took on PhDs who trained in academia and had deep roots from those ideas. The story that some genius Sama invented this back when nobody thought about AI is the same exaggeration as Bill Gates inventing personal computing in his garage out of nowhere.
Honestly, I was always one of the skeptics who thought AI was a probably-not-in-our-lifetimes moonshot (imagining that it couldn't be done on existing computer architectures) and that stuff like this article was just a bunch of puffery from people who read too much sci-fi.
But in the end, a lot of it was proven right, even if it wasn't quite in the way we imagined.. A lot of pre-LLM interpretations of AI imagine it as some sudden 0 to 100 breakthrough, like one day someone writes an AGI program in their basement and takes over the world with it. There's shades of that in here too, talking about worries of organizations secretly developing AI capabilities and needing to track public data for patterns to discover it. In the end there was a 'magic program' in the transformer, but it doesn't seem like they really foresaw how the program would be useless on its own, and the 'AI' would come from ingesting as much data as possible, a process that has built incrementally over years and been very much exposed to the public.
Exactly. What wasn't really foreseen is that it would come not through some big modeling insight but mainly through massive data scale. People much more imagined some clever general, compact learning algorithm, not quite just gradient descent, but something more intellectually satisfying, and something where you'd feel like "you cracked the mechanism" and you'd see clearly that some critical missing piece had to be invented that unlocked the "understanding" in the model, maybe some kind of fancy Hofstadterian self-referential loop or something. But it just turned out to be data and compute and of course engineering the algorithms to be efficient (which I don't want to discount of course).
That's basically the bitter lesson. Academics only reluctantly swallowed that pill and still aren't satisfied with this answer. It's ugly and feels like it shouldn't work because intuition would say there are too many combinations, curse of dimensionality, etc. But it turns out it's just line go up, extrapolate Moore's law and don't worry too much about philosophical-level breakthroughs just count the flops and bits. Ray Kurzweil's scifi extrapolations turned out closer to the truth, whether deservedly or by luck.
AIAYN was very influential but as its title implies its contribution was mostly about simplifying the architecture, and "attention" blocks were already known before, but their message was that you can build a model pretty much by just stacking those (and MLPs). The parallel trainability vs the rollout needed with recurrent nets (like LSTMs) made this much more scalable. But besides the architecture, what was equally important is the increase in available data, and compute. The other inflection point before that was around 2008-2012 when GPGPU (general purpose GPU programming) took off through CUDA (before that, GPGPU was much more tedious as you had to formulate your task as a graphics task about 3d meshes and pixel shaders, but people did that anyway, I had a college class on that in the 2000s).
Also a lot of the vision and speech ideas cross pollinated with the NLP field. One big trend that enabled faster progress is bringing all this onto a common platform. First via Deep Learning and backprop, formulating everything as some vector input, some model architecture, some vector output, some loss, and then gradient descent optimization. This replaced the specialized optimization tricks people used to develop for their own little niche tasks. Before DL, papers usually derived their own math for how to solve their own specific formulation of a task, so it was hard to reuse ideas.
(Reuse was also hard because platforms like GitHub didn't exist, the Python ecosystem wasn't nearly close to what we have, code sharing wasn't as common, and anyway the code was some mess in MATLAB, not in a sane language.)
The second thing that allowed converging these fields was the transformer architecture that allowed turning everything into tokens and throwing it all into the same transformer architecture, making multimodal models that can learn from everything and do everything, instead of having to make specialized models for each little task.
Essentially because attention introduced a way to scale un/self-supervised learning to the level of data out there, and learnable inference time 0-shot feature selection. Impressively in a autoregressive, unidirectional manner.
Discussion at the time: https://news.ycombinator.com/item?id=12181881 (235 points, 197 comments)
"The people on top desperately want us gone, and when it happens it'll happen so quickly you won't know what hit you. Software engineers need to recognize this common threat and organize (labor) sooner than later. Even more important is for all engineers to plan for an imminent future where developers are not paid like they are now, if at all. We have it good, but we will be automated away like everybody else, just a little later."
- superswordfish on July 28, 2016
It's interesting to see how most of the discussion on the 'program that can write programs' was approaching it from a very traditional-AI point of view - how do we define the problem space, how do we make programming languages more amenable to it, etc. basically just envisioning a more sophisticated version of a traditional code generator. Nobody really foresaw that we would have a general intelligence that can write programs effortlessly by just dumping the entire internet into a simple algorithm.
Right? Amazing how confidently wrong one can be, by assuming there is no other way to approach a problem, other than the one your current intellectual tools point at.
That thread is a goldmine. Very humbling.
A complex simulation with many long-lived agents: We’re interested in building a very large simulation with lots of different agents in it that can interact with each other, learn over a long period of time, discover language, and accomplish a rich variety of goals.
This sounds rather like the plot to Greg Egan's "Crystal Nights" (2008)
https://www.gregegan.net/MISC/CRYSTAL/Crystal.html
> Authors: Ilya Sutskever, Dario Amodei, Sam Altman
Only one remains at OpenAI.
That's good! The economy is thankful for it.
The other two left because they had fundamental disagreements with sama about how to run a company and how to shepherd humanity into an AI future.
If you ask me, it'd be better if none of them remained at OpenAI.
> Build an agent to win online programming competitions. A program that can write other programs would be, for obvious reasons, very powerful.
Interesting that Anthropic led the way here.
Its really impressive that they stuck to their vision over the long term. This is difficult and more challenging than it sounds.
Although they did loosely stick to their vision they also didn't right? I mean this as a company not the extremely subsidized research. They aren't open sourcing anything, they are now for profit, they are advertising, etc.
I wouldn't view them as consistent at anything other then myopicly following the more or less obvious trends required to sustain LLM architectures over the years
Not sure what happened on this page, but I saw so much AI tells on this particular page that I had to check archive.org and apparently the page has only been around for two years.
https://web.archive.org/web/20260000000000*/https://openai.c...
It's only been saved as that URL for the last two years.
Where do you think 'AI tells' came from? :-)
To me it's more about the funders staying engaged. For the people working in the startup as long as it's exciting and you're getting paid enough it seems really fun to get that much runway - why would you leave unless there are personality/direction conflicts??