Coding using @cursor_ai 0.45 with the @GoogleDeepMind (new) gemini-2.0-flash-thinking-exp model seems like the biggest step up in genai coding since Claude Sonnet 3.5 came out last June. This is unreal... forget about R1 folks - check out this new Gemini model! 🤯
The legendary @rao2z explains the difference between reasoning and memorisation using the example of the famous "why are manhole covers round?" interview question. Which do you think LLMs do?
This is @fchollet discussing deep learning guided program synthesis, reasoning, o-series models and the ARC challenge. We will drop the full video hopefully later today!
"o3 is the alexnet moment for program synthesis" - today is a pretty big day folks. We need to update our priors. o3 is zero-shotting extremely high scores on ARC-AGI which is a step towards the spirit of "developer aware generalisation" which Chollet originally envisioned.
o3 is really special and everyone will need to update their intuition about what AI can/cannot do.
while these are still early days, this system shows a genuine increase in intelligence, canaried by ARC-AGI
semiprivate v1 scores:
* GPT-2 (2019): 0%
* GPT-3 (2020): 0%
* GPT-4
Today @GoogleDeepMind released AlphaEvolve: a Gemini coding agent for algorithm discovery. It beat the famous Strassen algorithm for matrix multiplication set 56 years ago. @Google has been killing it recently. We had early access to the paper and interviewed the researchers.
Iman Mirzadeh from Apple @i_mirzadeh wrote the famous GSM-Symbolic paper a couple of months back which argued that LLMs are learning surface statistics and not genuinely reasoning due to their sensitivity to distractors and out of distribution examples.
We just released our interview with the father of Generative AI - @SchmidhuberAI!
The G, P, and T in "ChatGPT" (GPT means "Generative Pre-Trained Transformer") go back to Juergen's work of 1990-91 when he published what's now called "Unnormalised Linear Transformers,"
Neural Networks are not Turing Machines/Turing Complete (in any practical sense). Probably the most pervasive thing which (apparently!) nearly everyone misunderstands about neural networks. This means that the space of expressible programs in a NN, is a tiny fraction of that of a
We just released @fchollet keynote talk from #AGI24. He challenges LLM hype, discusses ARC-AGI benchmark, and proposes merging deep learning with program synthesis for true AGI progress.
Hot-take on Gemini
1) Surprised that Google managed to produce something roughly on GPT-4 level. They might be the only company other than OpenAI who will though.
2) The technology is SATURATING hard, scaling laws != competence
3) The elephant in the room is the AUTONOMY GAP -
