François Chollet

25.1K posts

François Chollet banner
François Chollet

François Chollet

@fchollet

Co-founder @ndea. Co-founder @arcprize. Creator of Keras and ARC-AGI. Author of 'Deep Learning with Python'.

United States Katılım Ağustos 2009
824 Takip Edilen623.9K Takipçiler
Sabitlenmiş Tweet
François Chollet
François Chollet@fchollet·
The 3rd edition of my book Deep Learning with Python is being printed right now, and will be in bookstores within 2 weeks. You can order it now from Amazon or from Manning. This time, we're also releasing the whole thing as a 100% free website. I don't care if it reduces book sales, I think it's the best deep learning intro around, and more people should be able to read it.
François Chollet tweet media
English
292
844
6.1K
822.3K
François Chollet
François Chollet@fchollet·
When the latest AI systems can't do something, there's a category of people who will immediately say, "well humans can't do it either!" - Then they stop saying it when AI improves a bit. Been hearing it for 4+ years, "humans can't reason either", "humans can't adapt to a task they haven't been prepared for", "humans can't follow instructions", "humans also suffer from hallucinations", etc. Until 2025 I was frequently told "humans can't do ARC 1 tasks either" (in reality any normally smart human would do >95% on ARC 1 if properly incentivized). Now that AI saturates ARC 1 they've completely stopped saying this.
François Chollet@fchollet

In general I've been sensing a new current deep learning maximalists recently, going from "our models can definitely reason" to "well our models can't reason, but neither can humans!"

English
32
7
121
14.4K
François Chollet
François Chollet@fchollet·
@zby To make it very short: reasoning generates causal models of the data, pattern matching uses associative/correlative models of the data.
English
6
1
13
359
Zbigniew Lukasiak
Zbigniew Lukasiak@zby·
I am curious about the difference you see, but for me both are mechanistic state transition under bounded resources. We are finite and computers are finite. We use the abstraction of infinite Turing Machine for computers - because it is a convenient way to think about the limits of machines that we can expand (add memory) - but at every point of time they are finite. You can say that reasoning is universal computation while pattern matching is just regular grammar - but the difference only happens at the limit.
English
1
0
2
408
François Chollet
François Chollet@fchollet·
This is more evidence that current frontier models remain completely reliant on content-level memorization, as opposed to higher-level generalizable knowledge (such as metalearning knowledge, problem-solving strategies...)
Lossfunk@lossfunk

🚨 Shocking: Frontier LLMs score 85-95% on standard coding benchmarks. We gave them equivalent problems in languages they couldn't have memorized. They collapsed to 0-11%. Presenting EsoLang-Bench. Accepted to the Logical Reasoning and ICBINB workshops at ICLR 2026 🧵

English
135
272
2.5K
209.6K
François Chollet
François Chollet@fchollet·
The fact that you need to provide a specialized harness clearly shows the model *does not* encode the kind of metalearning knowledge and problem-solving strategies that humans use. Humans solve novel problems without being told how to proceed step by step. AGI would *not* need a custom harness here. As an aside, the models still performed poorly at that point, they did not "crush" the task
English
3
0
8
468
François Chollet
François Chollet@fchollet·
"all reasoning is pattern matching" is a useless statement if you don't define "reasoning" and "pattern matching" first. You might as well say "all information processing is information processing." With grounded definitions of both, reasoning and pattern matching are *very* different modes of information processing.
English
2
0
40
2.4K
Zbigniew Lukasiak
Zbigniew Lukasiak@zby·
IMHO all reasoning is pattern matching - but the depth of that pattern matching is important. When the LLM does not know the language any reasoning about the program requires double depth - first about the language second about what needs to be done. This is supporting the same line of thinking as arxiv.org/abs/2602.01075 and arxiv.org/pdf/2509.21361
English
3
0
18
3.3K
François Chollet
François Chollet@fchollet·
You won't convince me that approaching a new programming language and working with it zero-shot is insurmountable. At my first job I had to work with a stack I had zero experience in (aside from Python) and I was shipping PRs in my first week. I had <1000 hours of programming experience in total by then.
English
10
2
97
9.8K
François Chollet
François Chollet@fchollet·
This is similar to how applying basic changes to how ARC tasks are encoded considerably degrades frontier model performance. If you're looking at the test for the first time, it really shouldn't matter what the encoding is. Unless you've studied specifically for the test, using a specific encoding
English
5
4
115
13.2K
François Chollet
François Chollet@fchollet·
Current AI is a librarian of existing knowledge. Science requires an explorer of the unknown. You don't win a Nobel Prize by staying in the library.
English
177
192
1.5K
70.9K
François Chollet
François Chollet@fchollet·
There is a poetic depth to the term "latent space" that transforms vector coordinates into a frontier of pure possibility
English
42
65
750
45K
Chris
Chris@chatgpt21·
@fchollet What do you think would fix the fundamental issues of the parametric learning paradigm.
English
5
0
39
6.4K
François Chollet
François Chollet@fchollet·
The next major breakthrough will branch out at a much lower level than deep learning model architecture. It will be a new approach. A better model architecture can lead to incremental data efficiency & generalization gains, but it won't fix the fundamental issues of the parametric learning paradigm.
Rohan Paul@rohanpaul_ai

Sam Altman just said in his new interview, that a new AI architecture is coming that will be a massive upgrade, just like Transformers were over Long Short-Term Memory. And also now the current class of frontier models are powerful enough to have the brainpower needed to help us research these ideas. His advice is to use the current AI to help you find that next giant step forward. --- From 'TreeHacks' YT Channel (link in comment)

English
101
54
879
140.4K
Draxler
Draxler@CausalEngineer·
@fchollet The psychologist Erik Erikson called this stage (Autonomy vs. Shame and Doubt.)
English
1
0
2
548
François Chollet
François Chollet@fchollet·
Knowing how to do a task is not proof of intelligence, but in reverse not knowing how to do it isn't proof of lack of intelligence. Intelligence is the rate of increase of your competence as a function of your resources/training, and that rate is independent of absolute levels.
English
65
57
697
35.6K

Keşfet

@zby @chatgpt21 @CausalEngineer @elonmusk @BarackObama @taylorswift13 @cristiano @BillGates