Rephrased: does G "emerge" from math + code RLVR? Or do you only get higher math + code skill and nothing else? Maybe G is a skill and it is isomorphic to math + code problem solving?
@fchollet
-
Jagged frontier: RLVR math/code vs human-data bottleneck
By
–
What if the jagged frontier is mainly math + code (which you can push arbitrarily far with RLVR), and everything else starts to plateau because it is still bottlenecked by human generated data? Model performance in non-verifiable areas has kept improving steadily, albeit much
-
65% of US population uses generative AI weekly
By
–
About 65% of people in the US use generative AI several times a week. That's about 72% of those who use the Internet, so actually quite close to saturation. They just don't pay for it.
-
Artificial consciousness should not be a goal of AI research
By
–
Also, to be clear, creating artificial consciousness should *not* be a goal of AI research. Understanding consciousness, sure, synthesizing it, no.
-
Argument on AI consciousness and computational substrate
By
–
Arguing that "consciousness arises from a computational substrate, and since AI models are computation, they are likely conscious," is as meaningless as arguing that "living creatures are made of atoms, and since this rock is made of atoms, it is likely alive." Everything is
-
AI consciousness debate: processing vs. sentience
By
–
Side note. The idea that current models are conscious is absurd and downright offensive. Do not confuse competent information processing and sentience, they are wholly unrelated. You can be entirely incompetent at everything yet fully conscious, like a toddler, or
-
AI sentience and the dystopian risk of quantifying suffering
By
–
The notion that current AI models are sentient and can suffer, combined with the foolish idea that suffering can be mathematically quantified and weighted between humans and non-humans, could lead us down an incredibly dark and dystopian path. But before it gets to that point,
-
Three key AI paradigm changes: inductive learning, tool use, code training
By
–
I think there are three extremely important changes: 1. Switch from transductive paradigm to inductive paradigm (by far the biggest jump)
2. Symbolic tool use, code execution, harnesses
3. Training models on code execution & harnesses, so that they start to encode symbolic -
Keras community call: pluggable backends, MLX and PaddlePaddle support
By
–
Summary of the Keras community call — 1. New pluggable backend plan, upcoming MLX and PaddlePaddle backends; new backends are installed as add-on libraries 2. Reducing number of ops needed to ship new backends; adding backend-agnostic implementations of all secondary ops 3.
-
Training models to be inductive and perform induction at test time
By
–
It should not matter how the inductive reasoning is encoded. What matters is that you train the model to be inductive and that you let it perform induction at test time.