Press Space to continue
Press Space to continue
Humans provide a fun comparison point here. In the 80s, the "cognitive consequences of programming" was a major psychological research focus. The prevailing hypothesis was that learning formal logic and algorithmic structures through computer programming would act as a form of mental gymnastics that could upgrade general reasoning, planning, and novel problem-solving abilities. But foundational research, followed by subsequent meta-analyses, demonstrated that this does not happen. Students who…
François Chollet@fchollet1d ago
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 slower than for math and code. But is that steady improvement a side effect of a higher G (itself driven by RLVR), or only a function of the amount of new human data getting injected into training (which is still continually happening on a massive scale)? A lot of things depend on the answer to this question