17 points | by silverpiranha 3 hours ago ago
3 comments
The paper is very light on details and doesn't offer any evidence of the stated hypothesis beyond referencing other work.
this position paper is actually bang on a direction we've been working on for the past year — scaling many specialized agents together with RL instead of just scaling one big model.
the thesis is that intelligence scales through interactions, not just individual capacity. E.g., prior work: https://x.com/t_ed_li/status/2038763049574879351
Too poorly written to take seriously. An LLM won’t help - the ideas are elementary
The paper is very light on details and doesn't offer any evidence of the stated hypothesis beyond referencing other work.
this position paper is actually bang on a direction we've been working on for the past year — scaling many specialized agents together with RL instead of just scaling one big model.
the thesis is that intelligence scales through interactions, not just individual capacity. E.g., prior work: https://x.com/t_ed_li/status/2038763049574879351
Too poorly written to take seriously. An LLM won’t help - the ideas are elementary