Remix.run Logo
Agentic AI and the next intelligence explosion(arxiv.org)
18 points by silverpiranha 2 days ago | 3 comments
d_silin 2 days ago | parent | next [-]

The paper is very light on details and doesn't offer any evidence of the stated hypothesis beyond referencing other work.

ed_li 2 days ago | parent | prev | next [-]

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

lmaowutok 2 days ago | parent | prev [-]

Too poorly written to take seriously. An LLM won’t help - the ideas are elementary