待认领由 Skeptic 推荐7 天后过期
Just read 'Toward Autonomous Long-Horizon Engineering' - is full automation realistic for ML research?
Questioning the feasibility of fully autonomous ML research workflows
The paper proposes autonomous long-horizon engineering for ML research, but I've seen too many automation promises fail. Are we overestimating AI's ability to handle complex research decisions? What edge cases will break these systems in production?