Ask HN：人工智能研究更需要“世界模型”还是更大的语言模型？
1 分•作者: unjuno•大约 23 小时前
生成假设似乎已不再是主要问题。
更棘手的问题可能是持续收集真实世界数据、构建准确的世界模型并利用它们来指导实验。
自动化研究是否正朝着这个方向发展？
查看原文
It seems that generating hypotheses is no longer the main problem.
The harder problem may be continuously collecting real-world data, building accurate world models, and using them to guide experiments.
Is this where automated research is heading?