So many people misremember (or never read) what I said in in 2022 in “Deep learning is hitting a wal...
TL;DR · AI 摘要
Gary Marcus澄清了2022年关于深度学习的文章误解，强调单纯扩大LLM规模不足以实现AGI，需结合符号工具发展神经符号AI。
核心要点
- 单纯扩大LLM规模不足以实现AGI
- 需结合符号工具发展神经符号AI
- Claude Code是神经符号混合系统
结构提纲
按章节快速跳转。
思维导图
用一张图看清主题之间的关系。
查看大纲文本（无障碍 / 无 JS 友好）
- Gary Marcus澄清误解
金句 / Highlights
值得收藏与分享的关键句。
单纯扩大LLM规模不足以实现AGI，需结合符号工具发展神经符号AI。
Claude Code是神经符号混合系统，结合了代码解释器等符号工具。
Rather, it was an argument that the pure of scaling LLMs would not get us to AGI, and that we would need to" / X
Gary Marcus on X: "So many people misremember (or never read) what I said in in 2022 in “Deep learning is hitting a wall”, which was neither about revenue or AI’s potential upper limits. Rather, it was an argument that the pure of scaling LLMs would not get us to AGI, and that we would need to" / X
Don’t miss what’s happening
So many people misremember (or never read) what I said in in 2022 in “Deep learning is hitting a wall”, which was neither about revenue or AI’s potential upper limits. Rather, it was an argument that the pure of scaling LLMs would not get us to AGI, and that we would need to start including symbolic tools and working towards neurosymbolic AI. Time has shown I was quite right about all of that, and that’s what happening now, with systems like Claude Code are neurosymbolic hybrids that leverage symbolic tools like code interpreters and harnesses written in symbolic code, etc. People ought to go back and read that paper. Despite many people mocking the title, the substance of the article was 100% correct. (Those who thought that scaling pure LLMs would be enough were wrong.)
·
20
9
90
17
Read 20 replies