Marcus戳穿了纯规模扩展的泡沫，做AI研究或关注AGI路线的人，看完会对当前方向产生反思。
Gary Marcus在推文中回顾其2020年arXiv文章《AI的下一个十年》，指出若当时采纳其建议，AGI可能已经实现。他认为过去三年纯规模扩展的路线是弯路，而神经符号AI（如Claude Code、DeepMind解决9个Erdos问题的系统）已取得进展。但其他三个目标——机器可解释知识库、可靠推理系统、显式世界模型——进展甚微。他呼吁聚焦这些方向，AGI或可在下一个十年内实现。
If we had done everything I suggested in my 2020 arXiv article “The Next Decade in AI”, we might act...
If we had done everything I suggested in my 2020 arXiv article “The Next Decade in AI”, we might actually have reached AGI by now. In the last three years, after a detour driven by the false promise of pure scaling, we have gone a long way to the first of the four goals I laid out: neurosymbolic AI. (Claude Code; DeepMind’s new math system that solved 9 Erdos problems, etc). Harnesses and tools, for example. But the field has made little progress on the other three: - databases of explicit, systematic, machine interpretable knowledge [LLMs are a partial but inadequate/unreliably substitute] - reliable reasoning systems that can work with partial and incomplete information - the ability to represent and induce explicit world models (which I called cognitive models in 2020) that AI systems can reason over. An excess commitment to pure scaling delayed all of these. If we focus on the other three AGI may be achievable, possibly in the next decade. 💬 4 🔄 0 ❤️ 5 👀 1162 📊 5 ⚡