Gary Marcus:LLM 的树搜索其实源自前 LLM 时代的经典符号 AI
Gary Marcus 这条推值得刷到:他把 LLM 的成功往回追到 1997 年 DeepBlue 那套树搜索。做模型的、写 AI 史的、或者总听人吹 LLM 全新范式的,都该看看这条溯源。
Gary Marcus 在 X 上发文指出,用树搜索暴力求解重要 AI 问题早于 LLM 诞生,1997 年 IBM 的 DeepBlue 击败国际象棋世界冠军就是典型例子。
他认为多数人不了解这段 AI 历史,并称当今 LLM 表现更好,部分原因是模型借鉴了经典符号 AI 中的树搜索方法。转发的讨论提到 AI 在高维搜索空间中的暴力计算能力,但强调定义核心思想、赋予意义的仍是人类。
reality: people have been brute forcing important AI solutions with tree search since before LLMs were born. DeepBlue, for example, in 1997. Sad that most people have no idea about the history of AI. Footnote: part of the reason that LLMs work better nowadays is because they finally learned to borrow tree search from classical symbolic AI. Kekius Maximus @Kekius_Sage AI can brute-force through high-dimensional search spaces faster than we can blink. First was Navier Stokes, now is Riemann hypothesis. Still, only a human brain can define the idea that changes how the next generation thinks. Human value does not shrink when AI gets smarter. We just move from being the calculators to being the meaning-makers. 🔗 View Quoted Tweet📊 6 ⚡
This report is categorized under AITOP's "Industry" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Key entities and technical topics under continuous tracking include Gary Marcus, DeepBlue, LLM, tree search.
Source attribution: curated from Gary Marcus (@GaryMarcus).