Marc Andreessen on X: Richard Sutton Is a Genius—And We Lack Definitions for ‘Novel’ and ‘Discovery’
TL;DR · AI Summary
Andreessen cites Sutton’s view that pre-AI, we lacked clear definitions of ‘novel’ and ‘discovery’, leading to AI-generated innovations that remain unnamed; Sutton argues supervised-learning-based generative AI cannot make true discoveries.
Key Takeaways
- Sutton claims: Supervised-learning-trained generative AI cannot make novel disco
- Andreessen observes: AI will generate many new things, but without precise defin
- Sutton distinguishes: AI creativity is recombination—not causal reasoning—thus l
Outline
Jump quickly between sections.
Andreessen introduces Sutton’s perspective, emphasizing the lack of formal definitions for ‘novel’ and ‘discovery’ before AI—and its consequences today.
In his video, Sutton asserts that supervised-learning-trained generative AI is fundamentally incapable of novel discovery, only recombining known data.
Without prior definitions of ‘novel’ and ‘discovery’, AI-generated artifacts will remain ambiguously labeled or unnamed.
Sutton contrasts AI’s statistical recombination with human causal reasoning—highlighting why AI lacks true discovery capacity.
Mindmap
See how the topics connect at a glance.
查看大纲文本（无障碍 / 无 JS 友好）
- AI 创造力与发现的本质边界
- Sutton 的核心论点
- 监督学习 AI 无法新发现
- 仅能组合已有知识
- 命名困境成因
- AI 前缺乏‘novel’/‘discovery’明确定义
- AI 后产生大量无名新物
- 创造力类型对比
- 人类：因果推理 + 突破性发现
- AI：概率重组 + 模拟创意
Highlights
Key sentences worth saving and sharing.
Sutton: 'Generative AI trained by supervised learning is incapable of making novel discoveries.'
Andreessen: 'We will use AI to create many new things and never quite know what to call them.'
Sutton’s key insight: AI creativity is about recombination, not causal reasoning — thus no true discovery.
Marc Andreessen 🇺🇸 on X: "Richard Sutton is of course a genius and a legend, and so self-recommending. My own view is that we didn’t have good definitions of either “novel” or “discovery” prior to AI, and so we will use AI to create many new things and never quite know what to call them." / X
Don’t miss what’s happening
Marc Andreessen 
](https://x.com/pmarca)
Richard Sutton is of course a genius and a legend, and so self-recommending. My own view is that we didn’t have good definitions of either “novel” or “discovery” prior to AI, and so we will use AI to create many new things and never quite know what to call them.
Quote
Richard Sutton
@RichardSSutton
·
4h
A new and possibly controversial perspective: In this video, I explain the sense in which generative AI trained by supervised learning is incapable of making novel discoveries. https://youtu.be/K5LAFEjTlBA The text of the speech: AI Creativity and Discovery Good day ladies and
38
17
305
168
Read 38 replies