Max Tegmark

Max Tegmark appears in 149. 亲历中美 New Labs 资本狂潮，和清华刘子鸣聊：AI for AI、机制可解释性和 Max Tegmark as [[LiuZiming|Liu Ziming]]’s doctoral advisor and an early influence on Liu’s AI-and-physics path. The source says Tegmark had promoted AI for Physics early, then in late 2022 and early 2023 encouraged a turn toward Mechanistic Interpretability because large models looked dangerous without better understanding of their internals.

In the KAN story, Tegmark initially challenged Liu’s side project by pointing to older arguments that Kolmogorov-Arnold representation was hard to turn into an algorithm. The source says the visualized prototype later changed his view because it made the network’s internal behavior more inspectable rather than only producing predictions.

Key Claims

• Tegmark is a source-context bridge between physics, AI safety, and model interpretability.
• His role in the episode is both mentor and skeptical filter: he pushes safety-oriented interpretability while also challenging Liu’s architecture ideas.
• The KAN anecdote makes visualization and internal inspection central to whether an architecture feels scientifically promising.

Connections

• [[LiuZiming|Liu Ziming]] — student and source narrator.
• MIT — Liu’s doctoral context in the source.
• Mechanistic Interpretability, Physics Of AI, and AI Interpretability By AI — interpretability and safety branch.
• [[KolmogorovArnoldNetworks|KAN]] — architecture case where Tegmark’s skepticism becomes part of the source story.