from Hacker News

Solving physics-based initial value problems with unsupervised machine learning

by opnac on 5/16/25, 9:45 PM with 8 comments

• by CGMthrowaway on 5/20/25, 2:07 PM

  One of the points seems really important: Because AI researchers almost never publish negative results, AI-for-science is experiencing survivorship bias.

  AI will accelerate greatly the survivorship bias crisis we have already seen. Because there are so many more reasons to reject an AI-driven result.

  The distribution chart, which I assume was pre-AI, is really scary. It implies that 85-90% of results are never published. https://substackcdn.com/image/fetch/w_1456,c_limit,f_webp,q_...

• by acc_297 on 5/20/25, 12:56 PM

  Today this front page post has a companion front page post probably worth reading in tandem:

  https://news.ycombinator.com/item?id=44037941

• by staunton on 5/20/25, 9:13 AM

  This paper is solving (basically) high-school-level problems by training neural networks on the "obvious" cost function. All of those problems can be solved much cheaper by standard numerical solvers for ordinary differential equations. They don't even compare to standard methods.

  So what's the point? Riding the neural network hype?