>>17066449
Do "scientists" not understand that any help anyone ever gave them against the problem, the AI just makes the best solution?

>>17066449
Schwartz had Claude reproduce results from one of his own physics papers. Claude did it in about 20 minutes. Schwartz says writing the original code had taken him weeks. Claude also pointed out a more efficient method he hadn't known about.

Claude then helped calculate 30 difficult particle-physics problems. Half reproduced known results using the new method; the other 15 had never been calculated before.

In ecology, Claude recognized that a mathematical tool from physics could be used on a problem researchers had been unable to solve at useful scale for about 20 years. Applying it to a heavily studied Panama forest showed that its tree-species mix changes about 4.5× faster than a major existing theory predicts.
In genetics, the team used Claude to analyze 5.7 billion pairs of nearby human mutations.

They found evidence that a DNA-shuffling process called gene conversion matters enough that some common genetics analyses may need to account for it.

In economics, Claude helped process the replication material from 4,452 published papers across five major journals. It converted roughly 30,000 pieces of research code from proprietary software into open-source code and checked the published numbers against what the code actually produced.

In linguistics, Claude helped build a database covering word stress across 6,072 languages, backed by a bibliography of about 160,000 research works. These kinds of databases are normally assembled manually and usually cover only hundreds of languages.

In mathematical physics, the team says Claude helped solve the last unsolved member of a family of problems George Watson began studying in 1939.

Other projects included models of ancient changes in Earth's atmosphere, gene activity inside individual cells, the life cycle of sunspots, and tools for analyzing the large-scale structure of the universe.

>zero progress on open physics problems
Yawn

>>17066452
>across 6,072 languages
name them or I call BS

>>17066452
Links, references?

>>17066463
https://www.anthropic.com/research/yes-claude-can-do-nine-loops

>>17066449
so there's like, one really desperate AIslopper spamming all these threads, isn't there?
that's sad.

>>17067145
Not an open physics problem, just computational dirty work nobody wants to do
Try again