www.lesswrong.com/posts/ZfbChZBXgje8T6Geu/excerpts-and-notes-on-mythos-model-car...
1 correction found
it wasn't real
Anthropic’s system card does not say the contribution was unreal. It says the contribution was real, but smaller or differently shaped than first understood, and often reflected execution of a human-specified approach.
Full reasoning
Anthropic’s own system card draws a narrower conclusion than this sentence does.
The official text says that after follow-up, "the contribution was real, but smaller or differently shaped than initially understood". It adds that in some cases what looked like autonomous discovery was actually "reliable execution of a human-specified approach."
So the correction here is not that Anthropic confirmed a major independent research contribution. It did not. But it also did not say the contribution "wasn't real." Their stated view is that the early claims were overstated or misattributed, not wholly nonexistent.
1 source
- Claude Mythos Preview System Card
“Early claims of large AI-attributable wins have not held up… When we followed up on each claim, it appeared that the contribution was real, but smaller or differently shaped than initially understood… In some cases what looked like autonomous discovery was, on inspection, reliable execution of a human-specified approach.”