News for hackers. Stuff that matters.
Re:DeepSeek-R1 by andrewla on Tuesday January 21, 2025 @11:40AM (#42782120)
The amazing thing continues to be that they can ever answer these questions correctly.

It's very easy to write a paper in the style of "it is impossible for a bee to fly" for LLMs and spelling. The incompleteness of our understanding of these systems is astonishing.

Re:DeepSeek-R1 by hcurtiss on Tuesday January 21, 2025 @09:42PM (#42788289)
Is that really true? Like, the data scientists making these tools are not confident why certain patterns are revealing themselves? That’s kind of wild.