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.
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.
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.