https://nvidia.github.io/cuda-python/cuda-core/latest/ https://developer.nvidia.com/nvmath-python

Let me just code that up real fast.

Because they used data that needs to be tokenized differently, and didn't really tune the models for use on that data. That's not really a limit of LLM tokenization per se.

>> We did not evaluate strategies known to improve LLM performance, including ... retrieval augmented generation

Which is a shame because this is exactly the kind of use case RAG is supposed to be good for and they largely observed problems it's supposed to help with.

Looking at the authors, it seems to me they're all subject matter experts in medicine and digital medicine, but their conclusion is the one in support of medical professionals and they really don't seem to have tried that hard to get good deep learning results.

I've had nightmares every time I've seen a doctor in the US, frequently because of things not being coded correctly. So honestly I'd just love to see a rigorous study of how often the human staff is messing it up too.

When I was not employed, I paid for the gas to drive 90 minutes each way and spent an entire day not looking at other opportunities to be hired or other ways to improve my resume, only to be told by the director that although I did very well in the interviews, they would be getting back to me in 3 months when they found out if their hiring freeze was over then, because at the present time the opening was not yet truly open.

Have some respect for human beings, regardless of their employment status.

But also, I did work on a project once where a test would scrape code snippets from documentation and make sure they still worked as expected. It's bad enough to break compatibility in your CLI, etc. but sometimes you need to do it. It's really bad when you have tutorials, recipes, etc. and they bit rot.