Machine Learning

Very cool application of LLM to screen potential protein-drug interactions.

This is an effort to get some discussion going.

I remember starting grad school and coming across reddit posts with themes like, "What research area will be hot in the next 10 years?", etc. In retrospect, the comments there were not very informed (talk of graphical models and bayesian non-parametrics). But, the heart of these posts is talking about a research area that you find exciting.

So, tell us what research area is currently exciting to you. Are you starting a new job, project, or graduate program to work on it?

I am looking to replace my old PC, and wondering what other people use.

Do you use your own hardware? If so, what do you have? What do you think gives you the most bang for your buck at the moment?

Do you use the cloud instead? If so, why? Which service(s) do you use?

Thank you!

Microsoft released a paper on progressive learning and fine-tuning using chain-of-though style data produces impressive results with little compute resource.

From the paper:

Our research indicates that learning from step-by-step explanations, whether these are generated by humans or more advanced AI models, is a promising direction to improve model capabilities and skills.

cross-posted from: https://lemmy.blahaj.zone/post/74156

From the latest commits:

We are happy to release our final 1T token version of OpenLLaMA 3B and 7B. We’ve updated the evaluation results. We are also happy to release a 600B token preview of the 13B model, trained in collaboration with Stability AI.

Haven't tried it yet, and the 13B model is still in the works, but hopefully this will be a better foundation than the leaked Meta AI model, not only for more reproducible research, but because nonacademics will be completely in the clear from a legal standpoint to run this stuff locally.

Interesting article, I wonder how much longer new language models will be detectable like this.

Weka implements a ton of statistics-based ML algorithms as well as some validation tools and graphs. All you need is some data in almost-CSV format and you can run some statistics analysis on it. This isn’t neural networks so you don’t need neither a powerful GPU nor gigabytes of data. Some tutorials online get useful results with 10-20 entries.