New collection testing if/how giant language fashions convert questions on a dataset into code that runs on-the-fly to research it, all net and client-side

Giant language fashions (LLMs) have confirmed exceedingly highly effective for tens of purposes, some not even anticipated upon design, and researchers (and plain customers) constantly unlock new capabilities -and limitations, after all. However there’s one particular urgent query about LLMs that I’ve had ever since GPT-3 got here out for programmatic use round 2 years in the past:

Can LLMs assist with information evaluation in ways in which make customers concentrate on the science or engineering downside, abstracting from any particular maths, algorithms or programming abilities required?

Or, poised in a special, extra private and simple means:

Can I ask an LLM questions on a dataset with my very own phrases and have it interpret these questions with the maths or scripting required to reply them?

Brief reply: Sure, however…

Lengthy reply: learn on and keep tuned to my upcoming articles, the place I’ll unveil various energy, and a few limitations, of LLMs for information evaluation. Trustworthy to my fashion, with code and net apps so that you can attempt on-line.

That is one thing I’ve needed to do for a very long time since I began working with GPT-3 programmatically round a 12 months earlier than the…