[GoLUG] LLMs can't reason

[kc-golug at chadwicks.me.uk]

I personally don't find LLMs all that useful. Checking the output carefully means Google is better but generating and checking something carefully like to create podcasts from pdfs works (notevooklm). LLMs are impressive but if you consider how much OpenAI have raised and the running costs (perhaps more than the whole US rail network cost) then it becomes woefully unimpressive.