A voice of reason in the AI LLM frenzy

https://www.msn.com/en-us/news/technology/large-language-models-will-never-be-intelligent-expert-says/ar-AA1RkXMM? large-language-models-will-never-be-intelligent-expert-says

My rant: A timely reminder that AI LLM are excellent at collating vast amounts of data and feeding it back out in readable formats which is extremely powerful in areas where there is a limited, verifiable, fact-based corpus of information to train on. AI has yet to exhibit any judgement on the quality of the data or create truly new understandings. Please don't be gullible - use it as a useful tool and understand how to use it properly.

Find more posts tagged with

Features

Desktop App

Comments

Dave Hooton

Keep initial queries in Logos relatively simple. A followup query is effective in restricting scope.
Use a collection of specific books rather than querying authors and getting opinions from others!

DMB

I think he's stuck on LLM's which are basically pattern matching. To get outside 'existing', we had to use genetic algorithms to climb outside of what exists (and far exceed human intelligence). I've no doubt they'll use similar techniques.

Ironically my current brain book suggests similar; competing/optimizing.

MJ. Smith

https://community.logos.com/discussion/comment/1387184#Comment_1387184

Yes, he is speaking specifically about LLM's as that is where false claims and public misunderstanding is focused currently.