Great thread detailing the extent to which LLMs are bullshit machines that make things up.

Never trust LLMs for any research or facts! They are built to give *any answer* so long as it sounds plausible.
Reposted from Patrick Galey
If you know me, you know what I think of gen AI - it can automate some tasks and write decent code but can only ultimately rephrase the Internet.

Yesterday I was (lazy/time-poor) so I tried Cerebras' Llama-3.1 interface.

I think this is quite illustrative of gen AI chain-of-thought 🧵

Comments