Hacker News Reader

Comment by krupan | Hacker News Reader

This is the best explanation of why I also don't use LLMs and I'm grateful it has been put into words here. I love every reason and explanation given, except for me personally I am still writing code. It's just that at this point in my career I either don't need help writing the code, or I have run into something really challenging and/or obscure and LLMs are no help at all!

View on HN

It is always been my understanding that the less competent one is, the more they find LLMs useful.

I don’t use them to generate code but I’ve found ChatGPT becoming increasingly helpful giving me ideas or working through things I’m stuck on.

I’ve started regularly sharing my ideas with it and either getting back a “yeah that’s correct, here’s why” or a “there’s an alternative which might be better …”

I’m actually shocked how good it’s getting. Used it on the weekend to help set up a terraria server. Yeah I could have worked out iptables was dropping the connection and changed the config myself. I’ve done it before. But having ChatGPT walk me through it was way faster.

You dont like using and building tools imbued with intelligence that you can control. Why.

There is no "intelligence" in LLMs; they're text predictors. As far as I can tell the whole LLM technology has limited applications in entertainment and that's about it. "Hallucinations" (even that term is problematic, as it suggests there's an actual consciousness/person, or the seed of one, there) as well as other "failures" - in fact features inherent to how the tech works - make it irrelevant for basically all other use cases.

Tech as an industry already had an atrocious reputation but the moment the insanely stupid "AI" bubble pops I suspect it'll get much worse. Ultimately it's pretty deserved, though. At this point the bullshit is so strong one almost wishes for a new AI winter.

The lack of control I feel when using them is one of the big reasons I don't like them

Reply on news.ycombinator.com