Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

I genuinely don't understand why some people are still bullish about LLMs

twitter.com

Post reply on HN
Print view View original

421–430 of 1001 posts

Re: I genuinely don't understand why some people are still bullish about LLMs

#421

Post by NBJack »

gilbetron wrote: post #109

I get so confused on this. I play around, test, and mess with LLMs all the time and they are miraculous. Just amazing, doing things we dreamed about for decades. I mean, I can ask for obscure things with subtle nuance where I misspell words and mess up my question and it figures it out. It talks to me like a person. It generates really cool images. It helps me write code. And just tons of other stuff that astounds me…

It's more like duct-taping a VR headset to your head, calibrating your environment to a bunch of cardboard boxes and walls, and calling it a holodeck. It actually kinda works until you push at it too hard.

It reminds me a lot of when I first started playing No Man's Sky (the video game). Billions of galaxies! Exotic, one of a kind life forms on every planet! Endless possibilities! I poured hundreds of hours into the game! But, despite all the variety and possibilities, the patterns emerge, and every 'new' planet just feels like a first-person fractal viewer. Pretty, sometimes kinda nifty, but eventually very boring and repetitive. The illusion wore off, and I couldn't really enjoy it anymore.

I have played with a LOT of models over the years. They can be neat, interesting, and kinda cool at times, but the patterns of output and mistakes shatters the illusion that I'm talking to anything but a rather expensive auto-complete.

Re: I genuinely don't understand why some people are still bullish about LLMs

#423

Post by gxs »

gilbetron wrote: post #109

I get so confused on this. I play around, test, and mess with LLMs all the time and they are miraculous. Just amazing, doing things we dreamed about for decades. I mean, I can ask for obscure things with subtle nuance where I misspell words and mess up my question and it figures it out. It talks to me like a person. It generates really cool images. It helps me write code. And just tons of other stuff that astounds me…

I wholeheartedly agree with you and it’s funny reading the replies to your comment.

Basically people just doubling down on everything you just described. I can’t quite put a finger on it but it has a tinge of insecurity or something like that, hope that’s not the case and me just misinterpreting

Re: I genuinely don't understand why some people are still bullish about LLMs

#424

Post by owenpalmer »

IncreasePosts wrote: post #340
ANighRaisin wrote: post #155

Earlier quoted context omitted.

This isn't really a problem in tool-assisted LLMs. Use google AI studio with search grounding. Provides correct links and citations every time. Other companies have similar search modes, but you have to enable those settings if you want good results.

Okay, but it's weird there is a "don't lie to me" button.

The "don't lie to me" button for a human is asking them, "where did you learn that fact?"

Grounding isn't very different from that.

Re: I genuinely don't understand why some people are still bullish about LLMs

#425

Post by imtheopposite »

gilbetron wrote: post #109

I get so confused on this. I play around, test, and mess with LLMs all the time and they are miraculous. Just amazing, doing things we dreamed about for decades. I mean, I can ask for obscure things with subtle nuance where I misspell words and mess up my question and it figures it out. It talks to me like a person. It generates really cool images. It helps me write code. And just tons of other stuff that astounds me…

What I find interesting is that my experience has been 100% the opposite. I’ve been using ChatGPT, Claude, and Gemini for almost a year (well only the ChatGPT for a year since the rest are more recent.) I’ve been using them to help build circuits and write code. They are almost always wrong with circuit design, and create code that doesn’t work north of 80% of the time. My patience has dropped off to the point where I only experiment with LLM a few times a week because they are so bad. Yes it is miraculous that we can have a conversation, but it means nothing if the output is always wrong.

But I will admit the dora muckbang feet shit is fucking insane. And that just flat out scares the pants off me.

Re: I genuinely don't understand why some people are still bullish about LLMs

#426

Post by chii »

nonethewiser wrote: post #355
cogman10 wrote: post #161

Earlier quoted context omitted.

> Wah, it can't write code like a Senior engineer with 20 years of experience! No, that's not my problem with it. My problem with it is that inbuilt into the models of all LLMs is that they'll fabricate a lot. What's worse, people are treating them as authoritative. Sure, sometimes it produces useful code. And often, it'll simply call the "doTheHardPart()" method. I've even caught it literally writing the wrong algor…

> What's worse, people are treating them as authoritative. … I've both seen online and heard people quote LLM output as if it were authoritative. Thats not an LLM problem. But indeed quite bothersome. Dont tell me what Chatgpt told you. Tell me what you know. Maybe you got it from ChatGPT and verified it. Great. But my jaw kind of drops when people cite an LLM and just assume it’s correct.

> when people cite an LLM and just assume it’s correct.

people used to say the exact same thing with wikipedia back when it first started.

Re: I genuinely don't understand why some people are still bullish about LLMs

#427

Post by kristopolous »

cogman10 wrote: post #161
gilbetron wrote: post #109

I get so confused on this. I play around, test, and mess with LLMs all the time and they are miraculous. Just amazing, doing things we dreamed about for decades. I mean, I can ask for obscure things with subtle nuance where I misspell words and mess up my question and it figures it out. It talks to me like a person. It generates really cool images. It helps me write code. And just tons of other stuff that astounds me…

> Wah, it can't write code like a Senior engineer with 20 years of experience! No, that's not my problem with it. My problem with it is that inbuilt into the models of all LLMs is that they'll fabricate a lot. What's worse, people are treating them as authoritative. Sure, sometimes it produces useful code. And often, it'll simply call the "doTheHardPart()" method. I've even caught it literally writing the wrong algor…

This is the problem of the internet writ large.

The solution is to be selective and careful like always

Re: I genuinely don't understand why some people are still bullish about LLMs

#428

Post by chii »

trial3 wrote: post #365
nonethewiser wrote: post #355

Earlier quoted context omitted.

> What's worse, people are treating them as authoritative. … I've both seen online and heard people quote LLM output as if it were authoritative. Thats not an LLM problem. But indeed quite bothersome. Dont tell me what Chatgpt told you. Tell me what you know. Maybe you got it from ChatGPT and verified it. Great. But my jaw kind of drops when people cite an LLM and just assume it’s correct.

It might not be an LLM problem, but it’s an AI-as-product problem. I feel like every major player’s gamble is that they can cement distinct branding and model capabilities (as perceived by the public) faster than the gradual calcification of public AI perception catches up with model improvements - every time a consumer gets burned by AI output in even small ways, the “AI version of Siri/Alexa only being used for mus…

Why is that a problem tho?

Branding for current products have this property today - for example, apple products are seen as being used by creatives and such.

Re: I genuinely don't understand why some people are still bullish about LLMs

#429

Post by skydhash »

Ferret7446 wrote: post #45
Panzer04 wrote: post #25

Earlier quoted context omitted.

Is such code hard to write in the first place? Automating the easy 80% sounds useful, but in practice I'm not convinced that's all that helpful. Reading and putting together code you didn't write is hard enough to begin with.

It's not hard, but it's time consuming.

The things I'm wary of are pitfalls that are often only in the command/function docs. Kinda like rsync with how it handles terminating slashes at the end of the path. Which is why I always took a moment to read them.

Re: I genuinely don't understand why some people are still bullish about LLMs

#430

Post by silisili »

Animats wrote: post #178
cogman10 wrote: post #161

Earlier quoted context omitted.

> Wah, it can't write code like a Senior engineer with 20 years of experience! No, that's not my problem with it. My problem with it is that inbuilt into the models of all LLMs is that they'll fabricate a lot. What's worse, people are treating them as authoritative. Sure, sometimes it produces useful code. And often, it'll simply call the "doTheHardPart()" method. I've even caught it literally writing the wrong algor…

> inbuilt into the models of all LLMs is that they'll fabricate a lot. Still the elephant in the room. We need an AI technology that can output "don't know" when appropriate. How's that coming along?

That's what I really want.

One thing I see a lot on X is people asking Grok what movie or show a scene is from.

LLMs must be really, really bad at this because not only is it never right, it actually just makes something up that doesn't exist. Every, single, time.

I really wish it would just say "I'm not good at this, so I do not know."

Post reply on HN