I genuinely don't understand why some people are still bullish about LLMs • Hacker News Board

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

#621

Post by sussmannbaka » Fri, Mar 28, 2025, 10:31 AM UTC

qwertox wrote: post #616

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…

An LLM is like a mouse. You no longer have the console as the primary interface, but a GUI, which 99.9+% of computer users control via a mouse. You no longer have the screen as the primary interface, but an AUI, which 99.9+% of computer users control via a headset, earbuds, or a microphone and speaker pair. You mostly speak and listen to other humans, and if you're not reading something they've written, you could hav…

I don’t even argue that they might get useful at some point, but when I point a mouse at a button and press the button it usually results in a reliable action. When I use the LLM (I have so far tried: Claude, ChatGPT, DeepSeek, Mistral) it does something but that something usually isn’t what I want (~the linked tweet). Prompting, studying and understanding the result and then cleaning up the mess for the low price of an expensive monthly sub leaves me with worse results than if I did the thing myself, usually takes longer and often leaves me with subtle bugs I’m genuinely afraid of growing into exploitable vulnerabilities. Using it strictly as a rubber duck is neat but also largely pointless. Since other people are getting something out of the tech, I’ll just assume that the hammer doesn’t fit my nails.

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

#622

Post by namaria » Fri, Mar 28, 2025, 10:35 AM UTC

ryandrake wrote: post #274

satisfice wrote: post #87

Earlier quoted context omitted.

I want to sit next to you and stop you every time you use your LLM and say, “Let me just carefully check this output.” I bet you wouldn’t like that. But when I want to do high quality work, I MUST take that time and carefully review and test. What I am seeing is fanboys who offer me examples of things working well that fail any close scrutiny— with the occasional example that comes out actually working well. I agree…

What's also scary is that we know LLMs do fail, but nobody (even the people who wrote the LLM) can tell you how often it will fail at any particular task. Not even an order of magnitude. Will it fail 0.2%, 2%, or 20% of the time? Nobody knows! A computer that will randomly produce an incorrect result to my calculation is useless to me because now I have to separately validate the correctness of every result. If I nee…

Better yet than the whole 'unknown, fluctuating and non deterministic rates of failure' is the whole 'agentic' shtick. People proposing to chain together these fluctuating plausibility engines should study probability theory a bit deeper to understand just what they are in for with these rube goldberg machines of text continuation.

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

#623

Post by f6v » Fri, Mar 28, 2025, 10:36 AM UTC

somenameforme wrote: post #554

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 mostly because of how they were initially marketed. In an effort to drive hype 'we' were promised the world. Remember the "leaks" from Google about an engineer trying to get the word out that they had created a sentient intelligence? In reality Bard, let alone whatever early version he was using, is about as sentient as my left asscheek. OpenAI did similar things by focusing to the point of absurdity on 'safety'…

> Remember the "leaks" from Google about an engineer trying to get the word out that they had created a sentient intelligence?

No, first time I hear about it. I guess the secret to happiness is not following leaks. I had very low expectations before trying LLMs and I’m extremely impressed now.

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

#624

Post by r00sty » Fri, Mar 28, 2025, 10:36 AM UTC

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 mean. How would you feel if you coded a menu in Python with certain choices but when you used it the choices were never the same or in the same order, sometimes there were fake choices, sometimes they are improperly labelled and sometimes the menu just completely fails to open. And you as a coder and you as a user have absolutely no control over any of those issues. Then, when you go online to complain people say useful stuff like "Isn't it amazing that it does anything at all!? Give us a break, we're working on it bro."

That's how I see LLMs and the hype surrounding them.

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

#625

Post by qwertox » Fri, Mar 28, 2025, 10:38 AM UTC

sussmannbaka wrote: post #621

qwertox wrote: post #616

Earlier quoted context omitted.

An LLM is like a mouse. You no longer have the console as the primary interface, but a GUI, which 99.9+% of computer users control via a mouse. You no longer have the screen as the primary interface, but an AUI, which 99.9+% of computer users control via a headset, earbuds, or a microphone and speaker pair. You mostly speak and listen to other humans, and if you're not reading something they've written, you could hav…

I don’t even argue that they might get useful at some point, but when I point a mouse at a button and press the button it usually results in a reliable action. When I use the LLM (I have so far tried: Claude, ChatGPT, DeepSeek, Mistral) it does something but that something usually isn’t what I want (~the linked tweet). Prompting, studying and understanding the result and then cleaning up the mess for the low price of…

These are the beginnings and it will only improve. The premise is "I genuinely don't understand why some people are still bullish about LLMs", which I just can't share.

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

#626

Post by jibal » Fri, Mar 28, 2025, 10:38 AM UTC

somenameforme wrote: post #554

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 mostly because of how they were initially marketed. In an effort to drive hype 'we' were promised the world. Remember the "leaks" from Google about an engineer trying to get the word out that they had created a sentient intelligence? In reality Bard, let alone whatever early version he was using, is about as sentient as my left asscheek. OpenAI did similar things by focusing to the point of absurdity on 'safety'…

> Remember the "leaks" from Google about an engineer trying to get the word out that they had created a sentient intelligence?

That's not what happened. Google stomped hard on Lemoine, saying clearly that he was wrong about LaMDA being sentient ... and then they fired him for leaking the transcripts.

Your whole argument here is based on false information and faulty logic.

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

#627

Post by rcxdude » Fri, Mar 28, 2025, 10:45 AM UTC

torginus wrote: post #523

Should we listen to Sabine in this case? Isn't this another manifestation of a generally intelligent person, who happens to be an expert in her field weighing in on something she's not an expert on, thinking her expertise transfers? This is the most common 'smart person' fallacy out there. As for my 2 cents, LLMs can do sequence modeling and prediction tasks, so as long as a problem can be reduced to sequence modelin…

Well, she's an expect in the field she's asking the LLM about, and she judges the responses to be nonsense. Would a non-expert be able to tell? This is mostly mirroring my experience with it, which is a straightforward report of "it doesn't work for me, even though I keep trying because everyone else is hyped about it and apparently it's getting better all the time".

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

#628

Post by hmans » Fri, Mar 28, 2025, 10:48 AM UTC

[dead]

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

#629

Post by topaz0 » Fri, Mar 28, 2025, 10:48 AM UTC

snitty wrote: post #286

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…

Folks really over index when an LLM is very good for their use case. And most of the folks here are coders, at which they're already good and getting better. For some tasks they're still next to useless, and people who do those tasks understandably don't get the hype. Tell a lab biologist or chemist to use an LLM to help them with their work and they'll get very little useful out of it. Ask an attorney to use it and…

Honestly it's worse than this. A good lab biologist/chemist will try to use it, understand that it's useless, and stop using it. A bad lab biologist/chemist will try to use it, think that it's useful, and then it will make them useless by giving them wrong information. So it's not just that people over-index when it is useful, they also over-index when it's actively harmful but they think it's useful.

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

#630

Post by jibal » Fri, Mar 28, 2025, 10:50 AM UTC

gloosx wrote: post #569

lnenad wrote: post #565

Earlier quoted context omitted.

> They push the narrative that they’ve created something akin to human cognition This is your interpretation of what these companies are saying. I'd love to see if some company specifically anything like that? Out of the last 100 years how many inventions have been made that could make any human awe like llms do right now? How many things from today when brought back into 2010 would make the person using it make it f…

No major company directly states "We have created human-like intelligence," they intentionally use suggestive language that leads people to think AI is approaching human cognition. This helps with hype, investment, and PR. >I'd love to see if some company specifically anything like that? 1. DeepMind researchers: Sparks of Artificial General Intelligence: Early experiments with GPT-4 - https://arxiv.org/abs/2303.12712…

ELIZA was not a natural language processor, and the fact that some people were easily fooled by a program that produced canned responses based on keywords in the text but was presented as a psychotherapist is not relevant to the issue here--it's a fallacy of affirmation of the consequent.

Also,

"4. Meta's AI chief scientist, Yann LeCun, has repeatedly said they are working on giving AI "common sense" and "world models" similar to how humans think."

completely misses the mark. That LLMs don't do this is a criticism from old-school AI researchers like Gary Marcus; LeCun is saying that they are addressing the criticism by developing the sorts of technology that Marcus says are necessary.