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

281–290 of 1001 posts

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

#281

Post by gilbetron »

Terr_ wrote: post #267
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 like computer graphics and VR: Amazing advances over the years, very impressive, fun, cool, and by no means a temporary fad... ... But I do not believe we're on the cusp of a Lawnmower-Man future where someone's Metaverse eats all the retail-conference-halls and movie-theaters and retail-stores across the entire globe in an unbridled orgy of mind-shattering investor returns . Similarly, LLMs are neat and have so…

Ah yes, the "computer graphics, which has generated billions upon billions of revenue and change are just a fad" argument.

What next, "This Internet thing was just a fad" or "The industrial age was a fad"?

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

#282

Post by jandrewrogers »

My anecdotal experience is similar. For any important or hard technical questions relevant to anything I do, the LLM results are consistently trash. And if you are an expert in the domain you can’t not notice this.

On the other hand, for trivial technical problems with well known solutions, LLMs are great. But those are in many senses the low value problems; you can throw human bodies against that question cheaply. And honestly, before Google results became total rubbish, you could just Google it.

I try to use LLMs for various purposes. In almost all cases where I bother to use them, which are usually subject matters I care about, the results are poorer than I can quickly produce myself because I care enough to be semi-competent at it.

I can sort of understand the kinds of roles that LLMs might replace in the next few years, but there are many roles where it isn’t even close. They are useless in domains with minimal training data.

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

#283

Post by tambourine_man »

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?

I’d love a confidence percentage accompanying every answer.

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

#284

Post by yCombLinks »

kylebenzle wrote: post #222
ffsm8 wrote: post #117

Earlier quoted context omitted.

It's definitely a tech that's here to stay, unlike block chain/nfts But I mirror the confusion why people are still bullish on it. The current valuation for it is because the market thinks that it's able to write code like a senior engineer and have AGI, because that's how they're marketed by the LLM providers. I'm not even certain if they'll be ubiquitous after the venture capital investments are gone and the servic…

That's a funny comparison, I can and do use cryptocurrency to pay web hosting, VPN and a few other things as it's become the native currency of the internet. I love llms too but agree with the parent comment that says it's inevitable they'll be replaced with something better well Bitcoin seems to be sticking around for the long long term.

In my office most people use chatGPT or a similar LLM every day. I don't know a single coworker that's ever used a cryptocurrency. One guy has bought some crypto stocks.

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

#285

Post by gtirloni »

mountainriver wrote: post #263
ffsm8 wrote: post #117

Earlier quoted context omitted.

It's definitely a tech that's here to stay, unlike block chain/nfts But I mirror the confusion why people are still bullish on it. The current valuation for it is because the market thinks that it's able to write code like a senior engineer and have AGI, because that's how they're marketed by the LLM providers. I'm not even certain if they'll be ubiquitous after the venture capital investments are gone and the servic…

It makes engineers a hell of a lot more efficient and opens up software to a whole new class of people. There is plenty of data to back this up. Based on that alone it’s worth quite a lot.

It does let beginners do a lot more than they are capable of currently.

But these are people that wanted to be in programming in the first place.

This "my mom can now code and got a job because of LLMs" myth, does this creature really exist in the wild?

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

#286

Post by snitty »

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 it's going to miss things that are blindingly obvious to the attorney.

Ask a professional researcher to use it and it won't come up with good sources.

For me, I've had a lot of those really frustrating experiences where I'm having difficulty on a topic and it gives me utter incorrect junk because there just isn't a lot already published about that data.

I've fed it tricky programming tasks and gotten back code that doesn't work, and that I can't debug because I have no idea what it's trying to do, or I'm not familiar with the libraries it used.

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

#287

Post by scarface_74 »

brookst wrote: post #227
ffsm8 wrote: post #117

Earlier quoted context omitted.

It's definitely a tech that's here to stay, unlike block chain/nfts But I mirror the confusion why people are still bullish on it. The current valuation for it is because the market thinks that it's able to write code like a senior engineer and have AGI, because that's how they're marketed by the LLM providers. I'm not even certain if they'll be ubiquitous after the venture capital investments are gone and the servic…

Remember the quote that IBM thought there would be a total market for maybe 10 or 15 computer computers in the entire world? They were giant, and expensive, and very limited in application.

That quote is as apocryphal as “640K should be enough for anyone”

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

#288

Post by viraptor »

sarchertech wrote: post #240
viraptor wrote: post #166

Earlier quoted context omitted.

> dumping out yet another UI that does the same thing we've done 1,000,000 times before isn't exactly novel As a yet that's exactly what people get paid to do every day. And if it saves them time, they won't exactly get bored of that feature.

They really don’t. People say this all the time, but you give any project a little time and it evolves into a special unique snowflake every single time. That’s why every low code solution and boilerplate generator for the last 30 years failed to deliver on the promises they made.

I agree some will evolve into more, but lots of them won't. That's why shopify, WordPress and others exist - most commercial websites are just online business cards or small shops. Designers and devs are hired to work on them all the time.

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

#290

Post by inopinatus »

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?

Unfortunately they are trained first and foremost as plausibility engines. The central dogma is that plausibility will (with continuing progress & scale) converge towards correctness, or "faithfulness" as it's sometimes called in the literature.

This remains very far from proven.

The null hypothesis that would be necessary to reject, therefore, is a most unfortunate one, viz. that by training for plausibility we are creating the world's most convincing bullshit machines.

Post reply on HN