"AI" is a failed technology, in the sense that it's neither fit for purpose, nor a likely path to its stated end goal of "AGI", but a lot of people's profits currently depend on pretending it's not.
-
ai does not have to be perfect to be useful. repeatability comes from tests, reviews, and engineering discipline. you are describing a tool that needs to be managed and pretending that proves it is useless.
You made me curious I admit. If you found out how to make agentic engineering not a waste of time... Then you probably...
1) Know how constrained agents and codebases have to be
2) Have a sense that 95% of what's critical in the job is still human
3) Invested a lot of time to find that out because it's not advertised muchYet your advice is "try and you'll see, AI can code*"
*If managed well.
🥸🥸🥸
That's a tricky statement. AI industry lives on AGI lies which produces despicable CLI tools and agentic IDE. Knowledge about how to use those things well is fragmented and contradictory. Default mode of any agent is shit.
I get why tech people are angry. Just using the tools isn't enough.
Or maybe you found some secret to make things smoother?
-
@knowprose "not all men"-ass take
@mayadev no idea what you're talking about, or why you did.
-
@joeinwynnewood @Gargron @skavish
what are the odds that, given the speed with which bad code can be created and that management definitely has no idea what good code looks like, they won't push their people to produce huge piles of crap
I would say those odds are very high. But management also pushes people to produce piles of crap without AI. I just spent most of last year fighting management on a poorly designed project where they just wanted to add more features before considering allowing a redesign, despite my proposed redesign making those features a lot easier to add. I've held multiple presentations on code quality that mostly fell on deaf ears. And this was before they switched to pushing AI.I'm not sure AI will make their code worse, because it is already really bad. The code quality of many corporate projects is atrocious. Despite their many automated code quality tools.
Been there, done that. But at least people produce code at rates that generally allow the few capable and competent people in the shop to keep the edifice from burning down.
When LLMs spit out orders of magnitude more code, absent guidance by those handful of capable and competent people, good luck, because nothing but luck will work.
-
it is genuinely mind-boggling. people who never used the tools, or know people who tried them and got shit results, confidently explain to people who use them successfully every day that they do not work. at some point, this stops being skepticism and starts looking like an ideology. the irony is fucking spectacular.
@skavish Yes, I am ideologically motivated in my rejection of LLMs. I'm not ashamed of that.
Vibecoders think a good prompt that explains the problem well, and usually results in working code, is basically the same as literate programming. I'm ideologically committed to the idea that programmers should understand the theory of their program. So I think the vibecoders are fundamentally unserious about their craft.
Yes, that's ideology -- and if you don't have ideology behind your own programming, that means you're just in it for the money.
-
@catsalad @zalasur @Gargron @mellowfish
It's already happening. The biggest recurring topic on developer mailing lists at work is squealing because the company has imposed token limits or withheld access to certain models and devs feel they can no longer do their jobs.
If you can't write code without using an #LLM, you can't write code. It's as simple as that.
-
@hllizi @Gargron @Luminous_J People arguing over things they won't even bother to unambiguously name, just another day on the internet.
-
"AI" is a failed technology, in the sense that it's neither fit for purpose, nor a likely path to its stated end goal of "AGI", but a lot of people's profits currently depend on pretending it's not. I am curious to see how long the hype will last once the Nvidia ouroboros ends and money stops being pumped into marketing.
@Gargron "neither fit for purpose"
you think there are no valid and effective uses for AI today? it works rather well today, not perfect but perfectly workable and could sustain multi-billion usd worth industry even if nvidia/openai/meta disappeared in a Thanos-like snap tomorrow.
I'm quite optimistic on AI rollout particularly because open source and open weight is due a strong catch up and likely even leadership in some domains. We already see open-claw lead and this is just the beginning.
-
@skavish @Gargron
I program for a living. I've reviewed AI generated code. It sucked. I've had AI review my code. It always got obvious shit wrong.It will continue to suck for the foreseeable future, because it's an imitation engine whose training data prominently features badly-written code and StackOverflow examples labeled ""Here's what I wrote; why doesn't it work?", and it has no actual intelligence.
The more reliable the coding AI gets, the more useless it will actually become, because folks will rely on it for bigger things, and finding the bugs that will inevitably turn up in major projects will be beyond them.
Software Engineer with 28 years of professional experience here (40 years of programming experience in total). Nah, that's not at all a good description of LLM software output and it also seems to be based on a misunderstanding on how LLMs learn from training material.
I have a bunch of equally senior software dev friends. We all use LLMs nowadays.
-
"AI" is a failed technology, in the sense that it's neither fit for purpose, nor a likely path to its stated end goal of "AGI", but a lot of people's profits currently depend on pretending it's not. I am curious to see how long the hype will last once the Nvidia ouroboros ends and money stops being pumped into marketing.
@Gargron
I totally agree. What I also dislike is renaming things with "AI" just to make it sound advanced.I heard an advert today for British Telecom saying their "AI business tool" could help your small business. That "tool" was booking tables (that's a CMS/database, not AI) and answering calls (that's an answering machine (at a push, IVR), not AI) etc.
-
"AI" is a failed technology, in the sense that it's neither fit for purpose, nor a likely path to its stated end goal of "AGI", but a lot of people's profits currently depend on pretending it's not. I am curious to see how long the hype will last once the Nvidia ouroboros ends and money stops being pumped into marketing.
@Gargron EV is a failed technology, in the sense that its purpose is to get rid of gas stations bjt that have not happened even after more than a decade.
The Covid vaccine is a failed technology, in the sense that some people will get vaccinated but still will get ill with covid.
Sure, AI is over-subsidized and over-hyped, but in its base is a great tool for some tasks. I doubt we'll stop using it.
-
"AI" is a failed technology, in the sense that it's neither fit for purpose, nor a likely path to its stated end goal of "AGI", but a lot of people's profits currently depend on pretending it's not. I am curious to see how long the hype will last once the Nvidia ouroboros ends and money stops being pumped into marketing.
no this is just false
It is true there is a lot of hype, but if you talk to any PhD or MD doing biomedical research, you are likely to find someone very excited about the potential of AI
-
@keisatsu @Gargron @imhotep @skavish So I need to be a dev to understand that using the argument being a maintainer of a FOSS project is enough expertise for him, but not for Linus?
In one answer he said: „I’m right, because I’m a FOSS maintainer. Period.“. Here he says, that being a FOSS maintainer isn’t a argument for being right.
I don’t need to be a dev to understand language and understand, that this both posts are contrary.
-
@keisatsu @Gargron @imhotep @skavish So I need to be a dev to understand that using the argument being a maintainer of a FOSS project is enough expertise for him, but not for Linus?
In one answer he said: „I’m right, because I’m a FOSS maintainer. Period.“. Here he says, that being a FOSS maintainer isn’t a argument for being right.
I don’t need to be a dev to understand language and understand, that this both posts are contrary.
-
if you’ve never used it, then yes, you’re not in a position to judge what it looks like when used properly. being an experienced engineer and maintainer gives you plenty of expertise in software engineering, but it doesn’t magically give you first-hand experience with a tool you explicitly said you’ve never used.
you can absolutely argue that you don’t want to use it. that’s a different claim.
-
"AI" is a failed technology, in the sense that it's neither fit for purpose, nor a likely path to its stated end goal of "AGI", but a lot of people's profits currently depend on pretending it's not. I am curious to see how long the hype will last once the Nvidia ouroboros ends and money stops being pumped into marketing.
@Gargron I have to respectfully disagree even though I know I'll be pilloried for it. Economically yes there's a huge bubble around the tech but it's far from failed. Right now there are thousands of small businesses and individuals doing so many creative things with both local and inexpensive hosted models. We're just scratching the surface of what will come for better or worse, but failed? No. Democratized coding yes. Where that leads is up to us.
-
@Gargron I have to respectfully disagree even though I know I'll be pilloried for it. Economically yes there's a huge bubble around the tech but it's far from failed. Right now there are thousands of small businesses and individuals doing so many creative things with both local and inexpensive hosted models. We're just scratching the surface of what will come for better or worse, but failed? No. Democratized coding yes. Where that leads is up to us.
I'd say it's "failed" in that there is no way it can live up to the level of hype that the Tech Bro's have made around it, but yes, there is a limited place for it.
CGI existed before this, it will continue to exist albeit augmented.
Machine Generated Art & Text existed before this, it will continue to exist albeit augmented.
Machine Learning existed before this and will continue too.
The annoying things is that all of this has been subsumed into the morass that is "AI": a technology which is not delivering what was promised.
I think people in the know need to revert to the old terms specifically because it displays an actual understanding of the tech behind the promise.
As it is, anyone that says "AI" to me when what they really mean is "ML" of "CG" etc. (particularly the ones that should know better) doesn't get my full attention or the benefit of the doubt from me.
-
if you’ve never used it, then yes, you’re not in a position to judge what it looks like when used properly. being an experienced engineer and maintainer gives you plenty of expertise in software engineering, but it doesn’t magically give you first-hand experience with a tool you explicitly said you’ve never used.
you can absolutely argue that you don’t want to use it. that’s a different claim.
@skavish "you have to step on rakes or else you can't say that stepping on rakes is a bad idea" ass response here
-
@Gargron That's a quite strong stance, that missed proper argumentation. I agree it's not perfect and it has risks, but 'failed"?What is your definition for 'failed'?
Is it the fact that radiologists now can better diagnose breast cancer using AI as a tool to diagnose than without?
Or Is it the fact that it immensely progressed mRNA technology?
Or perhaps is it your fear for a 'new technology' that's alien and could perhaps take over your job and eventually your life?
@Luminous_J @Gargron « This large-scale study examined the heterogeneous effects of AI assistance on 140 radiologists across 15 chest X-ray diagnostic tasks and identified predictors of these effects. Surprisingly, conventional experience-based factors, such as years of experience, subspecialty and familiarity with AI tools, fail to reliably predict the impact of AI assistance. Additionally, lower-performing radiologists do not consistently benefit more from AI assistance, challenging prevailing assumptions. Instead, we found that the occurrence of AI errors strongly influences treatment outcomes, with inaccurate AI predictions adversely affecting radiologist performance on the aggregate of all pathologies and on half of the individual pathologies investigated. »
https://www.nature.com/articles/s41591-024-02850-w
And that's ignoring the deskilling issues found with other medical professionals.
-
frankly, i do not give a shit about big corps with shitty management. AI is a massive multiplier for me and my team. if some incompetent corporate fucks manage to use it to create a mountain of shit, that is their problem. let them burn.
I'm tossing this in as another intrinsic aspect of LLM code.
"We demonstrate that hallucinations stem from the fundamental mathematical and logical structure of LLMs. Therefore, eliminating them through architectural improvements, dataset enhancements, or fact-checking mechanisms is impossible."
When people make mistakes, they're almost always mistakes we've seen before. LLM hallucinations? Unlikely.
LLMs Will Always Hallucinate, and We Need to Live with This
https://link.springer.com/chapter/10.1007/978-3-031-99965-9_39 -
@waltbaldwin @Gargron @mellowfish Great piece, loved it. But much of that was written by AI! I feel duped.
