Is it really a thought terminating cliche? It’s shorthand for the fact that LLMs and other gen “AI” produce hallucination ridden unoriginal output by their design. It’s a dead-end technology that cannot be relied upon for anything serious, everything it does must be verified in such a way to render the automation aspect useless. And the “art” aspect is insulting, there can be no message or human experience conveyed by a statistical average of real art.
It’s a dead-end technology that cannot be relied upon for anything serious
I’m sorry you are completely wrong. Do LLMs produce slop? Yes. Are LLMs useful? Also yes. The software industry, at least at the high end, is transforming rapidly because of this technology. The question of whether or not LLMs and other generative AI techniques along those lines are generally applicable is still an open one. But at least in some industries such as software it is certainly not a dead-end technology that cannot be relied upon for anything serious.
I work in tech and I develop software. LLMs generate code a lot faster than a human, but it is simply worse code; it is buggier, it is less secure, and it is less technically correct to the specifications. It takes more time to get that code up to snuff after the fact, especially given the verbose bloated code bases created by LLMs, than to write it by hand such that it was correct from the start. Also considering the real costs of running these things in such a way that you get any sort of “usable” result ends up costing more than having an engineer do it right, I stand by my point.
This may have been an accurate assessment 18 months ago, but it is not anymore. If you’re not getting good results out of the latest models it’s more of a you problem than a model problem at this point. And not to be completely rude here but at the level of people getting paid 7 figures USD to develop software LLMs are being widely used for basically all code. We need to move past this cope about their capabilities.
Sure, and thats why all of these companies have to force us to use this shit with quotas, mandatory tooling, and chiding about being 100x devs and being left behind. I’ve yet to meet an LLM booster in a senior technical position that actually handles code in any meaningful way. You can find plenty of execs making claims about it though. You’re just saying shit
I’ve been through it with this shit at my last two workplaces. People are willing to claim the output is solid because they can just keep running agents non stop to fix their issues and run a quick skim over the output, a bit of extra agentic use to run some tests and call it a day. They don’t care about code quality anymore, if it runs, ship it.
I’ve seen people ship a ton of features at once. 3 PRs in a day, 250 file changes. Just quick skim it, or even better, get a senior dev to run the diff through another agent and have that agent review the code because the senior is too overworked on reviewing slop he has to outsource his job too. Approve it.
And when it breaks we just cross our fingers because nobody knows how any of it works.
I am not just saying shit. I am telling you, as a person working for such a company, that this is real and this is happening. This is not executives talking shit. Although yes executives do talk a lot of shit and have unrealistic expectations. Coding was never really the slowest part of software development anyway.
everything it does must be verified in such a way to render the automation aspect useless.
While there’s no formal mathematical proof that P ≠ NP, all signs point it being so. Even if literally everything must be verified as you say, it still very often takes less time to verify a result than to generate it.
If my codebase needs to pass a manual code review which finds an endless slew of LLM hallucinations that need correcting, it adds an immense amount of time to the process to rectify them. And these tools create insanely long winded and bloated code bases so the review itself is much longer from the outset. We’re not talking about rigorous mathematical proofs, it’s much more akin to a smoke test, and even that is a huge burden to bear when using those tools such that I view them as useless from this caveat alone. And it’s basic best practice in the industry to perform these reviews, they are vital to quality results.
Even hobbyist software dev work would feel this impact. If you aren’t testing the functions an LLM writes you will be putting buggy work with major obvious edge cases and vulnerabilities out there. The tests they write for themselves are often incomplete, happy path, or plainly avoiding actual testing by just asserting true.
From seeing these shortcomings in a field where technical proof and automated testing is a possibility, I can only imagine the level of human review necessary on less quantitative work to ensure any sort of reliability or accuracy. They cannot be trusted as accurate or authoritative at any point in the chain. ML can be useful as statistics for fuzzy and approximate data, but that’s really where the usefulness ends
My point is that it doesn’t take less time to verify a result than to generate it. You can generate stuff all day but to verify it takes a lot more effort and tjme
And yet, I’ve gotten help from LLMs with code in two different ways: 1) Where I check everything at the door and manually write out or rewrite in my own words, only using the AI as an assistant for the method at most. 2) Where I don’t know enough in the language, so I go back and forth, trying a solution it gives me and then going back to it when corrections are needed.
2 is definitely more painful and I would probably be extra wary of using it in production business code and insist on review from someone who is literate in the language. BUT, there have been occasions where if I didn’t have 2 as an option, I may have never written the solution at all, or it would have taken me weeks or months of learning to get there. If I could be assured the “personal learning then completion” path would happen, that would lessen the value of 2, but being real, I cannot remotely guarantee it. Whether it’s a matter of time, motivation, etc., even someone who is already literate in some programming like myself, will not necessarily find it easy to learn every new language, library, etc., or always find it to be a motivating pursuit worth doing.
In other words, sometimes we are not talking about space-faring code and we’re talking about whether something with largely trivial consequences for mistakes, and a low friction path to fixing them, will ever get made at all. This is the kind of area where LLMs shine, since they don’t need to be without error, they just need to be “good enough” that it empowered somebody to make something that wouldn’t otherwise have gotten made.
there can be no message or human experience conveyed by a statistical average of real art.
Does a kid’s messy stick figure fridge drawing have a message and human experience conveyed? Or does it only mean something because the parents are proud of their kid for being able to do it at all?
Another way of looking at it is: Are Explosive Media’s Lego videos not conveying a message? AI not having the conscious will and worldview to actively choose to generate a specific message does not prevent humans from using AI generation as a means of helping them construct a message. The idea that AI is bad for crafting a specific message is a point that has become harder to defend over time, the better AI gen has gotten at listening to highly specific prompting.
Is it really a thought terminating cliche? It’s shorthand for the fact that LLMs and other gen “AI” produce hallucination ridden unoriginal output by their design. It’s a dead-end technology that cannot be relied upon for anything serious, everything it does must be verified in such a way to render the automation aspect useless. And the “art” aspect is insulting, there can be no message or human experience conveyed by a statistical average of real art.
I’m sorry you are completely wrong. Do LLMs produce slop? Yes. Are LLMs useful? Also yes. The software industry, at least at the high end, is transforming rapidly because of this technology. The question of whether or not LLMs and other generative AI techniques along those lines are generally applicable is still an open one. But at least in some industries such as software it is certainly not a dead-end technology that cannot be relied upon for anything serious.
I work in tech and I develop software. LLMs generate code a lot faster than a human, but it is simply worse code; it is buggier, it is less secure, and it is less technically correct to the specifications. It takes more time to get that code up to snuff after the fact, especially given the verbose bloated code bases created by LLMs, than to write it by hand such that it was correct from the start. Also considering the real costs of running these things in such a way that you get any sort of “usable” result ends up costing more than having an engineer do it right, I stand by my point.
This may have been an accurate assessment 18 months ago, but it is not anymore. If you’re not getting good results out of the latest models it’s more of a you problem than a model problem at this point. And not to be completely rude here but at the level of people getting paid 7 figures USD to develop software LLMs are being widely used for basically all code. We need to move past this cope about their capabilities.
Sure, and thats why all of these companies have to force us to use this shit with quotas, mandatory tooling, and chiding about being 100x devs and being left behind. I’ve yet to meet an LLM booster in a senior technical position that actually handles code in any meaningful way. You can find plenty of execs making claims about it though. You’re just saying shit
I’ve been through it with this shit at my last two workplaces. People are willing to claim the output is solid because they can just keep running agents non stop to fix their issues and run a quick skim over the output, a bit of extra agentic use to run some tests and call it a day. They don’t care about code quality anymore, if it runs, ship it.
I’ve seen people ship a ton of features at once. 3 PRs in a day, 250 file changes. Just quick skim it, or even better, get a senior dev to run the diff through another agent and have that agent review the code because the senior is too overworked on reviewing slop he has to outsource his job too. Approve it.
And when it breaks we just cross our fingers because nobody knows how any of it works.
I am not just saying shit. I am telling you, as a person working for such a company, that this is real and this is happening. This is not executives talking shit. Although yes executives do talk a lot of shit and have unrealistic expectations. Coding was never really the slowest part of software development anyway.
While there’s no formal mathematical proof that P ≠ NP, all signs point it being so. Even if literally everything must be verified as you say, it still very often takes less time to verify a result than to generate it.
If my codebase needs to pass a manual code review which finds an endless slew of LLM hallucinations that need correcting, it adds an immense amount of time to the process to rectify them. And these tools create insanely long winded and bloated code bases so the review itself is much longer from the outset. We’re not talking about rigorous mathematical proofs, it’s much more akin to a smoke test, and even that is a huge burden to bear when using those tools such that I view them as useless from this caveat alone. And it’s basic best practice in the industry to perform these reviews, they are vital to quality results.
Even hobbyist software dev work would feel this impact. If you aren’t testing the functions an LLM writes you will be putting buggy work with major obvious edge cases and vulnerabilities out there. The tests they write for themselves are often incomplete, happy path, or plainly avoiding actual testing by just asserting true.
From seeing these shortcomings in a field where technical proof and automated testing is a possibility, I can only imagine the level of human review necessary on less quantitative work to ensure any sort of reliability or accuracy. They cannot be trusted as accurate or authoritative at any point in the chain. ML can be useful as statistics for fuzzy and approximate data, but that’s really where the usefulness ends
It doesn’t matter whether that specific use case is fruitful or not. I wasn’t making a case for it.
My point is that it doesn’t take less time to verify a result than to generate it. You can generate stuff all day but to verify it takes a lot more effort and tjme
This is true in some use cases and false in others. In the use case you presented, it is true.
And yet, I’ve gotten help from LLMs with code in two different ways: 1) Where I check everything at the door and manually write out or rewrite in my own words, only using the AI as an assistant for the method at most. 2) Where I don’t know enough in the language, so I go back and forth, trying a solution it gives me and then going back to it when corrections are needed.
2 is definitely more painful and I would probably be extra wary of using it in production business code and insist on review from someone who is literate in the language. BUT, there have been occasions where if I didn’t have 2 as an option, I may have never written the solution at all, or it would have taken me weeks or months of learning to get there. If I could be assured the “personal learning then completion” path would happen, that would lessen the value of 2, but being real, I cannot remotely guarantee it. Whether it’s a matter of time, motivation, etc., even someone who is already literate in some programming like myself, will not necessarily find it easy to learn every new language, library, etc., or always find it to be a motivating pursuit worth doing.
In other words, sometimes we are not talking about space-faring code and we’re talking about whether something with largely trivial consequences for mistakes, and a low friction path to fixing them, will ever get made at all. This is the kind of area where LLMs shine, since they don’t need to be without error, they just need to be “good enough” that it empowered somebody to make something that wouldn’t otherwise have gotten made.
Does a kid’s messy stick figure fridge drawing have a message and human experience conveyed? Or does it only mean something because the parents are proud of their kid for being able to do it at all?
Another way of looking at it is: Are Explosive Media’s Lego videos not conveying a message? AI not having the conscious will and worldview to actively choose to generate a specific message does not prevent humans from using AI generation as a means of helping them construct a message. The idea that AI is bad for crafting a specific message is a point that has become harder to defend over time, the better AI gen has gotten at listening to highly specific prompting.