The "centaur" point is spot on. The only AI use cases that actually work are when a skilled dev uses it like an autocomplete on steroids.
The 0% success rate in enterprise AI is wild though. If you can't write good internal documentation and define clear processes, throwing a chatbot on top of it just makes the chaos faster. Trying to skip the foundational engineering logic is a huge mistake.
Doctorow's absolutist writing style is so tiresome.
"perhaps a few of the younger members of those families would have to do their own homework again."
Talk to any professor. AI has had a profound and irreversible impact on higher ed, it has changed students completely, it has changed how professors operate, and it is changing how papers are written. The leading voices in mathematics are having profound discussions on the nature and uses of the technology. To dismiss these radical changes with this kind of pathetic snark is actually just embarrassing.
I'll grant you that companies trying to force "AI" into everything are going about it mostly the wrong way but speaking from the position of someone who writes code and oversees people that write code: AI (aka LLMs) is absolutely changing everything.
I can't speak to other industries but LLMs have completely changed many if not all aspects of my job. We're producing more code, better tested code, better reviewed code, at a faster speed than ever before.
This will probably be downvoted since HN seems to be full of people who still think LLMs are ChatGPT 3.5-level "intelligence" and aren't willing or able to learn/try new methods. That sounds harsh but the truth is that writing code by hand is going away.
The people trying to shove "AI" into everything without thinking critically are wrong but so are the people who still, confidently, claim that LLMs cannot write good code. That might have been somewhat true 2 years ago but it hasn't been the case since at least Dec 2025 and has only improved since then.
As for the "0%" success rate in enterprise? Ehh, I'm not sure I buy that but I also work for a smaller startup and enterprise companies never cease to amaze me with their waste so maybe it's not so surprising. But it's not because LLMs are worthless, just that those people don't know what they are talking about and/or horribly overestimate how to use them effectively.
>The current generative AI is just spicy autocomplete on steroids—doing phrase and paragraph level prediction rather than character/word level prediction. It doesn't "answer" questions, it just extrapolates the statistically most likely answer-shaped string based on its training set.
My reply:
>How do you explain the multiple unsolved math problems that AI has recently solved?
Very simple. Out of 9000 garbage assessment, you end up with one correct proof, which from what i read is not really human-readable/elegant as such.
This is to be expected out of a slop machine that has no notion of higher level concepts.
No one claimed autocomplete-on-steroids could not have any use. Some of us claim the cons (destruction of the planet, destruction of human though capabilities) infinitely outweigh the pros (you solved a useless mathematical problem or built yet another note-taking app, congratulations!)
12 comments
[ 6.4 ms ] story [ 36.1 ms ] threadThe 0% success rate in enterprise AI is wild though. If you can't write good internal documentation and define clear processes, throwing a chatbot on top of it just makes the chaos faster. Trying to skip the foundational engineering logic is a huge mistake.
And maybe even more so since they carry the potential for bigger impact.
"perhaps a few of the younger members of those families would have to do their own homework again."
Talk to any professor. AI has had a profound and irreversible impact on higher ed, it has changed students completely, it has changed how professors operate, and it is changing how papers are written. The leading voices in mathematics are having profound discussions on the nature and uses of the technology. To dismiss these radical changes with this kind of pathetic snark is actually just embarrassing.
https://hermit-tech.com/blog/ai-mania-is-eviscerating-global...
I can't speak to other industries but LLMs have completely changed many if not all aspects of my job. We're producing more code, better tested code, better reviewed code, at a faster speed than ever before.
This will probably be downvoted since HN seems to be full of people who still think LLMs are ChatGPT 3.5-level "intelligence" and aren't willing or able to learn/try new methods. That sounds harsh but the truth is that writing code by hand is going away.
The people trying to shove "AI" into everything without thinking critically are wrong but so are the people who still, confidently, claim that LLMs cannot write good code. That might have been somewhat true 2 years ago but it hasn't been the case since at least Dec 2025 and has only improved since then.
As for the "0%" success rate in enterprise? Ehh, I'm not sure I buy that but I also work for a smaller startup and enterprise companies never cease to amaze me with their waste so maybe it's not so surprising. But it's not because LLMs are worthless, just that those people don't know what they are talking about and/or horribly overestimate how to use them effectively.
>The current generative AI is just spicy autocomplete on steroids—doing phrase and paragraph level prediction rather than character/word level prediction. It doesn't "answer" questions, it just extrapolates the statistically most likely answer-shaped string based on its training set.
My reply:
>How do you explain the multiple unsolved math problems that AI has recently solved?
No answer by Stross.
This is to be expected out of a slop machine that has no notion of higher level concepts.
No one claimed autocomplete-on-steroids could not have any use. Some of us claim the cons (destruction of the planet, destruction of human though capabilities) infinitely outweigh the pros (you solved a useless mathematical problem or built yet another note-taking app, congratulations!)