please, for all if our sakes, don't use chatgpt to learn. it's subtly wrong in ways that require subject-matter experience to pick apart and it will contradict itself in ways that sound authoritative, as if they're rooted in deeper understanding, but they're extremely not. using LLMs to learn is one of the worst ways to use it. if you want to use it to automate repetitive tasks and you already know enough to supervise it, go for it.
honestly, if I hated myself, I'd go into consulting in about 5ish years when the burden of maintaining poorly written AI code overwhelms a bunch of shitty companies whose greed overcame their senses - such consultants are the only people who will come out ahead in the current AI boom.
what's extremely funny to me is that this exact phrase was used when I was in college to explain why you shouldn't do exactly what the OpenAI team later did, in courses on AI and natural language processing. we were straight up warned not to do it, with a discussion on ethics centered on "what if it works and you don't wind up with model that spews unintelligible gibberish?" (the latter was mostly how it went back then - neural nets were extremely hard to train back then). there were a couple of kids who were like "...but it worked... " and the professor pointedly made them address the consequences.
this wasn't even some liberal arts school - it was an engineering school that lacked more than a couple of profs qualified to teach philosophy and ethics. it just used to be the normal way the subject was taught, back when it was still normal to discourage the use of neural nets for practical and ethical reasons (like consider almost succeeding and driving a fledgling, sentient being insane because you fed it a torrent of garbage).
I went back when the ML boom and sat in on a class - the same prof had cut all of that out of the curriculum. he said it was cause the students complained about it and the new department had told him to focus on teaching them how to write/use the tech and they'd add an ethics class later.
instead, we just have an entire generation who have been taught to fail the Turing test against a chatbot that can't remember what it said a paragraph ago. I feel old.
Microsoft Tay, after one day exposed to internet nazis
even less coherent. a neural net trained on a bad corpus won't even produce words. it's like mashing your face on the keyboard in a way that produces things that sound like words, inserted into and around actual, incomprehensible text. honestly, reading what gpt3 produced, I think that what was happening to a degree and they were doing postprocessing to extract usable text.
And like the LLMs themselves, they'll confidently be wrong and assume knowledge and mastery that they simply don't have, as seen in this thread.
did they get banned? I expected more angry, nonsense replies. "did chatgpt write this?" is such a fun and depressing game.
Never said they wouldn't. But you're saying the ONLY people benefitting from the ai boom are the people cleaning up the mess and that's just not true at all.
Some people will make a mess
Some people will make good code at a faster pace than before
those people don't benefit. they're paid a wage - they don't receive the gross value of their labor. the capitalists pocket that surplus value. the people who "benefit" by being able to deliver code faster would benefit more from more reasonable work schedules and receiving the whole of the value they produce.
I think it works well a a kind of replacement for google searches. This is more of a dig on google as SEO feels like it ruined search. Ads fill most pages of search and it's tiring to come up with the right sequence of words to get the result I would like.
The only thing you agreed with is the only thing they got wrong
Not really.
Third option, people who are able to use it to learn and improve their craft and are able to be more productive and work less hours because of it.
please, for all if our sakes, don't use chatgpt to learn. it's subtly wrong in ways that require subject-matter experience to pick apart and it will contradict itself in ways that sound authoritative, as if they're rooted in deeper understanding, but they're extremely not. using LLMs to learn is one of the worst ways to use it. if you want to use it to automate repetitive tasks and you already know enough to supervise it, go for it.
honestly, if I hated myself, I'd go into consulting in about 5ish years when the burden of maintaining poorly written AI code overwhelms a bunch of shitty companies whose greed overcame their senses - such consultants are the only people who will come out ahead in the current AI boom.
deleted by creator
as it turns out,
is a poor way to train both LLMs and people
deleted by creator
what's extremely funny to me is that this exact phrase was used when I was in college to explain why you shouldn't do exactly what the OpenAI team later did, in courses on AI and natural language processing. we were straight up warned not to do it, with a discussion on ethics centered on "what if it works and you don't wind up with model that spews unintelligible gibberish?" (the latter was mostly how it went back then - neural nets were extremely hard to train back then). there were a couple of kids who were like "...but it worked... " and the professor pointedly made them address the consequences.
this wasn't even some liberal arts school - it was an engineering school that lacked more than a couple of profs qualified to teach philosophy and ethics. it just used to be the normal way the subject was taught, back when it was still normal to discourage the use of neural nets for practical and ethical reasons (like consider almost succeeding and driving a fledgling, sentient being insane because you fed it a torrent of garbage).
I went back when the ML boom and sat in on a class - the same prof had cut all of that out of the curriculum. he said it was cause the students complained about it and the new department had told him to focus on teaching them how to write/use the tech and they'd add an ethics class later.
instead, we just have an entire generation who have been taught to fail the Turing test against a chatbot that can't remember what it said a paragraph ago. I feel old.
deleted by creator
even less coherent. a neural net trained on a bad corpus won't even produce words. it's like mashing your face on the keyboard in a way that produces things that sound like words, inserted into and around actual, incomprehensible text. honestly, reading what gpt3 produced, I think that what was happening to a degree and they were doing postprocessing to extract usable text.
did they get banned? I expected more angry, nonsense replies. "did chatgpt write this?" is such a fun and depressing game.
deleted by creator
I missed the local - link?
deleted by creator
It's absurd you don't think there are professionals harnessing ai to write code faster, that is reviewed and verified.
it's absurd that you think these lines won't be crossed in the name of profit
Never said they wouldn't. But you're saying the ONLY people benefitting from the ai boom are the people cleaning up the mess and that's just not true at all.
Some people will make a mess
Some people will make good code at a faster pace than before
those people don't benefit. they're paid a wage - they don't receive the gross value of their labor. the capitalists pocket that surplus value. the people who "benefit" by being able to deliver code faster would benefit more from more reasonable work schedules and receiving the whole of the value they produce.
I don't know if you know this but I can play neopets and chill once my ticket is written and passes tests. So now I work less hours a week
Also my personal project is developing faster.
I mean if you can get away with that, great. most people can't and it's just a tool to get more work out of them.
deleted by creator
I think it works well a a kind of replacement for google searches. This is more of a dig on google as SEO feels like it ruined search. Ads fill most pages of search and it's tiring to come up with the right sequence of words to get the result I would like.
deleted by creator