I see this as proof of how bad LLMs actually are. You have an AI trained on essentially humanity’s collective programming library. Languages of machines and computers. The result should be ungodly and near perfection. If there was any semblance of understanding in AI, it should be revealed in it’s capability to produce code.
Although… I can definitely see Microsoft thinking that their code is the example of perfection and training copilot on that rather than github.
Your proof of how bad LLMs are is the fact that there are a bunch of other companies producing way better coding agents and coding models than Microsoft is? I’m not sure how that follows. Those other agents are good, that’s the point of this.
It’s wild to me that people are constantly complaining how “bad” LLMs are, because what, it can make mistakes or it’s not orders of magnitude smarter than the smartest humans???
It makes mistakes with an alarming frequency, all with the same confident tone of writing as when it says anything else. It’s wrong so often that you’d need to use another source of information just to confirm it, which defeats the point in using AI in the first place!
It’s not smarter than the smartest humans, or even AS smart as them, or even as smart as REGULAR humans. It’s not even smart. It’s a magic eight ball that sometimes has the correct answer by pure chance. No intellect. No reasoning. Just probability.
Also: It steals from just about everyone; it encourages suicides and mass shootings; it’s frequently racist in its outputs; it’s burning, boiling and polluting the planet; it’s being forced on people who don’t want it; and billionares are using it to manipulate people. Probably some other stuff I forgot to mention.
Eh, no. If you think you can offload your mental burdens onto any single source, then that’s a you problem, not an AI problem. LLMs are still getting better, but I don’t think we should hold our breaths to them getting to a point where no verification is needed. If you asked a human subject matter expert an important question, would you verify or would you just assume not only that they’re right, but also that you understood them correctly?
But your post really embodies everything that is off with AI “critique” on Lemmy. One paragraph: LLMs are just random (unlike true intellect which somehow presumably don’t emerge from probabilistic phenomena?). Next paragraph: LLMs are racist.
To be clear, the way AI is being pushed is bad in many different ways, and you didn’t even mention the worst examples which in my mind would be how AIs are currently being used to kill people, for example it is likely that it helped the US to murder 170 children in Minab. But again, that’s not a technology issue, it’s an issue with how humans interact with technology.
LLMs are random. Its weighted randomness that frequently values racist outputs like whitewashing a crowd or calling black people monkies (to name two recent examples), but it is still random. That’s why you can ask it the same question twice and get two different answers.
Ever notice how AI defenders try to pretend the technology is better than it is, and brush past the countless failings and ethical failures inherent to the technology by condemning humanity? A bad toolsmith blames the worker, I guess. At a certain point, if the technology is only doing bad things, or doing things badly, it might just be bad technology.
If they’re random, they can not be racist. That’s my point. They’re returning output based on the data they’ve been fed (assuming we’re talking about training an LLM on Twitter).
I have noticed that North American grifters pretend that AI is much better than it is, thank you very much. I also notice that China is taking a different approach, with the population being significantly more hopeful about AI going forward as a result.
I think AI “critique” on Lemmy is, for the most part, North American backlash stemming from bad practices, overpromising, environmental destruction and a general financial grift that threatens jobs. Those are all very relevant and valid, but I think it completely misses the point to blame a technology rather than a political/economic system.
I see this as proof of how bad LLMs actually are. You have an AI trained on essentially humanity’s collective programming library. Languages of machines and computers. The result should be ungodly and near perfection. If there was any semblance of understanding in AI, it should be revealed in it’s capability to produce code.
Although… I can definitely see Microsoft thinking that their code is the example of perfection and training copilot on that rather than github.
Counterpoint: garbage in, garbage out
Oh right, that.
Your proof of how bad LLMs are is the fact that there are a bunch of other companies producing way better coding agents and coding models than Microsoft is? I’m not sure how that follows. Those other agents are good, that’s the point of this.
I would (probably not literally) wager that by “this” they meant *looks around at entire world*…
this
It’s wild to me that people are constantly complaining how “bad” LLMs are, because what, it can make mistakes or it’s not orders of magnitude smarter than the smartest humans???
It makes mistakes with an alarming frequency, all with the same confident tone of writing as when it says anything else. It’s wrong so often that you’d need to use another source of information just to confirm it, which defeats the point in using AI in the first place!
It’s not smarter than the smartest humans, or even AS smart as them, or even as smart as REGULAR humans. It’s not even smart. It’s a magic eight ball that sometimes has the correct answer by pure chance. No intellect. No reasoning. Just probability.
Also: It steals from just about everyone; it encourages suicides and mass shootings; it’s frequently racist in its outputs; it’s burning, boiling and polluting the planet; it’s being forced on people who don’t want it; and billionares are using it to manipulate people. Probably some other stuff I forgot to mention.
Just in case you weren’t sure why AI is bad.
Eh, no. If you think you can offload your mental burdens onto any single source, then that’s a you problem, not an AI problem. LLMs are still getting better, but I don’t think we should hold our breaths to them getting to a point where no verification is needed. If you asked a human subject matter expert an important question, would you verify or would you just assume not only that they’re right, but also that you understood them correctly?
But your post really embodies everything that is off with AI “critique” on Lemmy. One paragraph: LLMs are just random (unlike true intellect which somehow presumably don’t emerge from probabilistic phenomena?). Next paragraph: LLMs are racist.
To be clear, the way AI is being pushed is bad in many different ways, and you didn’t even mention the worst examples which in my mind would be how AIs are currently being used to kill people, for example it is likely that it helped the US to murder 170 children in Minab. But again, that’s not a technology issue, it’s an issue with how humans interact with technology.
LLMs are random. Its weighted randomness that frequently values racist outputs like whitewashing a crowd or calling black people monkies (to name two recent examples), but it is still random. That’s why you can ask it the same question twice and get two different answers.
Ever notice how AI defenders try to pretend the technology is better than it is, and brush past the countless failings and ethical failures inherent to the technology by condemning humanity? A bad toolsmith blames the worker, I guess. At a certain point, if the technology is only doing bad things, or doing things badly, it might just be bad technology.
If they’re random, they can not be racist. That’s my point. They’re returning output based on the data they’ve been fed (assuming we’re talking about training an LLM on Twitter).
I have noticed that North American grifters pretend that AI is much better than it is, thank you very much. I also notice that China is taking a different approach, with the population being significantly more hopeful about AI going forward as a result.
I think AI “critique” on Lemmy is, for the most part, North American backlash stemming from bad practices, overpromising, environmental destruction and a general financial grift that threatens jobs. Those are all very relevant and valid, but I think it completely misses the point to blame a technology rather than a political/economic system.
Do they still do Hungarian warts?