![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
baroque_mongooseI'm feeling woozy again this morning, which is why I'm here posting rather than trundling off to church. It's annoying, but I'm sure the bright side will eventually show up.
Anyway, there's been a survey on Discord lately (which closed rather abruptly last night) to find out what people thought about adding AI features; I should think most people - certainly most people I know - responded much as I did or even more strongly, which was "no, thank you".
The thing is, I am not philosophically opposed to AI on principle, any more than I am to hand axes. You can certainly use a hand axe as a weapon, but ideally you use it to chop firewood and similar useful tasks. AI is, when you get right down to it, another hand axe; it's a tool, which can be - and has been - used for good purposes. AI excels at finding patterns, so it is particularly useful to help with questions like "are there any risk factors for disease X that we don't know about?". No; my problem with AI is not about its basic existence. It's about how it's being used.
I've actually programmed some AI, at a basic level. I think it would be a really good idea if everyone who can program at all did that, because it teaches you, in a way that nothing else can, exactly what AI is and isn't. In particular, it teaches you that it's not magic and that it's not 100% reliable, and that there are some things it is extremely bad at. To switch metaphors here, no matter how good your hammer is, not everything is a nail.
So that's my first issue with it: it's being used for things it isn't good at, and, not only that, but this is happening on such a scale that it's getting difficult to find decent stuff (if you doubt my word, go and look for sewing/knitting/crochet patterns on Etsy, but for goodness' sake don't buy them). Then there is, of course, the environmental issue. At the level I was programming it, AI is harmless in that respect, but massive AI models which use huge data centres use a lot of resources; that's not to say they should never be built, but it's very important that sustainability is taken into consideration. They should, at a bare minimum, use entirely renewable power sources and recycle the huge amounts of water they use.
And then there's the matter of exactly who's profiting from it. It is, perhaps, a little simplistic to say "if something is making Elon Musk even richer, it's a bad thing", but, more generally, AI should not be reinforcing inequality, and therefore there need to be good open-source alternatives available which are not compromised by the drive to funnel a lot of money into the hands of a few people. But then, having said that, there also need to be strong mental health safeguards in place; probably everyone by now has heard the tragic story of the teenager who was encouraged to kill himself by ChatGPT, but that is just the tip of a very worrying iceberg. AI psychosis is a real problem, and it's the main reason why I think as many people as possible should have some experience of programming it. It's a great deal harder to be thrown off your mental rails by something if you wrote the underlying code.
And, finally, there's the whole copyright issue. AI models need to be trained, and that's fair enough; but there is a great deal of stuff out there that's in the public domain, and if anything more specialised is needed it should be written in-house. I strongly believe that AI models should not be allowed to go scraping copyright works or people's private conversations to use them for training, and that goes double if the models are producing anything for them that is going to be sold for a profit.
So, as far as I'm concerned, yes, let's have AI, but let's have it in its place, with due concern for all the potential dangers; and that place, to my mind, really does not include Discord.
Anyway, there's been a survey on Discord lately (which closed rather abruptly last night) to find out what people thought about adding AI features; I should think most people - certainly most people I know - responded much as I did or even more strongly, which was "no, thank you".
The thing is, I am not philosophically opposed to AI on principle, any more than I am to hand axes. You can certainly use a hand axe as a weapon, but ideally you use it to chop firewood and similar useful tasks. AI is, when you get right down to it, another hand axe; it's a tool, which can be - and has been - used for good purposes. AI excels at finding patterns, so it is particularly useful to help with questions like "are there any risk factors for disease X that we don't know about?". No; my problem with AI is not about its basic existence. It's about how it's being used.
I've actually programmed some AI, at a basic level. I think it would be a really good idea if everyone who can program at all did that, because it teaches you, in a way that nothing else can, exactly what AI is and isn't. In particular, it teaches you that it's not magic and that it's not 100% reliable, and that there are some things it is extremely bad at. To switch metaphors here, no matter how good your hammer is, not everything is a nail.
So that's my first issue with it: it's being used for things it isn't good at, and, not only that, but this is happening on such a scale that it's getting difficult to find decent stuff (if you doubt my word, go and look for sewing/knitting/crochet patterns on Etsy, but for goodness' sake don't buy them). Then there is, of course, the environmental issue. At the level I was programming it, AI is harmless in that respect, but massive AI models which use huge data centres use a lot of resources; that's not to say they should never be built, but it's very important that sustainability is taken into consideration. They should, at a bare minimum, use entirely renewable power sources and recycle the huge amounts of water they use.
And then there's the matter of exactly who's profiting from it. It is, perhaps, a little simplistic to say "if something is making Elon Musk even richer, it's a bad thing", but, more generally, AI should not be reinforcing inequality, and therefore there need to be good open-source alternatives available which are not compromised by the drive to funnel a lot of money into the hands of a few people. But then, having said that, there also need to be strong mental health safeguards in place; probably everyone by now has heard the tragic story of the teenager who was encouraged to kill himself by ChatGPT, but that is just the tip of a very worrying iceberg. AI psychosis is a real problem, and it's the main reason why I think as many people as possible should have some experience of programming it. It's a great deal harder to be thrown off your mental rails by something if you wrote the underlying code.
And, finally, there's the whole copyright issue. AI models need to be trained, and that's fair enough; but there is a great deal of stuff out there that's in the public domain, and if anything more specialised is needed it should be written in-house. I strongly believe that AI models should not be allowed to go scraping copyright works or people's private conversations to use them for training, and that goes double if the models are producing anything for them that is going to be sold for a profit.
So, as far as I'm concerned, yes, let's have AI, but let's have it in its place, with due concern for all the potential dangers; and that place, to my mind, really does not include Discord.
