I've been seeing various conversations about AI-generated art (henceforth 'AI art'), and I've found a lot of the perspectives generally unsatisfying. This is me attempting to work out my thoughts and feelings about it -- the usual caveats apply: this is a thought sketch rather than something more together, etc. etc.
The two most strident viewpoints I've heard are 'AI art is not different from a human artist using other works as reference/inspiration points' and 'AI art is inherently bad and represents theft'. I lean more towards the latter than the former in practice, but as I say, I don't find either suitably satisfying in and of itself.
I think, as with a lot of things these days, particularly when it comes to discussing emerging technologies (an AI in particular), people are conflating lots of disparate facets, and often end up talking (or arguing) at cross purposes, with their focus on slightly different things.
I think it's nonsense to argue that AI tools for generating art are either the same as a human artist or 'just another tool available to people' (and, what's more, this argument often seems to be trying to have it both ways). Demonstrating why I think it's nonsense is slightly harder. This is not some position against recognising non-human agents, but I don't think people are really trying to argue that DALL-E has personhood, for instance.
For me, the biggest difference between a human artist using other works as reference or inspiration and an AI art tool statistically modelling them en masse is that last part -- en masse. As with other technologies, the issue is perhaps not truly inherent, but comes down to a step change in scale and speed. A human simply cannot look at that many different works and assimilate them into something (other than perhaps, their own style). AI art tools are able to do that and fluidly move between them. This is not a situation we've had to face before when considering art, ethics, or copyright, and I think it's spurious, therefore, to adopt that as an analogous argument.
There's also the collapse between talking about people using these tools for their own interest and entertainment and people or organisations using these tools as part of commercial art pipelines. The former seems to carry more emotive weight -- 'why would you deny people access to something they've not been able to perform or afford before?' -- but also is the less worrisome bit. It doesn't have no issues, but it doesn't have the same problems as a tool that can scrape and in effect profit from the works of many, many artists without their opting in and essentially launder their contribution to enrich the tool's creator and/or cut those selfsame human artists out of work. These two elements -- personal and commercial -- obviously have no barrier between them in terms of what the tools can accomplish. It's very convenient to argue the emotive, personal case as a way of pushing back against criticism of the commercial one.
(There's also probably a distinction to be made around 'making art' and 'producing visual assets' but boy howdy do I not have time to get into that just now.)
As with other emergent technologies, often it comes down to a means -- intentional or otherwise -- of cutting out traditional forms of labour, or skirting regulations, organised workers, etc. Yes, it might be the case that the availability of AI gives people access to things they didn't before or makes things better in various contexts, but that doesn't mean it's automatically a net good, either.
Think about taxis -- in short, Uber and its cohort were a way of circumventing regulations and worker protections in the taxi industry in various locales. And yes, in some cases (like San Francisco), it sounds like the existing system was broken. But this became a mass movement that spent its way into undermining previously viable ways of doing things (e.g. by ploughing billions of dollars of investment into subsidising rides to undercut existing services, until they collapsed and Uber et al. were all that was left).
Is that model better for the consumer? Well, sure, initially, they get a ubiquitous and cheaper service! But as with other things, the only way this can continue to 'be viable' (if it ever was) is at the cost of heavily exploiting everyone in that loop which it is possible to exploit.
It's hard to argue that this has been a bad thing, because most people are exposed only to the end-user experience, which has been -- generally, most of the time -- extremely convenient and, historically, startlingly cheap. But of course there's another shoe. There's always another bloody shoe.
But, as with ridesharing, it feels like the AI is passing a threshold where people's desire for shiny things and convenience will continue to gather momentum faster than the -- very real, very valid -- arguments against it. Or, how about, maybe not even arguments against it. Maybe just arguments that we should be having meaningful conversations about how we fit this into our society rather than just moving ahead at Mach-3.
But it's not surprisingly, really, either, since that's how things seem to go with humanity. Someone needs to buy the truth some slip-ons.