Sister blog of Physicists of the Caribbean in which I babble about non-astronomy stuff, because everyone needs a hobby

Sunday, 6 April 2025

Upgrades

Today, a few more thoughts on AI, per some recent upgrades to ChatGPT's image generator. I generally stand by my previous assertions, but...

There is, of course, a Jurassic Park quote suitable for any occasion, and this time I think we need Ian Malcom's famous monologue :

Genetic force is the most awesome power the planet’s ever seen, but you wield it like a kid that found his dad’s gun. I’ll tell you the problem with the scientific power that you’re using here. It didn’t acquire any discipline to attain it. You read what others have done and you took the next step. You didn’t earn the knowledge for yourself so therefore you don’t take any responsibility for it. You stood on the shoulders of geniuses to accomplish something as fast as you could and before you even knew it you had it. You patented it and packaged it and slapped it on a plastic lunch box, and now your selling it! You wanna sell it! Well, your scientists were so preoccupied with whether or not they could they didn’t stop if they should!

It's not a perfect speech. "You read what others have done and you took the next step" is hardly much of a put-down : if anything is a pretty good description of how specialist research works. It has to work that way because we can hardly re-learn an entire discipline from first principles : if we all had to do that, research would have stalled probably sometime in the 18th century. Fortunately for the movie, Jeff Goldblum is very good indeed at delivering a genuine sense of imperfection, as though his character had really just thought all that up off the top of his head – so of course it has errors.

But what be errors so far as genetics is concerned might be pretty much straightforward truth when it comes to AI art. Not in the process of machine learning and programming and all that, which seems indeed to be something you have to be a right clever-clogs to understand. No, I mean in using AI to come up with new images for general consumption. Here you don't need any discipline, no technical skill in painting or drawing whatever – you just need a good idea. Or indeed, any idea.

My sympathies tend in the following direction :

"AI slop", though I despise the word "slop", is a perfectly legitimate term. An awful lot of AI "art" has a very distinctive and "samey" appearance that's usually good for no more than a cheap joke. You generally have to put some work in to get it to give anything interesting, and though that's perfectly possible, the basic content does tend to be slop. But then again, I strongly think that there's far, far too much modern art created by humans which is no better, and in fact can be a good deal worse.

So disparaging AI content on the sole grounds of its origin is patently ludicrous. Circumstances are important, but content should stand on its own. "Oh, they did it on a computer" is the same dumb attitude that morons have had for ages about CGI, not realising the skill that's gone in to creating iconic movies such as, well, that famous one about the dinosaurs. No, it doesn't matter that they used a computer, they still had to work bloody hard, and it's still art.

But the latest upgrades appear to be changing that. Whereas previously some considerable effort was required in writing and refining a prompt, sometimes ad nauseum to get a decent result, this appears to be less and less the case. We've getting ever closer to the point of having a good idea => producing content people want to consume, in one single rapid step.

In one sense this is harmless enough :


As I've said before, most AI-created content wouldn't have existed at all without it. I myself used an image of cybernetic Jane Austen in a recent talk; next week I'll use a cute cartoon of a radio telescope with boxing gloves. No money has been denied artists here, let alone "stolen" from them in any fashion : were AI not available, I simply wouldn't have made the images at all, and would either have used some other free image or thought of a different joke. Never in a million years would I have paid anyone anything, because we don't have a budget for that and never will.

But if I can't and won't pay anyone for joke images for my presentations, the same isn't true elsewhere. Indeed the quality of the above images show quite well how things are progressing towards a state of genuinely consumable, financially-viable content. Even more so is the case of the AI-generated children's "book", The Very Hungry Alien, of which I show here only my personal favourite image :


As far as I'm concerned this is adorable. Anyone still claiming that AI isn't useful is simply deluding themselves.

Now this particular style won't be to everyone's taste, but there's enough people who like this stuff that it has unavoidable consequences – well beyond making "shitty memes" for a cheap laugh. Last Christmas I bought a friend an actual, human-produced book Jonesey : Nine Lives on the Nostromo. Had it had content like the above, I would not have known or cared that it was AI produced, because it would still have met the quality threshold of "neat gift, here, take my money". So we are indeed reaching the stage where commercial opportunities for real, human artists with families to feed may be squandered. 

It might not be a problem if all art and illustration required exceptional levels of imagination, because then we could still pay professional artists to think of good ideas. The problem, of course, is that it doesn't. Sure, some of it does, but not all. Plenty of people have good ideas but lack the technical skills to produce them : the democratisation of creativity is a mixed blessing, to put it mildly.

I don't know what to do about this. Not a clue. I like this content and I'm glad it exists. To me, giving people the power to bring their ideas to life is a fundamentally good thing. But I can't deny I want people to earn a decent wage for doing what they love, and after all, it's the work of human artists that has made this possible in the first place.

Putting the genie back in the bottle is a strategy beloved by social media activists but almost never works. Yes, there will be diehards who refuse to even look at AI images on moral principle, but hardly anyone thinks that way. The vast majority, given the possibility to illustrate their ideas, and approving of the quality of the results, will happily both consume and generate this content. And to be honest, so will I. Trying not to is like saying, "don't use AI to help write your Python code, no no, spend the next ten hours Google searching or the next two days waiting for a condescending reply on Stack Overflow, or here, wade through this 300-page manual to find the two-line snippet you actually need". The idea of not using a freely-available, rapid, accurate solution when one exists is just not going to happen.

I have no conclusion. The only thing I can think of is that instead of xenomorphs, they should have had Terminator-style robots causing the apocalypse by taking away all the work of graphic artists. But maybe this would have caused the Universe to collapse in a singularity of sheer meta-irony. Anyone who wants to run with that idea, well, be by guest, I suppose.

No comments:

Post a Comment

Due to a small but consistent influx of spam, comments will now be checked before publishing. Only egregious spam/illegal/racist crap will be disapproved, everything else will be published.

Upgrades

Today, a few more thoughts on AI, per some recent upgrades to ChatGPT's image generator. I generally stand by my previous assertions , b...