[Coda]

I know in a community like this, this is going to be a bit of a controversial topic, but I've seen the vibe going around the Internet and I really don't agree with it.

You guys already know I'm a software engineer. It says "developer" right there under my name. But beyond just being a coder, I've got machine learning experience under my belt -- and an early start at it, too; my first attempts at dabbling in it were circa 1996. I say this not to try to claim any particular authority, but to be honest about my background and my opinions.

There are so many things in human society where good ideas get used by bad people, and it makes the subject poisonous to approach. For example, there's nothing inherently wrong with the concepts underlying blockchain technology or NFTs (even if there are issues with well-known implementations of them), but because of all of the NFT scams and the cryptocurrency speculation bubble, it's PR suicide to try to say that out loud.

And now it's happening again with "AI art".



Machine-learning image generators are very misunderstood. I think most people don't really get how they work. And while I'd be happy to dive into a technical explanation, it's kind of beside the point. People seem think that these tools are unethical. People think they're inherently tools of copyright infringement.

But an image generator is no more unethical and no more a tool of copyright infringement than a camera is.

Yes, you can use it in unethical ways. You can choose to use it for copyright infringement. But you can also use it for original creative work to make something you couldn't have done any other way.

Everyone is freaking out about the big companies with big models ripping off their artwork. Truth be told, those are the ones you should be fearing the LEAST. The bigger the model, the more original it can be. The examples of models trivially churning out near-duplicates of someone's work are mostly the result of an individual with a small model that's trained it specifically to try to copy one person's stuff. No amount of opting out or shaming corporations is going to stop some random guy from downloading your gallery from your website and feeding it to an AI.

All the fearmongering is going to accomplish is to create a pointless, stupid, emotional divide between people and deny a useful tool to people who would be best positioned to use it right.

And that's a big problem: Machine-learning image generators are the most powerful when they're used by artists. Just like a kid and a professional photographer will have different results even when they're using the same camera, someone with artistic skill is going to be able to make use of an image generator for much better results than the average person. Because it's not just tweaking text prompts -- the tools give you the ability to go back and forth between the AI and something like Photoshop so you can guide the generator to the perfect result.

Machine learning can't come up with new styles on its own. It can't come up with new ideas on its own. It isn't going to replace artistic skill -- it's going to augment it. People didn't stop painting just because cameras became popular. Drawing in Photoshop isn't cheating compared to drawing on paper.



None of this, of course, is to defend the people who DO choose to use the tools in an unethical way. But the anger and opposition should be directed towards the people who behave unethically, not towards the tools they're using. You don't blame graffiti on the paint!

I also acknowledge that there are other things to be worried about. I don't want to dismiss those uncaringly. And there's a lot more I could say, but this post is pretty long already. I'm more than happy to discuss anything.