A.I. in 2023, Continued

A couple of days ago, I wrote down a bunch of thoughts on A.I. I didn’t really get into how it works. I talked about what it produces, and how it is best to use it to assist artists rather than replace them. By not talking about how these large language models work, I skipped right over the biggest issue with this software, so that’s why I’m revisiting this topic so soon.

Current A.I.’s are trained on tremendous datasets. ChatGPT would not be able to create a story in the style of Mark Twain if it wasn’t fed a bunch of Mark Twain stories. Art A.I.’s are trained on art in much the same way. We feed a huge amount of data into the machine, and then based on what we’ve given it, we ask for extrapolations from its datasets in order to generate a “new” product.

It’s the training data where things get spicy. Were the writers and artists that produced the work that is fed into the machine asked if their work could be used in such a way? If not, how is that not theft?

As I said just after my birthday, you cannot copyright style. But let’s not be cowards and hide behind the law. Is it moral and right for Grammarly to feed its A.I. using customer data? Is it okay for Google to scan emails and Google Drive documents in order to feed its A.I.?

The cost to acquire this technological marvel is the dataset that is fed into the machine. These large companies are taking a socialist approach to handling that cost, but then preparing to turn around and sell the product in the most capitalist way. That’s one way to look at it, a way to unemotionally determine that what is happening with A.I. is wrong.

But fuck unemotional. No one wants their voice stolen and then reused without their consent. A writer or artist brings their voice to the material of their stories and their art. When we talk about an artist’s voice, are we not talking about style? Isn’t that what we’re training these machines to duplicate?

In my previous post, I truthfully state that I think the technology has the potential to be amazing. It does. Now I’m saying that the technology comes with a price tag, which is that your art, stories, and voice will get used, poured into a digital stew and served up to a machine god that will always be hungry for more.

Would it make any difference if the A.I. created from all of our combined voices could not be used to create profit for someone else?

I think that would help, but it wouldn’t solve the whole problem. I put this blog up for free, and I’m always hoping that my words will enter other people’s eyeballs or ears and provide pleasure or sustenance to someone else’s brain. If someone (foolishly) trained an A.I. using all of this publicly accessible version of my voice, even if they didn’t make a dime off of it (which they wouldn’t), they still stole my voice. Maybe I’d be okay with it if they used my voice to create some kind of public good. I think I would still want to be credited, or have some say in how my voice is used.

In summary, once again, I’m a fan of the technology in principle, but I am not a fan of how the technology is built. I think there are still questions of morality around building A.I. using the work of other people. And I still believe that artists, writers, and programmers are going to be hurt by this technology as companies learn to really misuse it.