The Machine Learning space has kind of exploded this year, with a lot of news around Generative models. This has caused a bit of an uproar in the copyright and creative worlds, and rightfully so. While it is an impressive leap going from being able to identify an apple in a picture to generating a picture of an apple. I don’t think it’s as revolutionary as people have been making it out to be.
I’ve taken to calling the things generated by ML models soulless, because I could never really think of a better way to put it. Regardless of medium, an artist pours their perspective, emotions, and ideas into a piece of art. When we experience it, those parts of the artist mix with parts of us, and it moves us. That’s the power of a heart-filled, soulful piece of art.
It’s something that I’m not sure can be captured with tensors and hyperplanes. Maybe someday, there will be something that can. But it will take a significantly larger jump than the ones we’ve seen recently.
These things are still tools; even with their limitations, they are extremely useful. If you’re an artist, writer, or even a coder, these tools can let you focus on the pieces of a project that can benefit from extra attention.
There’s also a danger here, these models need data. There’s talk about companies already looking at downsizing their junior software engineer pools. Because ChatGPT and Github CoPilot can both generate code. But if you slowly eliminate the junior positions, you’ll run out of seniors, and eventually may end up removing the human perspective from the role altogether.