We shouldn't really be surprised that tech companies inventing an AI powerful enough to generate shockingly realistic digital art just by feeding it a text prompt is stirring up some controversy in the creative community.
I wrote about my experience using one of these AI art tools, DALL·E, a few weeks ago, and I'm still having a great time poking around in it. I used it as recently as today to see if it could render what I would look like if I was a character in the live-action movie remake of Cats. The results were disappointing.
But in the last few weeks, reporting about what externalities this sort of software could have on the creative community has picked up. Business Insider has run a few really great pieces on the topic recently, one about an art contest in Colorado that was recently won by an artist who unbeknownst to the judges created his piece using the AI art tool Midjourney, and another about how these AI tools are indexing the styles of a bunch of well-known artists and allowing people to create new AI-generated art in the styles of these artists without their permission.
Once again, not surprising that these follow-on effects exist, but still interesting to read about exactly what they are and start a conversation around how the developers of this technology should respond to these concerns.
There are a few backend infrastructure changes that I think definitely need to take place if this technology is going to see expanded use in the future, which I absolutely assume it will. First, all of these companies should give artists the ability to remove their artwork from the software's training data, preventing the software from directly copying the style of a specific artist without their permission. One of these AI companies, Stable Diffusion, is already working on such a feature, and in general I think building this option is necessary to make sure artists still have an exclusivity on their artistic voice rather than have it be unjustly yanked onto the open market by unicorn startups.
The second change has to do with how the creation of a digital asset is recorded. We're living in the roaring 20s of digital content right now as all of us go about our daily lives documenting each and every event of note with our iPhones cameras. But this boom in digital content isn't just contained to cute pet pictures; AI-generated art and deepfake videos are making it harder than ever to separate what's real from what's not on the internet, and it's only going to get worse as these AI technologies get better. We desperately need a sort of "digital fingerprint"system for all digital content that logs exactly when and how that piece of content was generated, and we need that information stored on a secure ledger that cannot be edited or accessed by bad actors. We'll call it, I don't know, "the blockchain" or something silly. Regardless of what technology enables this to happen, though, we need this digital fingerprint system to be figured out sooner than later before AI-art and deepfake videos catch up to iPhone cameras in the eye test, at which point we have a lot more issues than can be covered in a single blog post.
But as far as how these AI art tools will impact the creative community specifically, I'm not actually too worried about it. First off, we always seem to overestimate the impact of exciting new tech in the short term. I'm sure some people were harping on about how negatively impacted the artistic community would be when Microsoft first demoed MS Paint. But more importantly, I think AI art tools have proven they can make realistic or pretty art, but they've yet to show they can make thoughtful art. Anyone who appreciates a piece of art as something more than just the sum of its prettiest colors knows that there is much more to being a great artist than just making something pleasing. You need to imbue it with emotion. You need to capture the subtlety that gives the piece character. You need to make it tell a story, not just show a scene. It's not clear to me that any AI art platform has been able to accomplish these yet, and it's not clear to me any of them have a path to getting there any time soon. Maybe you've seen AI try to tell jokes before; the results aren't glowing. The only people that might get killed by AI art software are people who were never creating anything thoughtful or original in the first place.
In other words, AI art tools won't kill creatives. They will kill uncreatives.
Even the winner of the art content I mentioned above admitted it was far more intensive to create his winning piece than just typing in a cool phrase into the generator. He said he created hundreds of images in the software and spent many weeks "fine tuning and curating" the results before solidifying three finalists to submit.
As I've also written about before, my favorite metaphor for how technology should be used is as a "bicycle for the mind," a Steve Jobs quote that captures how technology should amplify what humans are capable of, not just replace it. Without AI being advanced enough to make a genuinely thoughtful piece of art, I don't see it being able to replace the need for human artists right now. However, I think it has tremendous potential as a tool for artists to generate unique assets to include in a piece, mock up different ideas before dedicating time to pursuing them, or just generally create something outside-the-box for inspiration. I'm a horrible freehand artist, but with some DALL·E credits and a few hours on Photoshop, I think I could make a really unique and compelling custom piece. That's the epitome of what being a "bicycle for the mind" is. The more we can give people who don't self-sort as "artists" the tools they need to start making more art, the better. It's technology at its finest.
And let's be honest. Look at that contest-winning piece in the article I linked above. I don't care whether a human painted it by hand or extensively groomed it with AI. It's freaking beautiful. And it's hard to say the world won't be a better place with more art like that.
Comentários