Breaking news, the staff from Sound & Vision has been given severance packages, and all duties will be handled by ChatGPT-4 going forward; here's a link to the full article:
AI Takes Over Sound & Vision Magazine: Staff ReplacedMy sympathy goes out to David Vaughn, who will hopefully continue his duties reviewing movies for this site going forward.
OK, check the date, and then remember the tradition that has been going on for years from that publication, going back to the days of Lirpa Labs (whatever happened to them?).
This does lead to an opportunity to talk about the future of AI, though. No, not what’s been going on with Boston Dynamics or some Skynet scenario, but just in regard to the entertainment industry, and more specifically movies. The AI advancements made in the past couple of years have been remarkable, and the enhancements are coming along quicker than anyone would expect. There have been lots of articles lately about ChatGPT, but there are also advances made in other areas that are just as amazing.
One example is what’s happening in the art world. Programs such as Stable Diffusion can take ordinary language prompts and create images that have never been seen before. This is an absolutely fascinating program to me. As a student, I was one of the best in my class, and had aspirations of being a comic book artist at one time. But I soon realized that being great in a group of 30 kids isn’t good enough to cut it professionally, so I gave that up when reaching adulthood.
Using Stable Diffusion, a user can fashion their own visions into reality just by using descriptive language to prompt results. It does take some learning to get used to, but the results can be striking. To illustrate, here are two pictures. One I created using Stable Diffusion. The other is from an industry award-winning published role-playing game rule book:
OK, I picked some low-hanging fruit for this one. It might be hard to believe, but the image on the left was from an actual published work. The one on the right is a wholly original illustration generated by me using a series of refined prompts based on the styles four fantasy artists I admire. The difference isn’t subtle.
This is an example of fantasy work, but there are users out there creating photo-realistic pieces that would be hard to distinguish from the real thing. It’s gotten remarkably good at producing faces, while it still struggles with things like limbs and fingers.
There are similar advances being made in the written world. While the Sound & Vision article was a joke, it was actually produced using chatGPT-4. While that sounds wild, various publications have been using AI to create content, including CNET and BuzzFeed. Some of these have been filled with errors, which, like the case of appendages in art, are still a problem, but it’s likely that you’ve run across some text that was completely computer generated and not even known it.
The Writers Guild of America is actively discussing and debating how AI might be used to create screenplays going forward, and what rules might need to be created to regulate its use.
Then there’s voice generation. Computer speech has been around for a long time, but that is also evolving. The recent Obi-Wan series on Disney+ featured an AI version of Darth Vader’s, with James Earl Jones signing off on the change. They did a similar thing for Luke Skywalker’s appearance in The Mandalorian. The improvement over the old stilted and monotone voicing is remarkable.
You may very well have seen this in practice too, when combined with AI-generated text. There are YouTube channels that combine AI articles with an AI voiceover, combine it with generic footage, producing videos with click-bait titles. There are often tell-tale signs of this that signal it as not right, but they’re produced in such a way that they still get revenue-producing clicks.
The next big advancement will come with motion pictures. While the writing, voicing, and conceptual art might already be affected by AI, producing moving images will take much more time and refinement. There are already some short films made this way, but that’s not at the level of a Hollywood-grade photorealistic movie.
So where does that leave us for practical application today? It’s easy to imagine a scenario where a creator sits down with the bones of a story idea, and uses AI to hammer out a script, letting it lay down the basic plot, with many variations, and then goes in to polish it, fixing any logical inconsistencies or clunky dialog.
It’s also easy to imagine that same creator working with an AI imaging tool to create their own conceptual artwork based on words taken directly from their script. It could even be used to storyboard the whole thing, although it might be faster to work directly with an actual artist at that point.
Once that’s done, voicing AI could be used to cast various actors to read sections of the script without having them come in for an audition. The total result is a pitch-package ready for studio review. I actually wouldn’t be shocked if much of that was already going on today.
There are all kinds of legal and ethical questions and concerns that will need careful review. Copyright is the biggest one, both for the final product and for any pre-existing works that may have been used in its creation. Early rulings were that AI-created art couldn’t be copyrighted, since it wasn’t made by a person, but that may change in the future based on how much human input there was during the process. There are also lawsuits from artists arguing that AI art is stealing from their prior creations. But that raises the question of fair use and substantial transformation.
All of that will still be argued as time goes on, but the existing technology will only continue to improve, with the results becoming harder and harder to distinguish from those that were created from scratch by real humans. And the only thing that’s certain is that the creative arts will never be the same.
Scott
2 Replies
326 Views