Genie 3 and what it really is
It's not what you think it is...yet
While investors seem to think Google’s Genie 3 is ready to flip the gaming world on it’s head, people who know how the new AI tool works are less worked up about it. I talked to a number of devs about what Genie 3 is, what it isn’t, and what it may yet be down the road. While a load of what I write here today is my opinion, it ain’t based on nothin’.
This is what Google says Genie 3 is:
Given a text prompt, Genie 3 can generate dynamic worlds that you can navigate in real time at 24 frames per second, retaining consistency for a few minutes at a resolution of 720p.
Yep, it does that. And I think it looks like complete shit. (Sorry, I suppose slop is what everyone calls it.) My point is, it’s a very neat trick, but it’s not a game. If anything, it’s a next generation copyright violation generator. I swear you can almost hear Nintendo lawyers chewing their lips.
If you believe link-baiting X posters, then you already believe the video game industry and development environment is cooked. You’d be wrong due to a myriad of reasons, of which I will share just a few:
AI does not create these “games” or settings, or anything. AI can only assemble something from what it was trained on. It can only remix elements. Which is why you are seeing mostly results that look like GTA clones and overt Nintendo IP duplication.
The tool does not make a “game” as we would define it, with all the interaction, story, pacing, and complexities designed for human consumption. What it does offer is a way to prompt janky slightly explore-able spaces you can sort-of walk around in for three minutes at 24fps, 720p.
What it DOES do it kinda neat. Like a lot of gen-AI, there’s a strong first impression, until the truth of that you are seeing, much less playing, starts to show through the cracks.
I keep seeing people write things like, “this is the worst it will ever be!”. Sure, the tools and tech will improve, no doubt about that. Will anyone want to play these “games”? I don’t care HOW good it looks, or how quickly you generated them with prompts, these are not interesting outside of the initial contrivance.
I think its more likely, like a lot of gen-AI, the “game” will be in just screwing around making crap of it all. Like the Sora app and the like. Make stuff, get bored, still play Arc Raiders with your friends on Friday night.
Some might say, “well, with the help of the tools in the hands of game devs, writers and designers, the sky’s the limit!”
Yeah, what you are describing is game development. They’ve always had tools, and those tools are only as good as the devs, writers and designers using them to create this art particular collaborative and technical art form.
People think they would love a robot to just make a TV show or write a book based on a couple simple prompts that appeal to you. But we’ve learned pretty quickly that the results are never what you actually REALLY want. What you want is rare, and special, and human, and hard to duplicate simply by “duplicating” them.
They may ultimately assist, speed up, or otherwise affect how things are made, but this human thinks it may never reach what we are actually trying for. And if it does, that might not be something we never actually wanted at all in the end.
Reminds me of Roblox games
This is well said. I seem to recall a scene from the matrix where the oracle said “cookies need love, like everything does”. AI lacks the ability to put forth love to the code as developers have learned to do. I’ve said it before and I’ll say it again. Video games are an art form. Slap it in a gallery and watch the intricate pieces mesh together in perfect harmony. It doesn’t even require tons of money as is evident by so many indie developers. It just takes understanding how to use the most of your resources and composing them into a beautiful symphony. This gen-ai slop is just a bad attempt at mimicking what real developers try and succeed at.