I’ve got a cool new patent pending. The idea behind it goes like this.
I think the world will have AI making movies, games, music and educational material on demand within the next 24 months or so. The trouble is that even the greatest directors produce no shortage of flops, and AI may take a long while before it’s a good “Director.” A machine that can make anything will make a great many things nobody asked for.
Hence my patent-pending twist: don’t spin out a whole movie. Produce it in real time, and modify the content — also in real time — based on how the viewer is reacting. Imagine an AI Seinfeld episode that gets funnier as you watch. Educational content produced to generate high gamma waves in children and keep their attention. An AI DJ that learns the crowd, until everyone is dancing.
Formally, the patent covers real-time content modulation based on passive, biometric user feedback. Passive is the operative word. Nobody rates anything, nobody fills in a survey. The content reads the room the way a good comedian does — except the room is you, and the reading is biometric. It’s the difference between a recording and a performance.
The AI Seinfeld is less hypothetical than it sounds. In December a stream called Nothing, Forever started running almost non-stop on Twitch — “a never-ending, AI-generated spoof of ‘Seinfeld,’ the show about nothing,” made by a two-person outfit calling itself Mismatch Media, with its script coming from an OpenAI GPT-3 model. It generates without end, and it has no idea whether anyone is laughing. That’s the gap.
And the generating side is arriving on schedule. To wit: Runway’s Gen-2 just released. The company describes it as “a multimodal AI system that can generate novel videos with text, images or video clips,” and as of this week it’s available in the browser and on iOS, though the clips are only seconds long and a bit choppy. I created a rather fantastic little video from text with it this morning.
