Ian Eck

we were sloppily vibe-coded the night before a deadline

Everyone working in AI is frantically preparing for this massive wave, and everyone outside of AI thinks all those people ARE the massive wave and should all stop. Both are correct

What it feels like building your custom little agent harness while the singularity swells between China and California

It’s always a balancing act between the sponsor money and the creative community. Made doubly hard in such a volatile, dynamic space. We try to create a symbiotic relationship and hope to improve this balance every year. As one of the producers of the conference, what I want to bring in next year is MUCH more experimentation—between media-code, physical-digital, ai-human. I see us evolving past “ai film” into something more expansive. I’d urge you to engage with our simulation (app.aionthelot.com/assistant) in which you can query our conference ai however you like. “Find me a mentor, find me an investor, find me someone to move into a hacker house with for a week, etc” The simulation data is liquid and can be spooned out by any prompt you write. Then imagine this is baked into the physical space. Baked into the badges. Baked into the volumetric stage, and gamified. We didn’t quite land the concept this year, but it’s the type of thing I want to encourage more of. Breaching the canvas of flat-screen linear narrative. Creating more cross-functional experiences. Cutting through the typical social hierarchies and smooth-talkers. Elevating the true prompt-creatives. Thanks for coming out!

New prompt technique -Create two opposing bots to debate a topic -Give them both acid

7.) The future affordance of all this is the ability to make this content more dynamic, reactive, and alive. Gen-AI will enable nonlinear storytelling, more like games than movies.

6.) This is a new flavor of creativity that feels more like discovery. The skills it rewards are curation, taste, writing, editing, patience, and an open mind. We’re about to see an explosion of “AI dreamers” that will have more in common with bedroom DJs than traditional film-makers.

5.) In this way, playing with Sora feels like making a stew with a slot machine. You have this sort of idiot-savant as your sous chef, not doing exactly what you want, but surprising you with its ingenuity. You can test ideas rapidly and go down rabbit holes tangential to your own imagination.

4.) The “blend” feature is another highlight. AI comes up with transitions that would be impossible for a camera to pull off, and too resource-intensive for any VFX artist to render on a whim. But more than that, Sora devises transitions between shots that no human would even think of.

3.) Your canvas is a minimalist timeline (5-20 seconds long) where you can write in your keyframes, and drag/drop any image/video in whatever order you like. Then the model interpolates all the frames that might come in between. -80% of these generations are nightmarish/useless. -10% are passable attempts at your request -And 10% fall into the “happy accidents” category where some unexpected magic happens.

2.) The video model itself is good but not way-ahead-of-the-pack good. It still struggles with simple physics and character consistency. And we still haven’t crossed the uncanny valley when it comes to ~humans talking~ Animation always seemed like a natural fit for gen-AI for these reasons. You’re freed from the shackles of realism. Unfortunately, you can tell Sora hasn’t been trained much on animation, illustration, or fine art, so it doesn’t have the aesthetic breadth of a Midjourney.

1.) First, Sora feels like a faster and more buttery instrument than other AI video interfaces. Everything just works.

Here’s a short film I made with OpenAI's new Sora video model. I was lucky enough to be a part of the Sora Alpha Creator program—testing the interface, diving into the tool’s capabilities, affordances, and shortcomings. Here’s some takeaways (a thread):

AI is the latest on-ramp for creativity. New paths are always a little janky and upsetting for people on the well-trodden paths—but that's more of a reason to lean into this experiment, so we can design it thoughtfully. @MikeGioia1 @_B_Contreras_ latimes.com/business/story/2…

Ever wanted to copy/paste a writing style? Now you sorta can

Is prompt engineering a thing? Is it a science? A fad? An art form? Guess we'll find out