May 31, 2025

Veo, Tarkovsky, and the Sweet 'n' Sour Taste of Instant Cinema

“Modern mass culture, aimed at the ‘consumer,’ … is crippling people’s souls.” Tarkovsky wrote that. It hits different when a shot now renders faster than popcorn pops.

I boot Veo-2. Prompt: early-morning freight yard, 1979 Kodachrome, slow dolly. Thirty seconds—wet rails, lens flare, perfect fog. Gorgeous. Great for pre-vis, dubs, quick VFX. But it's just tokens. My gut flips.

Veo-3 lands: new engine, plus generated dialogue. The implication is a means to fully automate human performance in video. No crew call, no 4 a.m. coffee, no carrying a human history in a human body. I'll be interested to see but I sense this is where we'll see many creative-minded folks draw a line. Fine for sensible ads, or interactive mediums like games and world simulations where realtime performance generation is actually doing something new. Cinema is literature at 24 fps; yank out the human and you’re left with HD noise

My ninth grade English teacher (shoutout to Mrs. S!) drilled it in: literature is a reflection of the human condition. Remove the humanity, the mirror shatters. So my north star: 1. Amplify, don’t automate. 2. The AI should prompt the human to be creative, not the other way around. And 3. every output needs a “break-glass” toggle exposing exactly how the illusion fails.

Disinformation? Still a beast. Watermarks and detectors help, but coverage is patchy. Until C2PA-style provenance lives in every phone and camera, we’re heading for a balkanized web: gated zones that verify content, free-for-all zones that don’t. (Side bet: 35 mm film prices climb—proof you can hold up to the light.)

Yet none of this kills the fun. Rapid previs saves budgets, AI cleanup rescues shaky takes, world-building goes turbo. I just keep the sweat in the loop so Tarkovsky’s warning stays theory, not prophecy.

Wonder and queasiness, shoulder to shoulder. Honest—and motivating.

- Noah (views are my own)

Apr. 2, 2024

A Preface on Gen AI Design

When I first started tinkering with AI, I had no idea it would become an obsession. But as I've dug deeper into the field, working alongside engineers and researchers, I've realized that this technology is going to completely upend how we create, think, and share ideas. For journalists, creators, academics, and anyone who traffics in knowledge, AI won't just change how we work; it'll reshape the very nature of what we consider work.

But here's the thing: as much as I'm excited about the possibilities, I'm also keenly aware of the pitfalls—not least of which is the post-truth, synth-flooded internet we're quickly arriving at today. We're playing with some seriously powerful tools here, and if we're not careful, things could go sideways fast. I'm talking about issues of trust, transparency, privacy, liability, bias, intellectual property, geopolitical stability, and a burst of new questions around human consciousness—the kinds of thorny issues that keep philosophers up at night.

For me, charging headlong into the AI revolution, I want to make sure we're building safeguards and thinking hard about the human impact. It's not enough to just build cool stuff; we have a responsibility to do it thoughtfully and with integrity. Call me old-fashioned, but I think the endgame here cannot be about machines replacing humans. It cannot be about LLMs writing screenplays, nor some sagacious GPT-whatever that knows everything about everything—it's got to be about machines and humans working together in a chaotic, beautiful symbiosis of creativity and intelligence. It's got to be about trying to achieve all the great things that might now be possible sooner thanks to AI—advances in scientific research and sustainable materials, breakthroughs in robotics for prosthetics and accessibility, and unprecedented cultural exchange through universal language—just to name a few.

The road ahead is going to be a wild ride, but I'm stoked to be part of the crew trying to map out an ethical path forward.

- Noah