OpenAI Is OUT: Google Just Won the AI War

While two billionaires were screaming at each other in a courtroom, Google walked on stage and lapped the entire industry.

In this transmission of Sector M31, host Yasemin breaks down Gemini Omni — Google's new AI that doesn't just generate video, it understands physics. Drop in a voice note, a photo, a sketch, and it builds a realistic world with sound, weight, and gravity that obeys the rules of reality. Oh, and Google is spending $190 billion on AI this year alone — with analysts predicting it overtakes OpenAI by the end of 2026.

The company everyone wrote off in 2023 just became the one to beat. And it happened while you were looking the other way.

━━━━━━━━━━━━━━━━━━

🛰️ ABOUT SECTOR M31

Sector M31 is a weekly transmission from the edge of the AI revolution. Each episode, Yasemin breaks down the AI news that actually matters — the lawsuits, the breakthroughs, the quiet power moves — translated for normal humans, minus the jargon and the hype. We're living through the moment sci-fi predicted. Welcome to the broadcast.

━━━━━━━━━━━━━━━━━━

🎬 IN THIS EPISODE

• Google's stunning comeback — from the Bard disaster to the front of the pack
• What a "world model" is, and why physics changes everything
• The marble, the bell, and the demo that broke people's brains
• Why $190 billion means nobody can slow down now
• When video stops being proof — and what that does to all of us

━━━━━━━━━━━━━━━━━━

🔗 SUBSCRIBE & FOLLOW

‣ Subscribe to Sector M31 for new transmissions every week ‣ Hit the bell so you don't miss the next signal ‣ Coming up: Elon dissolves his own AI company, and the AI that found a 27-year-old secret

━━━━━━━━━━━━━━━━━━

🎙️ HOST

Yasemin is a writer, performer, and broadcaster. Sector M31 is her transmission from the edge of the AI revolution.

━━━━━━━━━━━━━━━━━━

#️⃣ TOPICS

#GeminiOmni #Google #OpenAI #ArtificialIntelligence #AINews #WorldModel #AIVideo #DeepMind #ChatGPT #Claude #AI2026 #FutureOfAI #DeepFakes #SectorM31

Transcript

SPEAKER_00 0:08

All right. So while Elon Musk and Sam Albin were in a courtroom in Oakland yelling at each other about who really owns the future of artificial intelligence, Google was on stage at their developer conference quietly going, oh hey, by the way, we built an AI that understands physics. It can watch a video and predict what happens next. You can tell it a story out loud and it'll turn it into a little movie. And we're spending $190 billion on AI this year. Have a great night. And basically nobody noticed because two billionaires were busy fighting over charity. Today, the company everyone wrote off two years ago just realized what might be the most powerful AI tool ever shipped to regular people. It's called Gemini Omni. And I think we need to talk about what it actually is, what it actually does, and the part nobody's saying out loud, which is that this thing might change what real means on the internet. Welcome back. I'm Yasmin, and we are getting into Google. Okay, quick history. Because to understand why this is a huge deal, you need to remember what Google where Google was about three years ago. November 2022. ChatGPT launches, the whole world loses its mind, and inside Google, there is reportedly a full-blown panic. Because Google had been working on AI quietly in their yeah in their labs for like a decade. They invented a lot of the technology that ChatGPT is built on. And then a tiny startup with 200 employees beat them to the punch and stole all the oxygen. Google launches their own chat bot a chatbot called BARD. It's terrible. The first public demo gets a fact wrong on stage. The stock drops $100 billion in a day. There are articles about how Google is finished, how they had their lunch eaten, how they're the Kodak of AI. Cut to today, May 2026. Google's AI is now called Gemini. There are over 600 million people using it. They just walked on stage at their developer conference. That's I.O. by the way. That's what the I.O. thing is. It's basically Google's version of the Apple keynote. And they unveiled a new model called Gemini Omni. They also casually mentioned that they are spending $190 billion on AI infrastructure this year. Just this year, which is, for reference, that is six times what they spent in 2022. It is more than the entire GDP of like Hungary. And here's the analyst quote that's been bouncing around the tech press this week. There's a guy named Nate Elliott. He watches this industry and he basically said by the end of 2026, Google will have overtaken OpenAI as the leading source of consumer AI engagement, which is a fancy way of saying the company everyone wrote off three years ago is about to lap the entire field. They're closer to Chat GPT than most people realize. And this Omni thing they just announced is the move that might do it. Alright, so what is this thing? The simplest way to explain Gemini Omni is you give it anything, your voice, a photo, a sentence, a rough drawing on a napkin, and it makes a video with sound, with realistic physics of uh whatever you described. That's it. That's the pitch. Create anything from any input. That's literally how Google described it. Now, AI that makes video is not new. We've had this for a couple of years. Open AI had a thing called Sora. Google had a thing called Vio. They were all fine. They could make five-second clips of a cat walking across the kitchen. Sometimes the cat had six legs, sometimes the kitchen melted. It was clearly AI. But Omni is different. And the reason is different is one word, physics. Let me describe one of the demos that they showed on stage. They gave the model a prompt. I think it was something like marble rolling down a wooden ramp into a bell. And the video it generated, the marble actually behaved like a marble. The wood actually sounded like wood. When the marble hit the bell, the bell rang with the correct sound. At the correct moment, with the right amount of vibration, the whole thing obeyed gravity. The whole thing obeyed weight, which sounds like the world's most boring sentence. The marble obeyed gravity. But this is actually a massive deal because every other AI video model up until now has been doing basically really fancy guessing. It looks at a million videos of marbles and it goes and it goes, marbles usually look like this. So I'll draw something that looks like that. Omni is doing something different. Google is calling it a world model, which means, and I'm simplifying, the AI has built an internal understanding of how the actual physical world works. Things like weight, liquids pour, fabric drapes, light bounces, sound matches the action. And the implication of that is enormous because if an AI understands how the world works, it can predict what happens next. It can take a video you give it and continue it. It can take a photo of you in your kitchen and animate you cooking dinner. It can take a sketch of your uh a sketch your kid drew and turn it into a Pixar short. There was another demo where they made a claimation video explaining how protein folding works, not by editing the footage. The AI just built the claymation from scratch, from a description. And here's the kicker: you can edit it by talking to it, like the video already generated, and you go, actually, uh, can you make the lighting more cinematic and put the character in a red jacket instead? And the AI redoes the video, keeping the character the same person, keeping the scene consistent, just changing what you asked it to change. This is something AI has been notoriously bad at. Character consistency, like you'd ask it to draw your character five times and you'd get five different people. Omni keeps the character. The character is the character across multiple edits, across multiple shorts. Okay, where can you actually use this thing? Because the answer is more interesting than I expected. The version they're shipping right now is called Gemini Omni Flash. Flash in Google AI naming basically means the fast cheap one. There's apparently a more powerful version called Omni Pro coming later. If you pay for Google's AI subscription, which starts at about eight bucks a month, you get it today, right now, in their Gemini app and in this thing called Google Flow, which is their video editor. If you don't pay for anything, you get it free in YouTube Shorts, which is when you stop and think about it for a second, kind of a wild distribution strategy. Because YouTube Shorts is TikTok. It's the swipe up endlessly scroll video thing. It has, I don't have the exact number, billions of users. And Google just put the most powerful video generation AI ever shipped directly into that app for free for everybody, which means in the next few weeks, your YouTube Shorts feed is going to be different. There's going to be way more AI content, some of it amazing, some of it horrifying, all of it indistinguishable from real video unless you're really paying attention. Now, Google is trying to be responsible about this. Every video Omni makes has something called a synth ID watermark, which is basically an invisible signature baked into the file that says this was made by AI. The idea being if a news outlet or fact checker needs to verify whether something is real or fake, they can scan it and find out, which is great. In theory, in practice, I don't know. How often does the average person fact-check a video they see on shorts? Be honest, you don't, I don't, nobody does. We see a thing, we react, we scroll. The watermark exists. Whether it matters in the real world is a completely separate question. Also worth mentioning, Omni won't edit speech in videos yet. Like you can make a video of a person, but you can't make them say specific words yet. Google's holding that one back. They said, and this is a real quote, they want to bring this capability to users responsibly. Which I read as we know exactly how bad the deep fake problem is, and if we ship this one day, we better be careful. So we are going to wait until everyone else does it first, and we're not the ones who get blamed. Okay, big picture. Why does any of this matter to people who don't work in tech and don't care which billionaire is winning? Here's the thing I keep coming back to. Up until basically this week, video was kind of the last form of evidence we had, right? Like photos can be photoshopped, audio can be cloned, text has never been reliable, but video, video was always the gold standard. If you saw a video of something, you basically believed it happened. And the reason video stayed reliable for so long is that faking video used to be incredibly hard. It took Hollywood Studio, it took specialists, it took millions of dollars, which meant the supply of fake video was very small. Omni. And the things that will come after Omni basically destroys that. Anyone with a phone and eight bucks a month can now generate convincing video of anything they want. And before you say, Oh, that's been true for a while, deep fakes have been around for years. Yes, they have, but they were bad. They were uncanny, valley, creepy. You couldn't tell. Your brain knew, this is different. The physics demo I described earlier, most people couldn't tell. The marble looks right, the sound is right, the lighting is right. So we are now living in a world starting basically this week where the default assumption about a video you see online has to flip. You don't get to assume it's real anymore. You have to ask, how do I know this is real? And nobody has trained us for that. Nobody has thought about what that does to a society, to a courtroom, to a news cycle, to a marriage where one person sees a video of the other person doing something they didn't do. And I'm not saying that to be alarmist. I'm not saying AI is going to ruin everything. I'm saying we just crossed a line this week, very quietly, while two billionaires were yelling at each other about a charity in Oakland. And then the other thing, the thing I keep coming back to is the money. $190 billion in one year from one company on AI. That's not a science project, that's not a research grant, that's not a bet. Google is betting that this technology is going to remake the entire economy, and they want to own the company that does the remaking. They have to win. Once you're 190 billion dollars in, you can't decide, oh, you know what? Maybe we should uh slow down and be careful. The shareholders won't let you, the market won't let you, the board will fire you. So all of these companies, Google, OpenAI, Antropic, Meta, they are now locked in. They have to ship, they have to win. Caution is now a luxury they cannot afford. And we, meaning normal humans, are the ones who are going to live in the world that produces. Alright, so that's Gemini Omni. Google just shipped what might be the most powerful video generation tool ever released to the public. It understands how physics works. It edits by conversation. It's already in YouTube Shorts. It's already running on $190 billion of infrastructure. And it's just the Flash version. The pro version is coming. My honest reaction: equal parts excited and worried. Like the protein folding claymation demo was genuinely cool. There's a version of this where Omni helps teachers make incredible lessons, and small filmmakers make movies they couldn't afford before, and grandparents make little videos for the grandkids. Real good stuff. And there's also a version of this where my mom sees a video of me on Facebook doing something I never did, and there's no way to convince her it's not real. Both of those futures are now possible. This week. Next week, and this one is also bonkers, we are going to talk about what Elon did the week he lost his lawsuit. Spoiler, he dissolves his entire AI company, folded it into SpaceX, which, yeah, we're going to get into that. If this helped you understand what the hell is going on, share it with somebody who keeps seeing AI headlines and going, what does this mean? I'm trying to make this show that thing. I'm Yasmin. Thanks for listening. See you in the next one.