# What Everyone Gets Wrong About AI — Bittensor Co-Founder Explains ## Метаданные - **Канал:** Cointelegraph - **YouTube:** https://www.youtube.com/watch?v=Q39FSBCZN0k - **Дата:** 10.06.2026 - **Длительность:** 4:45 - **Просмотры:** 237 ## Описание Bittensor co-founder Ala Shaabana joins us to discuss what most people get wrong about AI, why he doesn't believe in AI apocalypse scenarios, and where decentralization can actually add value. We explore the origins of Bittensor, the challenge of incentive design, the future of decentralized AI, deepfake detection, AI agents, and how artificial intelligence is likely to reshape the economy and everyday life over the next decade. 00:00 Intro 00:15 What Do Most People Get Wrong About AI? 00:40 How Did Bittensor Start? 01:15 Should AI Be Decentralized? 01:45 What's the Hardest Problem in Decentralized AI? 02:15 What Is Crucible Labs Building? 03:10 Can AI Really Detect Deepfakes? 03:30 Is an AI Apocalypse Actually Coming? 03:50 How Will AI Change the Economy? 04:05 What Will AI Change About Everyday Life? #AI #ArtificialIntelligence #Bittensor #TAO #Crypto #DecentralizedAI #MachineLearning #AIAgents #Deepfakes #FutureOfAI Follow COINTELEGRAPH: Website: https://cointelegraph.com/ Telegram: https://t.me/cointelegraph Facebook: https://www.facebook.com/cointelegraph Twitter: https://twitter.com/cointelegraph Mobile app: https://linktr.ee/cointelegraph_mobile Cointelegraph covers everything Bitcoin, bringing you the latest news, prices, breakthroughs, and analysis, with emphasis on expert opinion and commentary from the digital currency community. ## Содержание ### [0:00](https://www.youtube.com/watch?v=Q39FSBCZN0k) Intro I think most people don't really understand how the flywheel works in AI. They're going to change the economy and change the ecosystem and change everything. Can AI models actually share data between each other and build this way, and kind of train together, instead of just training separately in a different silo? The ### [0:15](https://www.youtube.com/watch?v=Q39FSBCZN0k&t=15s) What Do Most People Get Wrong About AI? fact that AI is going to blow us all up one day, I don't think it's true. I think most people don't really understand how the flywheel works in AI, and it's like continuously creating incentives and building on top of those and kind of creating this cycle where you're continuously feeding back into the initial ecosystem in the first place. ### [0:40](https://www.youtube.com/watch?v=Q39FSBCZN0k&t=40s) How Did Bittensor Start? So at the very initial stages, what the biggest problem I found with AI is hardware. It was extremely difficult to get hardware and use that to build. If you were not in, say, the big labs like Google or in the big universities like Stanford or MIT and so on. So I started looking into the problem of, can AI models actually share data between each other and build this way and kind of train together instead of just training separately in a different silo? And that was the early stages of what eventually led me to meet Jacob, my co-founder, and eventually build a tensor. ### [1:15](https://www.youtube.com/watch?v=Q39FSBCZN0k&t=75s) Should AI Be Decentralized? That's a good question. Not every AI problem should be decentralized. There are certain problems that should be and certain problems shouldn't be. If you're building AI for, say, safety critical systems like F-35 Raptor airplanes, yes, you don't need to decentralize that. But there are certain AI otherwise that benefit from decentralization, things like, let's say, um, training or inference or even, uh, deepfake detections, right? It's good to have that decentralized. So there's no bias and ### [1:45](https://www.youtube.com/watch?v=Q39FSBCZN0k&t=105s) What's the Hardest Problem in Decentralized AI? you remove that sort of problem. Now, the most difficult part I'd say is the incentive design. And that's always a difficult problem when you're building. How do you create an ecosystem where your participants are optimizing for what it is you want to build, instead of trying to game it, or trying to find ways around it or to cheat the system? That's usually the thing most people struggle with, and that is usually the one piece that if you get it right, you're good to go. ### [2:15](https://www.youtube.com/watch?v=Q39FSBCZN0k&t=135s) What Is Crucible Labs Building? Certainly, yes. So, uh, most recently I've co-founded Crucible Labs. Crucible labs is an incubator that's sort of sitting between the growth and the product intersection of bit tensor. And what we're trying to do is build tooling for, say, um, different investors that can speculate on the ecosystem and on the different subnets and bit tensor. And we've also have been working on Incubations. We're trying to bring out projects that we believe are interesting and that are built by very smart people and that have the potential to break into the real world instead of just staying inside of crypto. So most recently has been a genetic, uh, design project. So, uh, Oro is one subnet that we've recently incubated. They have been building shopping agents to help people really, uh, market research to understand what it is that they're shopping for and how to properly, uh, how to properly retrieve it. Another one has been, uh, more on the research side of things. That was Cognito. That's more doing ### [3:10](https://www.youtube.com/watch?v=Q39FSBCZN0k&t=190s) Can AI Really Detect Deepfakes? um, a mixture of experts research on bit tensor. And our very initial one was actually called bit mind. It was more on the deepfake detections. They built a subnet to effectively detect deepfakes as you are scrolling the internets. And basically you can ask it and say, okay, is this picture real or fake? And it's the browser extension, the ability to actually tell you whether it is or not. ### [3:30](https://www.youtube.com/watch?v=Q39FSBCZN0k&t=210s) Is an AI Apocalypse Actually Coming? So I would say that the fact that AI is going to blow us all up one day, I don't think it's true. I think a lot of these models are really just very sophisticated, uh, word predictors. And they're very good at predicting the next word, but they're not going to blow us up anytime soon. They're going to make things very different. They're going to change the economy and change the ecosystem and change everything. But the level of ### [3:50](https://www.youtube.com/watch?v=Q39FSBCZN0k&t=230s) How Will AI Change the Economy? nuclear apocalypse people are expecting is not coming. I think in five, ten years, we're going to see AI change the way we live day to day. We're going to see it change the way that we go about our daily tasks is going ### [4:05](https://www.youtube.com/watch?v=Q39FSBCZN0k&t=245s) What Will AI Change About Everyday Life? to change the way that people start their careers, for example. And I think it's also going to change the way that, um, really the way we live our lives effectively. I think that's through either new job creation or changing different positions at the same time. Also change the way that we learn. And really, it's going to be much more pervasive in our everyday lives, kind of like the cell phone. --- *Источник: https://ekstraktznaniy.ru/video/53154*