#Ai Agi

Skynet is here 🤖 Everything you see and hear in this video was generated using AI prompts, no real humans or any image reference that is not a text prompt. Images: FLUX (with LoRA) Video: Kling Music: Suno [ 🎥: @amitayboneh ] #ai #chatgpt #aitools #openai #aitips #machinelearning

If artificial general intelligence is achieved, it would be able to outperform humans on most intellectual tasks. When might that happen—and how worried should you be? We asked Sam Altman, the boss of OpenAI. Click on the link in our bio to learn what else he predicted. ⁠ ⁠ #artificialintelligence #AGI #SamAltman #OpenAI #technology #AI

what’s “agi”? 🍵 one of the most widely interpreted words right now, i think there are a lot of incentives for how people use this term i hope this vid provides a bit of clarity! #art #ai #design #future #agi

This is not a joke nor an exaggeration #agi #aisafety #ai #marczellklein #marczell

AI is going to replace you, unless… #fyp #ai #startup #taste #agency #tech #content

The most important idea to understand the current limits of AI and what it means for AGI comes from complexity science. Stuart Kauffman talks about the adjacent possible - local, specific things that can be brought into being by an intelligence embedded in a living system, where new functions and possibilities emerge that could not have been specified in advance, and therefore cannot be fully explored by systems that operate over predefined spaces. Why it matters for AGI is pretty direct. If Kauffman is right, then scaling AI—more data, more compute, better architectures—does not get you to AGI in the strong sense people imply. It gets you better search and recombination within a space. But it does not give you participation in the ongoing creation of the space itself. That’s a different kind of process. It looks more like evolution than optimization. There’s still an open question, though, and it’s where current debate actually sits: whether sufficiently embodied, interactive, and open-ended AI systems could start to approximate that kind of space-expanding behavior—or whether Kauffman’s argument is a fundamental limit. #cognitivesovereignty #ai #stayhuman #agi

Jensen Huang, co-founder and CEO of Nvidia, suggested that AGI may already be within reach. His view reflects how quickly AI capabilities are advancing across coding, reasoning, and real world applications. 👉 Follow @uncover.ai to stay updated with the latest AI news — Media: Lex Fridman

The stages of AI explained. The last one changes everything—for better or worse

What happens if AI wins and AGI actually arrives in the next few years? Without jobs or purpose what becomes of us? #ai #tech #education #fyp #chatgpt

What’s smarter than AGI? And no, it’s not ASI! Here’s my (very different) take on the tech (+ we define it!) . . . Hi, I’m Harper, AI educator, engineer and advisor. ••• My goal is to democratize AI, so everyone can understand it, use it, and make informed decisions about it. Happy you’re here. ••• Join us here if you want to learn about AI, and let me know in the comments if there’s anything specific you want to learn about. #aieducator #airesearch #advancedai #aiexpert #aiforall #aiforgood #techtalks #artificial_intelligence #alwaysbelearning #moderntechnology #techvideos #techindustry #techfacts #techtalk #techinnovation #futureready #scienceandtechnology #techupdates #technologytrends #instatechnology #techgeek #techtrends #deeplearning #techlover #technologynews #machinelearning #artificialintelligence #ai #agi #openai

Demis Hassabis: The Jobs AGI Could Never Replace. Can AGI really do every job a human can do? In this clip, Google DeepMind CEO Demis Hassabis breaks down a simple but powerful example: an AI “doctor” versus a human nurse at the bedside. He explains why some tasks might be automated, but other roles will stay deeply human, even in a world of advanced AI. Today, studies from groups like the World Economic Forum and McKinsey show that AI is reshaping millions of jobs, but most roles are being transformed, not instantly wiped out. What matters is which parts of a job are about pure information and which parts are about empathy, touch, and trust. Diagnosis might be handled by an AI system in the future, but would you really want a robot holding your hand in a hospital bed? This short is about more than doctors and nurses. It is about what we actually value in human work and what we do not want to outsource to machines, no matter how smart they become. #agi #futureofwork #aiinsights #ceotalk #demishassabis #googleceo

AI has a giant problem that nobody is talking about — and it’s not hallucinations, bias, or any of the usual suspects. It’s coordination. Google DeepMind published a paper late last year — “Distributional AGI Safety” — arguing that AGI won’t emerge as one giant all-knowing model. Instead, it will emerge the way an ant colony works: thousands of specialized agents coordinating together, none of them AGI on their own, but collectively exhibiting AGI-level capabilities. They call it Patchwork AGI. It’s a beautiful idea. And it might be right. But it depends entirely on one thing: AI agents being able to actually work together. So has anyone checked if they can? Researchers at ETH Zurich just did. In their paper “Can AI Agents Agree?”, they gave multiple AI agents the simplest possible task — agree on a single number, no stakes, no conflict, no tricks. The results were not good. Agents got stuck, went in circles, and timed out without ever reaching agreement. And the more agents you added, the worse it got. Introduce even one adversarial agent into the group and the whole system collapses to zero consensus. This is a seriously understudied problem. And the research we have so far is just the beginning — this study only tested two models from the same model family, both on the smaller side at 8 and 14 billion parameters. But the question it raises is already important: if we’re building an entire agent economy — AI assistants negotiating on your behalf, autonomous companies, agent-to-agent transactions — all of it assumes coordination works. Most multi-agent systems today only function because one agent is in charge. That’s not coordination. That’s a dictatorship. The moment you need two agents with opposing interests to reach an agreement, we’re in uncharted territory. Papers mentioned: — Can AI Agents Agree? (Berdoz, Rugli, Wattenhofer — ETH Zurich, 2026) — Distributional AGI Safety (Tomašev et al. — Google DeepMind, 2025) 📩 Get my weekly AI newsletter — This Week in AI — in the link in bio. #agi #airesearch #googledeepmind #aiagents #patchworkagi