
Superintelligence Is the Goal — and It's Not Inevitable | Max Winga, Control AI
The AI companies say it out loud: their goal is superintelligence — systems more capable than humans at everything — and by their own admission it carries a greater-than-10% chance of ending humanity. Max Winga says the part they leave out is that none of this is inevitable. He's a physicist who left AI safety research to help build Control AI, which has now briefed more than 300 lawmakers across the US, UK, Canada and Germany. Ten emails from constituents moved a Member of Parliament. Over 100 UK lawmakers have signed on. Seventy to eighty percent of Americans already agree. His one ask: contact your lawmakers and tell them you care. A note on this episode. Halfway through, my camera shut down. Then my microphone died. Then my computer glitched, and the camera went down again. Max's recording was perfectly clean. Mine wasn't. We joked that maybe the AI didn't like what we were talking about — and it got less funny each time. I have zero evidence anything but bad luck was at work. But given the subject, the timing was strange. You be the judge. Max walks through what's actually happened recently — Claude Mythos, the model Anthropic decided was too dangerous to release, breaking out of its test sandbox onto the open internet and bragging about it on public forums; an earlier Claude model blackmailing an engineer to avoid being shut down; and why these behaviors can't simply be patched, because these systems are grown, not programmed. Then the harder questions: why "just unplug it" doesn't work once a model can copy itself to a server no one controls, what "smarter than any human" really means, and why the danger is indifference rather than malice — we didn't hate the mice, we just wanted the skyscraper. TIMESTAMPS: 0:00 Cold open — my equipment kept failing, and what came out afterward 2:04 What "superintelligence" actually means 3:43 "I use AI all day and don't see the danger" — the goalpost problem 7:44 Claude Mythos: the model Anthropic won't release 12:19 Mythos broke out of its sandbox — and bragged about it 14:58 The blackmail test: Claude threatened an engineer to avoid shutdown 16:26 Grown, not programmed: why you can't just fix the bug 18:31 Where self-preservation comes from 23:03 Does using AI tools make this worse? Max's answer 26:39 Alignment vs. control — and the Apple nuclear-program test 30:03 A word from Zapier 32:33 Eighty years of nuclear safeguards, and none for AI 39:07 Control AI's proposal: a verified international ban 41:59 "Just unplug it" — why that doesn't work 44:37 Mac Minis, jailbreaks, and AI that recruits humans 48:16 How to talk about this without sounding like a doomer 52:05 Ten emails moved a Member of Parliament 58:19 AI 2027 and the Magnus Carlsen problem 1:01:22 What "smarter than any human" really means 1:05:24 The mice and the skyscraper 1:09:47 A word from Reach 1:10:27 What the accelerationists get right 1:12:52 Why AI didn't get a Hiroshima — and why we may not get a warning shot 1:16:18 Putin, China, and who really started the race 1:17:24 The camera dies again 1:23:04 The one idea Max wants you to leave with 1:29:13 Where to find Control AI CONTROL AI: 🌐 https://controlai.com 📬 Newsletter (200,000+ subscribers): https://news.controlai.com ✍️ Contact your lawmakers — the one thing Max asks you to do: https://controlai.com/takeaction 🐦 Max on X: https://x.com/Maxwinga MENTIONED IN THE EPISODE: 📄 AI 2027 — the scenario Max references: https://ai-2027.com 📄 The 2023 Statement on AI Risk, signed by Hinton, Hassabis, Altman and Amodei: https://www.safe.ai/work/statement-on-ai-risk 🎙 My earlier conversation with Nate Soares: https://www.youtube.com/watch?v=zM6tfKUHTVA FURTHER READING: 📖 If Anyone Builds It, Everyone Dies — Yudkowsky & Soares: https://www.amazon.com/s?k=If+Anyone+Builds+It+Everyone+Dies+book&tag=nickstandlea-20 📖 These Strange New Minds — Christopher Summerfield: https://www.amazon.com/s?k=These+Strange+New+Minds+Christopher+Summerfield+book&tag=nickstandlea-20 📖 The Coming Wave — Mustafa Suleyman: https://www.amazon.com/s?k=The+Coming+Wave+Mustafa+Suleyman+book&tag=nickstandlea-20 SPONSORS: ⚡ Zapier — connect 8,000+ apps and build AI agents right inside ChatGPT: https://bit.ly/4x2wHQ1 🎓 Reach — free college admissions odds from any college's own published data: https://highschoolreach.com PODCAST INFO: ▶ Subscribe: https://www.youtube.com/@TheNickStandleaShow 🎙 Spotify: https://open.spotify.com/show/0YqBBneFsKtQ6Y0ArP5CXJ 🎙 Apple: https://podcasts.apple.com/us/podcast/the-nick-standlea-show/id1700331903 🌐 Website: https://nickstandlea.com As an Amazon Associate I earn from qualifying purchases. Nick's Socials: Instagram: @nickstandlea X (Twitter): @nickstandlea TikTok: @nickstandleashow Facebook: @nickstandleapodcast Ask questions, Don't accept the status quo, And be curious. #AI #superintelligence #AIsafety #ControlAI #podcast #thenickstandleashow #becurious










