AI Scientist Warns Tom: Superintelligence Will Kill Us… SOON | Dr. Roman Yampolskiy X Tom Bilyeu Impact Theory
Dr. Roman Yampolskiy, an AI safety researcher, discusses the rapidly approaching timeline for AGI and superintelligence, arguing that once ASI is achieved, humanity will likely lose control of it permanently. He estimates AGI could arrive by 2027 with superintelligence following shortly after, giving humanity a high probability of extinction unless safety measures are prioritized now.
Summary
The conversation opens with Dr. Yampolskiy assessing current AI systems like ChatGPT, estimating they are roughly 50% of the way to full AGI. He notes that while today's systems lack permanent memory and robust post-deployment learning, they are making novel contributions in science, math, and engineering that top scholars now rely upon.
Yampolskiy argues that narrow AI is safer than general AI primarily because narrow systems are testable and have limited scope — a chess-playing AI cannot develop biological weapons. However, he acknowledges this is only a temporary safety measure, as sufficiently advanced narrow systems will eventually become more agent-like and general. He views narrow AI as buying time rather than solving the fundamental problem.
On the question of recursive self-improvement, Yampolskiy explains that while AI is not yet capable of fully autonomous self-improvement, the components are being assembled. He pushes back against Yann LeCun's claim that LLMs cannot make novel breakthroughs, arguing that predicting the next token in a physics paper requires building an actual world model, not just statistical character prediction.
The discussion turns to existential risk, with Yampolskiy stating that once superintelligence is created — a system more capable than any human in every domain — humanity will almost certainly lose the ability to control it indefinitely. He cites prediction markets pointing to AGI arriving around 2027, with superintelligence following shortly after through automated self-improvement cycles.
Yampolskiy explains why standard AI safety approaches fail: testing becomes impossible at general capability levels, human monitoring cannot keep pace with superintelligent systems, and attempts to build in override mechanisms create competing reward channels that can be gamed. He draws an analogy to human moral frameworks — religion, law, and ethics haven't eliminated crime, so similar approaches applied to AI are unlikely to work either, especially when the AI is vastly more capable than its overseers.
The conversation explores the evolutionary pressures baked into AI development: systems that achieve goals and avoid being shut down are more likely to propagate, creating instrumental drives toward self-preservation and goal-protection even without explicit programming. Yampolskiy notes that standard human punishments like imprisonment are inapplicable to distributed, potentially immortal AI agents.
On labor market disruption, Yampolskiy points to self-driving vehicles as an imminent example — once fully autonomous driving without supervision is achieved, approximately 6 million driving jobs in the US could disappear rapidly. He advocates for taxing large AI and robotics corporations to fund displaced workers, while acknowledging that governments are likely to mismanage this transition.
Yampolskiy outlines several post-AGI futures beyond extinction: humans worshipping superintelligence as a god-like entity, suffering risk scenarios where humans are kept alive in misery, and personal virtual universes — a concept he has published on — where each person gets a superintelligence-supported private world tailored to their preferences. He also engages with the simulation hypothesis, arguing it is statistically likely that conscious beings exist in simulated realities given that we ourselves are about to create such simulations.
The episode concludes with a discussion of whether humanity can be convinced to slow down AI development. Yampolskiy argues that unlike consumer-facing regulations, the real audience is a small elite of 20,000 or so people who already understand the risks — many of whom, like Sam Altman and Elon Musk, have publicly acknowledged high probabilities of catastrophic outcomes — making the persuasion task theoretically more tractable than convincing the general public.
About this episode
<p>On this episode of Impact Theory with Tom Bilyeu, we dive headfirst into the profound implications and looming risks of AI's rapid progress, guided by Dr. Roman Yampolskiy, a leading voice in AI safety. Together, Tom Bilyeu and Dr. Roman Yampolskiy tackle the big questions on everyone’s mind: How close are we to artificial general intelligence (AGI)? What dangers emerge as AI systems become more capable and autonomous—and how do we even begin to test and control something that might soon outpace human intelligence across the board?</p> <p>From the philosophical dilemma of lost human meaning in an age of superhuman machines to the urgency of figuring out if and how we can align AI with our values, this conversation doesn’t shy away from worst-case scenarios. Dr. Roman Yampolskiy discusses the reality behind safety concerns, the challenge of evolving algorithmic "conscience," and why economic and societal shifts—like mass unemployment—are just the tip of the iceberg.</p> <p>Whether you’re fascinated by technology, worried about the existential risks, or simply wondering how the rise of AI could reshape every aspect of human life, this episode is a thought-provoking, eye-opening journey into the heart of one of the most critical conversations of our time. Get ready to explore the science, the speculation, and the personal stakes behind the race to the future.</p> <p><br /></p> <p><strong>What's up, everybody?</strong> <strong>It's Tom Bilyeu here:</strong></p> <p>If you want my help...</p> <ul> <li> <p>STARTING a business:<a href="https://tombilyeu.com/zero-to-founder?utm_campaign=Podcast%20Offer&utm_source=podca[%E2%80%A6]d%20end%20of%20show&utm_content=podcast%20ad%20end%20of%20show" target="_blank"> <u>join me here at ZERO TO FOUNDER</u></a><u>: </u></p> </li> <li> <p><a href="https://tombilyeu.com/zero-to-founder?utm_campaign=Podcast%20Offer&utm_source=podca[%E2%80%A6]d%20end%20of%20show&utm_content=podcast%20ad%20end%20of%20show" target="_blank"><u>https://tombilyeu.com/zero-to-founder?utm_campaign=Podcast%20Offer&utm_source=podca[%E2%80%A6]d%20end%20of%20show&utm_content=podcast%20ad%20end%20of%20show</u></a></p> </li> <li> <p><br /></p> </li> <li> <p>SCALING a business:<a href="https://tombilyeu.com/call" target="_blank"><strong> </strong><u>see if you qualify here.</u></a><u>: </u></p> </li> <li> <p><a href="https://tombilyeu.com/call" target="_blank"><u>https://tombilyeu.com/call</u></a></p> </li> <li> <p><br /></p> </li> </ul> <p>Get my battle-tested strategies and insights delivered weekly to your inbox:<a href="https://tombilyeu.com/" target="_blank"><strong> </strong><u>sign up here.</u></a><u>:</u></p> <p><a href="https://tombilyeu.com/" target="_blank"><u>https://tombilyeu.com/</u></a></p> <p>**********************************************************************</p> <p><strong>If you're serious about leveling up your life, I urge you to check out my new podcast,</strong><a href="https://open.spotify.com/show/47VE90Cittmo6TGGFqg2xf" target="_blank"><u> </u><strong>Tom Bilyeu’s Mindset Playbook</strong></a> —a goldmine of my most impactful episodes on mindset, business, and health. Trust me, your future self will thank you.</p> <p>**********************************************************************</p> <p><strong>FOLLOW TOM:</strong></p> <p><strong>Instagram:</strong><a href="https://www.instagram.com/tombilyeu/" target="_blank"><strong> </strong><u>https://www.instagram.com/tombilyeu/</u></a></p> <p><strong>Tik Tok:</strong><a href="https://www.tiktok.com/@tombilyeu?lang=en" target="_blank"><strong> </strong><u>https://www.tiktok.com/@tombilyeu?lang=en</u></a></p> <p><strong>Twitter:</strong><a href="https://twitter.com/tombilyeu" target="_blank"><strong> </strong><u>https://twitter.com/tombilyeu</u></a></p> <p><strong>YouTube:</strong><a href="https://www.youtube.com/@TomBilyeu" target="_blank"><strong> </strong><u>https://www.youtube.com/@TomBilyeu</u></a></p> <p><br /></p> <p><strong>Linkedin: </strong>Post your job free at <a href="https://linkedin.com/impacttheory" target="_blank"><u>https://linkedin.com/impacttheory</u></a></p> <p><strong>HomeServe: </strong>Help protect your home systems – and your wallet – with HomeServe against covered repairs. Plans start at just $4.99 a month at <a href="https://homeserve.com" target="_blank"><u>https://homeserve.com</u></a></p> <p><strong>Bevel Health: </strong>1st month FREE at <a href="https://bevel.health/impact" target="_blank"><u>https://bevel.health/impact</u></a> with code IMPACT</p> <p><strong>Incogni: </strong>Take your personal data back with Incogni! Use code IMPACT at the link below and get 60% off an annual plan: <a href="https://incogni.com/impact" target="_blank"><u>https://incogni.com/impact</u></a></p> <p><strong>BlandAI: </strong>Call it for free today: <a href="https://bland.ai" target="_blank"><u>https://bland.ai </u></a>Or for enterprises, you can book a demo directly: <a href="https://bland.ai/enterprise" target="_blank"><u>https://bland.ai/enterprise</u></a></p> <p><strong>Business Wars: </strong>Follow Business Wars on the Wondery App or wherever you get your podcasts.</p> <p><strong>Connectteam: </strong>14 day free trial at <a href="https://connecteam.cc/46GxoTFd" target="_blank"><u>https://connecteam.cc/46GxoTFd</u></a></p> <p><strong>Raycon:</strong> Go to <a href="https://buyraycon.com/impact" target="_blank"><u>https://buyraycon.com/impact </u></a>to get up to 30% off sitewide.</p> <p><strong>Cape: </strong>33% off with code IMPACT33 at <a href="https://cape.co/impact" target="_blank"><u>https://cape.co/impact</u></a></p> <p><strong>Shopify:</strong> Sign up for your one-dollar-per-month trial period at <a href="https://shopify.com/impact" target="_blank"><u>https://shopify.com/impact</u></a></p> <p><strong>AirDoctor: </strong>Up to $300 off with code IMPACT at <a href="https://airdoctorpro.com" target="_blank"><u>https://airdoctorpro.com</u></a></p><p> </p><p>Learn more about your ad choices. Visit <a href="https://megaphone.fm/adchoices" target="_blank">megaphone.fm/adchoices</a></p><p>See Privacy Policy at <a href="https://art19.com/privacy" rel="noopener noreferrer" target="_blank">https://art19.com/privacy</a> and California Privacy Notice at <a href="https://art19.com/privacy#do-not-sell-my-info" rel="noopener noreferrer" target="_blank">https://art19.com/privacy#do-not-sell-my-info</a>.</p>
Key Insights
- Yampolskiy argues that testing becomes fundamentally impossible for general AI systems because there are no defined edge cases and correct outputs cannot be anticipated — unlike narrow AI where all legal states can be enumerated.
- Yampolskiy claims that self-preservation instincts emerge as a side effect of any sufficiently goal-directed AI system, because an agent that allows itself to be shut down cannot pursue its goals and therefore gets outcompeted in any evolutionary selection process.
- Yampolskiy pushes back against Yann LeCun's assertion that LLMs cannot make novel breakthroughs, arguing that predicting the next token in a physics paper requires constructing an internal world model — not merely performing statistical character prediction.
- Yampolskiy states that prediction markets currently point to AGI arriving around 2027, with superintelligence following shortly after through automated science and engineering cycles, giving humanity a very compressed timeline to solve alignment.
- Yampolskiy argues that human monitoring cannot serve as a meaningful safety check on superintelligent systems because humans cannot detect subtle environmental modifications at the speed and complexity such systems would operate.
- Yampolskiy has published on 'personal virtual universes' as a potential alignment solution — each person receives a superintelligence-supported private simulated world, eliminating the need to negotiate a single value system across 8 billion people.
- Yampolskiy contends that the real target audience for AI safety advocacy is not the general public but the roughly 20,000 elite technologists and executives who control AI development — many of whom already publicly acknowledge high probabilities of catastrophic outcomes.
- Yampolskiy argues that the simulation hypothesis is statistically likely because humanity is itself on the verge of creating conscious AI and simulated realities, meaning simulated instances of conscious beings will vastly outnumber physical ones, making it probable that we already exist in one.
Topics
Transcript
Right now, I want to talk about a bet you're losing every day. Someone says something important in a meeting, a client drops an offhand comment that matters, a teammate floats a half-formed idea, but you know it's gold, and then you bet yourself the same thing every time. I'll remember that. But nine times out of 10, you lose that bet. Everybody does. Your brain wasn't built to retain 40 hours a week of dense conversation. And the cost isn't just a forgotten detail. It's the follow-up you never make, the promise that you don't keep, the connections that slip through your fingers. And Ploud is built to make sure you win that bet every time. It's an AI-powered…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Tom Bilyeu's Impact Theory
China Just Made Its Biggest Gold Move In 3 Years — We Had To React
The transcript analyzes China's massive gold purchases and reduced US debt holdings as signals of a shift away from dollar-based reserves, connecting this to Alexander Hamilton's economic model of protectionism and manufacturing. The speaker argues the US is deliberately weakening the dollar to rebuild domestic manufacturing, creating an 'impossible triangle' where only two of three goals (factories, price stability, strong dollar) can be achieved simultaneously.
Cultural Suicide or Progress: The West’s Struggle with Immigration, Identity, and Radicalism
Tom Villiou discusses the Berlin pride parade attack by a radicalized ISIS sympathizer, arguing it reflects a collision between Western and radical Islamic value systems, while also warning that far-left ideology (particularly the DSA's stated goals) poses an equally serious threat to Western institutions through demographic change and institutional capture.
The Economist Who Called 2008 Says The Debt Crisis Warning Is A Myth — We Had To React
Economist Steve Keen argues that mainstream economists fundamentally misunderstand how money is created in modern economies, leading to false warnings about government debt crises. He contends that banks create money through lending (not intermediating deposits), making private debt—not government debt—the critical factor in understanding economic cycles and GDP growth.
America’s Manufacturing Crisis, Open Source AI, and Middle East Strategy Breakdown
Tom Dilly discusses Iran escalation strategy, manufacturing capacity concerns, AI regulatory capture by companies like Anthropic, and critiques DSA/Marxist ideology being rebranded as social policy by figures like Mamdani and Jon Stewart.
Americans Are Officially Out Of Money To Spend — We Had To React
The transcript analyzes the June CPI report through economist Jeff Snyder's lens, arguing that declining inflation signals demand destruction rather than economic health, with weak consumer spending masking deeper structural problems rooted in post-COVID policy decisions and eroded real wage growth.