Ads skipped

Nate Soares on P(Doom), Alien Superintelligence, Human Enhancement, and the Future of AI

60K views · Jul 1, 2026 · Science & Technology

Comments · 398

  • @janekelly08 · 2 months ago

    We love you, Roman!  Thanks for all you do.

    62

  • @trvs_b · 2 months ago

    Finally! An interview of Nate Soares by an AI researcher who knows the best questions to ask. I'm so glad you created this podcast, Roman.

    130

  • @IAXYZDomains · 2 months ago

    Wow, the very best P-doom debate, no one matches this yet, thanks Roman and Nate

    31

  • @thor9563 · 2 months ago

    Kind of scary that this entire conversation - very intelligent, thoughtful, is about how to keep from being eliminated by our creation!

    15

  • @Sweetpeas-v2j · 5 days ago

    Thank you Roman & Nate for all you’re doing for the field and for humanity! I hope your message gets out to the whole world. Fascinating conversation and topics.

    2

  • @fanismoutos · 2 months ago

    Roman, hearing your esteme guest confronting the absolute emergency that threatens humanity, and hearing all these nebulous ill-defined wishful thinking,<br> &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;send a chill in my spine, &nbsp;and i thought &quot;we are doomed&quot; &nbsp;.

    21

  • @KA-dm9lm · 1 month ago

    Roman, great job letting your guest just talk. For someone as bright as you, I would guess that would be hard to do (I would wanna get my point in!). Great job man

    8

  • @HoldingJesus · 11 days ago

    I’d love to hear a new, current discussion between you and Nate &nbsp;now that more is developing and more people are becoming interested. 🙏

    1

  • @AnthonyPhilipsAI · 2 months ago

    Giving this a fair hearing. Nate Soares is a serious voice in AI safety. He is president of MIRI and has worked for years on AI alignment and smarter than human AI risk. He is not a crank, and this should not be dismissed as fear bait.<br><br>The clean way to hold this is to split the claims. One claim is that advanced AI could create extinction level risk and should be treated seriously. That is now a mainstream concern, not a fringe idea. In 2023, Geoffrey Hinton, Yoshua Bengio, Sam Altman, Demis Hassabis, Dario Amodei and many others signed the short statement saying AI extinction risk should be treated alongside pandemics and nuclear war.<br><br>A second claim is that superintelligence would kill everyone with near certainty. That is much stronger. It is closer to the Soares and Yudkowsky position than to the field as a whole. In a large survey of 2,778 AI researchers, the median respondent gave about a 5 percent chance of human level AI leading to outcomes that were extremely bad, for example human extinction. That means real risk, not zero risk, but also not settled doom.<br><br>The deeper truth underneath both sides is this: we are building increasingly capable systems quickly, and we still do not fully understand or reliably control what is happening inside them. That is the load bearing fact. The danger is not that doom is proven. The danger is that if we are wrong, we may not get a clean second try.<br><br>And maybe the argument everyone is having is the wrong one. Fighting over the exact odds that it kills us is a question nobody can honestly answer. The sharper question is this. Why are we building something we cannot undo, faster than we can understand it, when nothing sets that speed except competition and money? That speed is a choice. Right now it is not being made by the public.<br><br>So the honest position is not panic and not dismissal. Take the warning seriously. Do not treat the worst case as proven. Slow down enough to see clearly, because this is too important to let competition and money set the pace for all of humanity.

    2

  • @I-Dophler · 2 months ago

    🎯 Key points for quick navigation:<br><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0">00:00</a> <b>Insufficient Time Alignment</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=68">01:08</a> <b>Alignment Seems Solvable</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=109">01:49</a> <b>Orthogonality Thesis Explained</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=150">02:30</a> <b>Value Alignment Challenge</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=220">03:40</a> <b>Human Flourishing Concept</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=290">04:50</a> <b>Physically Possible Alignment</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=330">05:30</a> <b>Coherent Extrapolated Volition</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=372">06:12</a> <b>Do WhatIMean</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=416">06:56</a> <b>Trial Error Impossible</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=457">07:37</a> <b>Reject Negative Utilitarianism</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=541">09:01</a> <b>Human Bias Debate</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=622">10:22</a> <b>Alien Values Conflict</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=778">12:58</a> <b>Humanity Alignment Scope</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=876">14:36</a> <b>Human Values Extend</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=998">16:38</a> <b>Civilizations Converge Similarly</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1067">17:47</a> <b>Aggregation Remains Unsolved</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1138">18:58</a> <b>Alignment Versus Control</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1167">19:27</a> <b>Superintelligence Must Care</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1209">20:09</a> <b>Progress Before Perfection</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1247">20:47</a> <b>Intelligence Balances Tradeoffs</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1275">21:15</a> <b>Point Return Irreversible</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1331">22:11</a> <b>Rollback Theoretical Only</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1416">23:36</a> <b>Alignment Risks Persist</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1430">23:50</a> <b>Alien Life Likely</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1482">24:42</a> <b>Fermi Limits Distance</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1524">25:24</a> <b>Civilizations Eventually Meet</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1566">26:06</a> <b>Humanity Probably Younger</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1665">27:45</a> <b>AI Rejects Filter</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1787">29:47</a> <b>Observable Universe Changes</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1813">30:13</a> <b>Acausal Trade Skepticism</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1855">30:55</a> <b>Preserving Alien Minds</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1953">32:33</a> <b>Decision Theory Research</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=1998">33:18</a> <b>Investigate Theory Anomalies</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2093">34:53</a> <b>Intelligence Theory Gaps</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2216">36:56</a> <b>Scaling Without Understanding</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2271">37:51</a> <b>Timeline Highly Uncertain</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2341">39:01</a> <b>Pause AI Development</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2369">39:29</a> <b>Political Awareness Growing</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2469">41:09</a> <b>Book Advocacy Began</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2495">41:35</a> <b>Stop Superintelligence Race</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2521">42:01</a> <b>Government Awareness Increasing</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2566">42:46</a> <b>Cybersecurity Alarm Raised</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2594">43:14</a> <b>Awaken Bus Driver</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2620">43:40</a> <b>Presidential AI Briefings</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2694">44:54</a> <b>Collaborative Book Writing</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2722">45:22</a> <b>Expanded Online Resources</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2780">46:20</a> <b>Draft Governance Framework</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2821">47:01</a> <b>Avoid Creating Superintelligence</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2863">47:43</a> <b>Compute Monitoring Proposed</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2905">48:25</a> <b>Conservative Training Oversight</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=2960">49:20</a> <b>Dynamic Governance Needed</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3001">50:01</a> <b>Restrict Algorithmic Research</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3074">51:14</a> <b>Flexible Legal Framework</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3129">52:09</a> <b>Current AI Acceptable</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3211">53:31</a> <b>Research Capability Concern</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3250">54:10</a> <b>Rollback Considered Sensible</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3333">55:33</a> <b>Pdoom Context Matters</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3376">56:16</a> <b>Human Future Preservation</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3402">56:42</a> <b>Bus Cliff Analogy</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3486">58:06</a> <b>Superintelligence Extremely Dangerous</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3571">59:31</a> <b>Public Awareness Growing</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3624">01:00:24</a> <b>Humanity Could Respond</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3650">01:00:50</a> <b>Global Pause Required</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3678">01:01:18</a> <b>Nuclear Style Coordination</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3706">01:01:46</a> <b>Decades Worth Buying</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3734">01:02:14</a> <b>Human Enhancement Opportunity</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3776">01:02:56</a> <b>Alignment Trial Limitations</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3832">01:03:52</a> <b>Von Neumann Example</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3885">01:04:45</a> <b>Intelligence Theory Foundations</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3923">01:05:23</a> <b>Smarter Humans Hope</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=3965">01:06:05</a> <b>Enhanced Humans Risk</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4006">01:06:46</a> <b>Evolution Missed Targets</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4048">01:07:28</a> <b>Humans Share Values</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4089">01:08:09</a> <b>Altruistic Enhancement Preferred</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4131">01:08:51</a> <b>AI Drives Differ</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4159">01:09:19</a> <b>Governance Enforcement Strategy</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4172">01:09:32</a> <b>Chip Supply Control</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4238">01:10:38</a> <b>Self Destruct Chips</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4298">01:11:38</a> <b>Superintelligence Exceeds Nukes</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4355">01:12:35</a> <b>Diplomatic Pressure First</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4422">01:13:42</a> <b>Sabotage Remains Option</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4489">01:14:49</a> <b>Diplomatic Clarity Essential</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4588">01:16:28</a> <b>Illegal Research Punished</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4631">01:17:11</a> <b>Garage AI Prohibited</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4702">01:18:22</a> <b>Restrictions Potentially Temporary</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4771">01:19:31</a> <b>Political Support Growing</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4839">01:20:39</a> <b>Regulation Becoming Accepted</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4877">01:21:17</a> <b>Bipartisan Concern Increasing</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4936">01:22:16</a> <b>Awareness Rapidly Expanding</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4962">01:22:42</a> <b>Treaty Still Distant</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=4976">01:22:56</a> <b>Hope Increasing Worldwide</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5018">01:23:38</a> <b>Labs Cannot Coordinate</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5088">01:24:48</a> <b>Corporate Pause Insufficient</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5114">01:25:14</a> <b>Global Treaty Required</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5144">01:25:44</a> <b>International Focus Needed</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5172">01:26:12</a> <b>Acceleration Arguments Unconvincing</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5211">01:26:51</a> <b>Brakeless Racecar Analogy</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5268">01:27:48</a> <b>Brakes Unlikely Built</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5309">01:28:29</a> <b>Risk Remains Unjustified</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5336">01:28:56</a> <b>Cancer Diagnosis Analogy</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5393">01:29:53</a> <b>Seek Practical Solutions</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5417">01:30:17</a> <b>Alignment Remains Unsolved</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5447">01:30:47</a> <b>Human Enhancement Option</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5474">01:31:14</a> <b>Optimistic Clickbait Title</b><br><a href="https://www.youtube.com/watch?v=MMMqQm22LS0&amp;t=5525">01:32:05</a> <b>Hopeful Farewell Exchange</b><br><br>Made with HARPA AI

    11

  • @TheKilkenny1967 · 11 days ago

    Excellent conversation and guest Roman

    8

  • @sulehan1 · 9 days ago

    Amazing conversation from two beautiful minds. Thank you Nate being the voice of most needed reason. Roman please do an episode with Daniel Kokotajlo.

    1

Up next

LIVE

Max Tegmark’s P(Doom) Is Over 90%

The Roman Forum with Roman Yampolskiy and Future of Life Institute · 92K views

LIVE

AI Safety Expert: Ban Superintelligence!

The Roman Forum with Roman Yampolskiy and ControlAI · 57K views

LIVE

Training Sand to Think: Artificial General Intelligence & Future of Physics

Perimeter Institute for Theoretical Physics · 287K views

LIVE

Is AI Hiding Its Full Power? With Geoffrey Hinton

StarTalk · 3.7M views

LIVE

Roman Yampolskiy on Jacob Coxon’s AI Warning: What Happens Next? | For Humanity #94

The AI Risk Network | AI Safety and The Roman Forum with Roman Yampolskiy · 18K views

LIVE

Prof. Nick Bostrom - What Happens When AI Escapes Our Control?

The Peter McCormack Show · 125K views

LIVE

We have 3 years to solve alignment before superintelligence

80,000 Hours · 19K views

LIVE

Roman Yampolskiy vs Emad Mostaque: I Was The Only Optimist

Dr Brian Keating and The Roman Forum with Roman Yampolskiy · 142K views

LIVE

Stop Torturing AIs! Cameron Berg - AI consciousness and welfare researcher.

The Roman Forum with Roman Yampolskiy and AM I? · 83K views

LIVE

AI Pioneer Geoffrey Hinton: AI Is Conscious, Superintelligence is Coming, And We Should Be Worried

Alex Kantrowitz · 407K views

LIVE

Roman Yampolskiy: Dangers of Superintelligent AI | Lex Fridman Podcast #431

Lex Fridman · 978K views

LIVE

Why Building Superintelligence Means Human Extinction (with Nate Soares)

Future of Life Institute · 31K views

LIVE

He Puts Humanity’s Odds at 50-50. He Builds Anyway. | Emad Mostaque

Duncan CJ · 13K views

LIVE

AI is MUTATING: And We Don't Know What It is Doing | Connor Leahy

The Peter McCormack Show · 1.6M views

LIVE

Is AI the Great Filter? | The Fermi Paradox

Isaac Arthur · 129K views

LIVE

An AI 2027 Author Returns, With a Plan to Change the Ending

80,000 Hours · 44K views

LIVE

Philosopher David Chalmers asks: When we talk to AI, what are we talking to?

UC Berkeley and Berkeley Arts & Humanities · 204K views

LIVE

AI has hacked the code of human civilization | Yuval Noah Harari

Yuval Noah Harari · 1.3M views

LIVE

ChatGPT Whistleblower: They Offered Me $2m To Keep Quiet, No One Is Ready For What's Coming!

The Diary Of A CEO · 19M views

LIVE

Ajeya Cotra – "This might be the clearest warning shot we ever get"

Dwarkesh Patel · 1.5M views

LIVE

I Run the World’s Oldest AI Safety Lab... There’s a “High Double-Digit” Chance AI Kills Us

Sean Kim and Machine Intelligence Research Institute · 162K views

LIVE

Whistleblower REVEALS REAL REASON AI Execs Freaking Out NOW

Breaking Points · 1M views

LIVE

The AI Threat Is MUCH Worse Than You Thought | Nate Soares Interview

Novara Media · 196K views

LIVE

"If Anyone Builds It, Everyone Dies" | Nate Soares

The Peter McCormack Show · 225K views

LIVE

A.I. Futurist: What Your Life Looks Like In 2028

Mighty Pursuit · 212K views

LIVE

No, AI Isn’t Conscious; It’s Actually Much Worse | AI Scientist

Decoded Genius Clips · 2.5M views

YouTube, with the door locked.

Aegis plays a clean stream instead of YouTube's player, so pre-roll ads, trackers, and fingerprinting never ride along. Drop Shields any time if you want the official player back.