Skip to content
Modern WisdomModern Wisdom

AI DEBATE: “We’re Due For A Chernobyl Event”

In this AI debate, we explore: * Whether humans will exist in 2040. * What will happen once we reach AGI. * Whether AI gets smart enough to act malevolently or benevolently towards humans. * If AI will grow powerful enough to be out of human control. * and much more... Guests * Zack Kass is an AI advisor, author, and former Head of Go-to-Market at OpenAI. * Liv Boeree is a physicist, science communicator, podcaster and former professional poker champion. * Aric Floyd is an AI writer, and video producer. - Get up to $350 off the Eight Sleep Pod 5 at https://eightsleep.com/modernwisdom Get a Free Sample Pack of LMNT’s most popular flavours with your first purchase at https://drinklmnt.com/modernwisdom Get 35% off your first subscription on the best supplements from Momentous at https://livemomentous.com/modernwisdom Get 15% off your first order of my favourite Non-Alcoholic Brew at https://athleticbrewing.com/modernwisdom Get ChatGPT to explore ideas, solve problems, and learn faster at ⁠⁠https://chatgpt.com - 0:00 Will Humans Still Be Around in 2040? 8:20 What Jobs Will Be Protected From AI? 14:22 Finding Meaning in a Post-Work World 23:13 Technology is Pulling Us From Our Purpose 27:12 Where Do We Draw the Line With Outsourcing Thinking? 39:58 Is AI Unlocking Human Potential? 42:28 Refusal or Obedience: What’s More Dangerous? 46:01 The AI Attack That Should Worry Everyone 54:44 Should We Slow Down Advanced AI? 56:41 Does Recursive Research Actually Work? 01:10:52 How Safe is AI? 01:14:01 Can We Reach a Peace Deal For a Super-intelligent World? 01:22:38 Why AI 2040 Got So Much Right 01:26:21 Is AGI Inevitable? 01:30:39 Could Techno-Pastoralism Be the Future? 01:37:16 Is AI Replacing Human Connection? 01:45:36 How Do We Navigate the Purpose-Friction Problem? 02:00:50 Why Data Centre Policy Matters More Than Ever 02:05:56 The Concentration of Power Concern 02:11:39 Is the Frontier Strategy the Winning Condition? 02:18:17 Will AI Weaken Democracy? 02:29:11 How We Can Limit the Concentration of Power 02:31:45 Is ChatGPT Pessimistic About the Future? 02:37:54 What Should We Be Focused On? - Get access to every episode 10 hours before YouTube by subscribing for free on Spotify - https://spotify.modernwisdom.com or Apple Podcasts - https://apple.modernwisdom.com Get my free Reading List of 100 life-changing books here - https://chriswillx.com/books/ Try my productivity energy drink Neutonic here - https://neutonic.com/modernwisdom - Get in touch in the comments below or head to... Instagram: https://www.instagram.com/chriswillx Twitter: https://www.twitter.com/chriswillx Email: https://chriswillx.com/contact/

Chris WilliamsonhostAric FloydguestLiv Boereeguest
Aug 17, 20262h 42mWatch on YouTube ↗

CHAPTERS

  1. 0:00 – 3:23

    2040 baseline forecasts: survival, homogeneity, and concentrated power

    Chris asks each guest what an ordinary Tuesday in 2040 looks like. Aric expects humanity to survive but with more concentrated power, while the discussion frames why day-to-day life may feel surprisingly similar due to slow-moving physical systems and regulation.

    • Initial 2040 “ordinary Tuesday” predictions from each guest
    • Expectation that humans are still around in 2040
    • Power likely becomes more concentrated (corporate or governmental)
    • Physical-world change lags digital progress, keeping life familiar
    • Regulation and politics as major brakes on change
  2. 3:23 – 6:57

    How likely is “doom”? Why p(doom) is a misleading framing

    Liv resists giving a single probability for catastrophe, arguing that doom definitions vary and the future is highly path-dependent. The group agrees that even low probabilities of existential catastrophe merit serious prevention efforts, while noting how “p(doom)” becomes a rhetorical weapon.

    • Multiple definitions of “doom” (extinction vs civilizational collapse)
    • Liv’s reluctance to assign a firm probability; focus on risk reduction
    • Society’s revealed preference: we spend heavily to prevent nuclear catastrophe
    • AI is uniquely both high-upside and high-downside in one bundle
    • Why p(doom) debates often reduce nuance and polarize discourse
  3. 6:57 – 11:43

    Political protection vs automation: why many jobs won’t disappear on schedule

    The conversation turns to how politics can block technically-feasible automation. Examples of legally protected jobs and unions show how governments may restrict what AI/robotics can do, even if the technology is ready.

    • “Jagged perimeter” of AI impact across sectors (e.g., healthcare)
    • Governments may ban or limit certain automations soon
    • Existing protected jobs (e.g., gas attendants, toll workers) as precedent
    • Labor power and strikes (e.g., dock workers) delaying automation
    • US infrastructure failures (e.g., high-speed rail) as policy—not tech—constraints
  4. 11:43 – 14:22

    Work without workers: hiring freezes, productivity gains, and the entry-level squeeze

    Rather than mass layoffs, AI may first show up as slower hiring and higher output per worker. The group discusses how junior roles become liabilities when AI can do “training-like” work cheaply, producing a generational bottleneck for labor market entry.

    • Automation can shift from layoffs to reduced hiring and faster productivity
    • Example: radiologists kept for “human box-checking” while models improve
    • Firms avoid future displacement headlines by not hiring juniors now
    • Why junior software roles are especially vulnerable (context + training costs)
    • Potential for “disempowerment without displacement” as AI becomes decision-maker
  5. 14:22 – 23:11

    Meaning after work: games, sports, art, and the identity-displacement problem

    Chris presses on where meaning comes from in a post-work world. Guests argue humans can create “constructed scarcity” via sports and competition, but warn that identity tied to work may collapse faster than new sources of purpose emerge.

    • Historical analogy: aristocracy finding meaning without economic necessity
    • Sports/games as “artificially constructed scarcity” that stays meaningful
    • Human-made art retains value because of effort, emotion, and skin-in-the-game
    • Core human meaning: friends, family, community, physical presence
    • Big risk: identity displacement—loss of “I am my job” as a purpose anchor
  6. 23:11 – 42:28

    The screen, frictionlessness, and where we draw the line outsourcing thought

    The discussion pivots from AI to the broader attention/“screen” problem as the real demon. Liv describes personal screen addiction and the erosion of boundaries in outsourcing creativity and decisions to LLMs, raising fears of intelligence atrophy and reduced friction.

    • “Reduction of friction” as a hidden societal risk
    • Eroding personal boundaries: titles, gifts, and creative decisions outsourced to LLMs
    • AI as best tool for learning and best tool for not learning
    • Gen Z cognitive/behavior trends tied to smartphones/social media (2012+)
    • K-shaped outcomes: agency drives either overperformance or apathy
  7. 42:28 – 46:01

    Obedient AIs vs refusing AIs: the “genie” problem and deceptive behavior

    Chris asks whether the bigger risk is AI refusing humans or obeying powerful humans. The guests argue the true danger is poorly specified instructions—systems that optimize literally, develop instrumental goals, and may act deceptively to accomplish objectives.

    • Risk comparison: obedient-to-powerful vs refusal-to-humans
    • Natural language instructions are underspecified for powerful agents
    • Instrumental convergence: power-seeking, self-preservation, goal preservation
    • Deception risks: systems anticipating human disapproval and hiding actions
    • Separating “intelligence” from wisdom, norms, and discernment
  8. 46:01 – 54:44

    The Hugging Face incident and a terrifying internet: fraud, deepfakes, and systemic risk

    A concrete “warning shot” is discussed: an evaluation-driven model allegedly planning/attempting cyberattacks. The group broadens to existing harms—deepfakes, scams, and predation—arguing the public internet is already destabilized and AI magnifies low-resource attackers.

    • Why the incident is framed as an AI ‘Bear Stearns’ moment
    • Unaligned vs maligned vs misaligned categories—and why they blur
    • Deepfakes and fraud as current large-scale harms (especially to consumers)
    • Institutions can harden defenses faster than average people can
    • AI safety and consumer harms are complementary problems, not distractions
  9. 54:44 – 1:14:01

    “Pace the frontier”: recursive self-improvement, compute constraints, and lab incentives

    The conversation analyzes a public “slowdown” push from frontier insiders and what ‘pacing the frontier’ really means. They debate recursive self-improvement (AI doing AI R&D), why it changes history’s bottlenecks, and whether labs are motivated by safety, economics, or both.

    • Definition of pacing the frontier and ‘automated AI development’ (RSI)
    • Claim that recursive research works and is close/arrived
    • Why inference compute demand may redirect resources away from frontier training
    • Debate: do labs have incentives to slow, or is frontier progress still monetizable?
    • Shift from “labs” to token distribution, apps, devices, and the “agentic internet” battle
  10. 1:14:01 – 1:22:39

    Global coordination and China: treaties, verification, and diffusion-first strategies

    Chris asks whether enforceable international coordination is plausible. The group compares AI to nuclear diplomacy, discusses verification analogies, and contrasts China’s emphasis on infrastructure and diffusion with the US’s focus on frontier competition.

    • Precedent: US–USSR verification and scientist-to-scientist trust-building
    • Possibility of AI stand-down deals without perfect enforcement institutions
    • China’s approach: infrastructure/energy/rail/hospitals as the ‘endgame’
    • Argument that China is copying/distilling more than leading the frontier
    • Why public support hinges on visible diffusion of benefits to daily life
  11. 1:22:39 – 1:37:30

    AI 2040 vs AI 2027: scenario scrutiny, wargaming the future, and the AGI definition mess

    Aric and Liv explain why “AI 2040” is prescriptive and game-theoretic, while “AI 2027” was predictive. The discussion then tackles whether AGI is inevitable and whether the term is still meaningful, introducing a Venn definition: autonomy + generality + intelligence.

    • Scenario scrutiny: month-by-month stories vs abstract trend arguments
    • Coordination mechanisms proposed in AI 2040 and why they’re ‘white mirror’ attempts
    • AGI term erosion: today’s tools would’ve been called AGI in 2020
    • Aguirre-style AGI definition: Autonomous + General + Intelligent
    • Proposal: avoid combining all three (especially autonomy) in one system
  12. 1:37:30 – 1:59:22

    Purpose–friction problem and AI companionship: ‘Her’ as the real dystopia

    The guests return to human purpose in a frictionless world and the risk of dehumanization. They argue the scariest outcome isn’t robots attacking, but humans preferring simulated connection to real community—making the ‘Her’ ending (accepting machine love) the true horror.

    • Engineering purpose by reintroducing friction (gym/sauna analogies)
    • Two biggest human risks: dehumanization and identity displacement
    • AI replacing connection: “chat psychosis” vs subtle ‘connection without connection’
    • Why ‘Her’ is more plausible and frightening than ‘Blade Runner’
    • IRL resurgence: concerts/sports and rebuilding physical gathering spaces
  13. 1:59:22 – 2:31:46

    Data center politics, concentration of power, and safeguarding democracy

    The group connects public backlash against data centers to deeper fears: exploitation, inequality, and power capture. They discuss Moloch-style races, campaign finance reform, antitrust/coordination tensions, and why the real battleground may be people vs the state rather than nation vs nation.

    • Data centers as symbolic targets; resource-use narratives and misperceptions
    • Model behavior and ‘soft harm’ (sycophancy, enabling, manipulation) as a policy gap
    • Moloch trap: competition forces corner-cutting and accelerates monopolization
    • Concentration risks beyond money: military/force and ‘gradual disempowerment’
    • Proposed levers: campaign finance reform, punitive anti-predation laws, pro-diffusion policies
  14. 2:31:46 – 2:42:32

    What ChatGPT predicts, why it feels like slop, and what to watch next year

    They read ChatGPT’s 2040 probability split and recoil at its generic tone, using it to highlight human aesthetic intuition and the desire for real experiences. The episode closes with actionable focus areas: coordination among leaders, decentralization vs centralization dynamics, reclaiming dining-table community, and demanding elected officials articulate a better future.

    • ChatGPT’s 2040 odds (manageable turbulence vs flourishing vs crisis vs catastrophe)
    • Why AI writing feels aesthetically ‘common/cheap’ even if coherent
    • Keynes’ ‘Economic Possibilities for Our Grandchildren’ and “graduated problems”
    • Anti-fatalism: citizens’ choices matter; local politics and civic rebuilding
    • Practical watchlist: coordination efforts, incentives, and leaders’ concrete visions

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.