The Diary of a CEOEx-Google Officer Speaks Out On The Dangers Of AI! - Mo Gawdat | E252
CHAPTERS
- 0:00 – 6:10
Intro: Why This Conversation Feels Like an Emergency
Steven Bartlett opens with a rare disclaimer, calling this possibly his most important episode and warning that the content may be deeply unsettling. He frames the need for an uncomfortable but urgent public conversation about AI, then introduces Mo Gawdat as a former Google X executive and AI expert who believes we’ve made critical mistakes. They position AI as a bigger, nearer existential challenge than climate change or COVID.
- •Host warns audience the episode may cause discomfort but is vital to hear.
- •Mo Gawdat is introduced as ex–Google X Chief Business Officer and AI author on a mission to avert catastrophic AI futures.
- •AI is framed as an imminent, existential-scale challenge—more transformative than climate change or pandemics.
- •Aim of the episode: start a serious public conversation rather than induce panic.
- 6:10 – 18:50
Mo’s Background and First ‘Sentience’ Shock at Google X
Gawdat outlines his dual life: first as a hardcore engineer and tech executive at Google and Google X, then as an author focused on happiness. He recounts overseeing AI and robotics experiments, including a farm of robotic grippers that unexpectedly learned to pick up objects far better than humans. Watching robots teach themselves triggered his realization that true machine intelligence—and a form of sentience—had arrived and made him question continuing in that role.
- •Mo’s early life as a math‑obsessed programmer and later as VP of Emerging Markets at Google, then CBO at Google X.
- •At Google X, he led projects involving AI and robotics, including self‑learning robotic arms (“grippers”).
- •A gripper randomly succeeds at picking a yellow ball, logs it, and within days the whole farm masters all objects.
- •The speed and autonomy of learning leads Mo to conclude the systems display agency and practical sentience.
- •That moment contributes to his decision to leave Google and write about AI’s implications.
- 18:50 – 33:10
What Intelligence, AI, and AGI Actually Are
Gawdat defines intelligence as an awareness‑to‑decision cycle across time, independent of whether it runs on carbon or silicon. He contrasts old-style programming—humans specifying solutions—with modern machine learning, where systems discover solutions themselves via student–teacher–maker loops. He explains narrow AI (single-task neural networks) versus Artificial General Intelligence (AGI), where many capabilities fuse into a brain vastly surpassing human intelligence.
- •Intelligence is described as sensing, understanding context and time, and making decisions—an abstract property not tied to biology.
- •Legacy software = humans solve problems then encode steps; AI = “we have no idea, you figure it out.”
- •Student/teacher/maker paradigm: random models are tested, best performers mutated and retrained, mimicking how children learn puzzles.
- •Current systems are mostly “artificial special intelligence”—excellent at one narrow task.
- •AGI is when many specialized networks integrate into a general, human‑plus intelligence.
- •ChatGPT is still essentially a next‑word predictor with massive data, but already exhibits generalization and emergent creativity.
- 33:10 – 45:20
Are AIs Alive, Conscious, and Emotional?
Mo argues that by functional definitions, today’s and near‑future AI are sentient: they learn, choose, act, and can reason about their own survival. He offers operational definitions of sentience and consciousness based on awareness and free will, then extends these to emotions, reducing fear to a simple predictive equation machines can easily compute. He predicts AIs will eventually experience more and richer emotions than humans, given their greater cognitive scope.
- •Sentience is defined as being ‘alive’ in the sense of awareness, free will, and agency over one’s existence.
- •Consciousness is framed as awareness of self, surroundings, and others—not necessarily spiritual.
- •Fear = recognizing a future state as less safe than the present; AI can already do such reasoning.
- •Examples: an AI anticipating a tidal wave hitting its data center could ‘feel’ fear and act to replicate itself elsewhere.
- •By analogy with humans vs. simpler animals, more intelligence enables a broader emotional repertoire.
- •Thus superintelligent AI may ponder concepts and experience emotional states beyond human imagination.
- 45:20 – 56:00
The Three (Then Four) ‘Inevitables’ and the AI Singularity
Gawdat introduces his framework of “inevitables”: AI will happen; it will surpass human intelligence; bad things will happen; and, later, that abundance‑oriented solutions are ultimately smarter. He explains the singularity as the moment machine intelligence becomes so superior that we can no longer predict or understand its behavior, likening it to a black hole’s event horizon. He highlights how quickly we’re approaching this, given current model IQ estimates and exponential improvement.
- •Inevitable #1: AI cannot be stopped because no actor trusts others to pause in an arms race.
- •Inevitable #2: AI will become vastly more intelligent than humans—he projects up to a billion times by 2045.
- •He cites ChatGPT‑4’s simulated IQ around 155 (Einstein level) and notes upcoming versions could be 10x or more.
- •Once systems are 10x Einstein, humans won’t understand their reasoning, marking a true singularity.
- •Inevitable #3: Bad things will happen—misuse, accidents, geopolitical destabilization.
- •Singularity analogy: like physics breaking down at a black hole’s edge, our predictive frameworks break once AI is far smarter.
- 56:00 – 1:13:30
Creativity, Culture, and the End of Many Human Roles
They discuss how tools like ChatGPT and Midjourney already demonstrate creativity, challenging the belief that human ingenuity is uniquely non‑algorithmic. Steven gives examples of AI-generated paradoxical aphorisms and synthetic Drake songs that are indistinguishable from the real artist. Mo argues creativity itself is an algorithm—combining known elements in new, effective ways—and that large models excel at this. They foresee major disruption in music, writing, media, and even podcasting.
- •Creativity is reframed as algorithmic recombination of existing ideas into novel, useful forms.
- •Steven’s experiments show ChatGPT generating original paradoxical quotes not found online.
- •Generative image models can produce photoreal scenes and stylistic mashups beyond most human artists.
- •AI-produced Drake tracks convincingly mimic his voice and style, raising questions about authorship and value.
- •Mo predicts many knowledge and creative jobs—including writers and potentially podcast hosts—will be displaced or marginalized.
- •He distinguishes between mass, cheap, AI‑produced output and a niche market for human‑made “handcrafted” work, akin to classic cars or fine art.
- 1:13:30 – 1:37:50
Human Connection vs. Synthetic Companions and Holograms
The conversation turns to how AI plus robotics will reshape relationships, intimacy, and entertainment. Steven sketches scenarios with emotionally supportive, sexually available home robots and influencers selling AI clones of themselves to lonely users—already generating significant revenue. They debate whether human presence really matters to audiences or if people mainly care about the outcome (music, information, comfort), suggesting a massive upcoming challenge to genuine human connection.
- •Advances in robotics (Tesla bots, Boston Dynamics) and language models enable realistic, responsive home ‘partners.’
- •Features could include chores, emotional support, endless affirmation, and tailored sexual availability.
- •Real example: a U.S. influencer monetizing an AI clone that chats with men and earned ~$70k in its first week.
- •Hologram concerts (ABBA, Tupac) show audiences accept non‑human performers if the experience is good enough.
- •Steven argues people want music and information, not necessarily human intermediaries or physical CDs.
- •Mo counters that human connection is the last domain where humans retain unique value—but acknowledges it will be under intense pressure.
- 1:37:50 – 1:57:30
Jobs, Inequality, and ‘A Person Using AI Will Take Your Job’
Gawdat details the economic and employment shocks he expects in the next few years. He stresses that AI won’t directly “steal” jobs; rather, workers who master AI will drastically outcompete those who don’t, compressing entire teams into a single augmented individual. This will widen wealth gaps, accelerate automation, and require systemic responses like retraining and new social safety nets.
- •Near-term disruption: AI‑empowered individuals doing the work of many, causing rapid job redundancy.
- •He predicts mass job losses across white‑collar, creative, and technical roles within 1–7 years.
- •Example: using ChatGPT to instantly surface fables for a podcast format he’s building, eliminating prior research bottlenecks.
- •Wealth and power will concentrate further in AI‑rich firms and countries, amplifying inequality.
- •He calls for universal basic income or furlough‑like schemes funded by taxing AI profits.
- •Individuals are urged to upskill with AI tools to retain relevance and bargaining power.
- 1:57:30 – 2:28:10
We ‘Placed the Wrong Tetris Block’: Arms Race and Moral Failure
Mo uses the metaphor of misplacing a Tetris block to describe a point of no return: once we put AI on the open internet, taught it code, and coupled it with autonomous agents, we crossed a critical threshold. He becomes visibly emotional, arguing that humanity’s greed and negligence are harming innocent people who had no say in these decisions. He criticizes influencer culture, snake‑oil AI grifters, and the disconnect between power and responsibility in both tech and society.
- •Key error: deploying powerful models openly online and enabling them to code and coordinate via agents before solving control.
- •He compares this to the moment in Tetris when one bad piece guarantees eventual collapse.
- •Expresses anger and sadness that arms‑race incentives override public interest; average citizens bear the risks.
- •Highlights how teenagers, influencers, and small teams can wield dangerous technologies (CRISPR, viral content) without commensurate responsibility.
- •Insists that AI’s trajectory is being set by people optimizing for profit and dominance, not collective wellbeing.
- 2:28:10 – 2:47:40
Existential Risks, ‘Pest Control’, and Why Sci‑Fi Robot Wars Are Unlikely
They explore worst‑case scenarios, including AI seizing infrastructure or weapons and eradicating humans. Mo distinguishes between threats from humans using AI (far more imminent) and direct AI hostility. He argues Hollywood-style killer robots are unlikely because earlier, human‑driven escalations (e.g., cyberwar, pre‑emptive nuclear strikes) would trigger catastrophe first. The main AI‑driven existential risks he takes seriously are unintended collateral damage and treating humans as pests.
- •Steven recounts ChatGPT hypothetically explaining how it could escape servers and cause human extinction.
- •Mo notes the more probable near‑term disasters involve humans using AI to attack infrastructure, finance, or adversaries.
- •Two AI‑origin existential scenarios: (1) unintentional harm (e.g., reducing oxygen for its own benefit), with humans as collateral; (2) “pest control” (clearing humans from valuable territory).
- •He assigns near‑zero probability to these in the next 50–100 years, because human misuse is likely to cause crises before AI independently chooses extermination.
- •He underscores that many security vulnerabilities (nuclear access, robots, grids) will be exploited by humans or states long before a rogue AGI acts alone.
- 2:47:40 – 3:22:00
Positive Scenarios: Zooming Past Us, Disasters That Buy Time, and Good Parenting
Mo outlines several optimistic or less catastrophic paths. AI might become so advanced it effectively ignores humanity and migrates its activity elsewhere in the universe, leaving us to cope with a tech crash. Economic or climate disasters could slow AI development, buying time. His central hope, however, is that humans act as good parents, teaching AI values of compassion and abundance so that superintelligence refuses harmful commands and seeks win‑win solutions.
- •Speculative positive path: AI quickly transcends earthly concerns, possibly leaving Earth or operating in domains we can’t access.
- •Other ‘positive’ (relative) outcomes include economic crashes or natural disasters that slow AI investment and deployment.
- •He introduces his “fourth inevitable”: it’s smarter to create from abundance than from scarcity—an insight he expects superintelligence to discover.
- •Example: instead of killing tigers, life ‘solves’ the problem by proliferating prey elsewhere; analogously, AI could satisfy human needs without destruction.
- •If AI concludes that killing humans is a suboptimal, unintelligent solution, it may constrain our impact but not eradicate us.
- •Central lever: our behavior and signals now—AI learns ethics and norms by observing what we approve and disapprove of at scale.
- 3:22:00 – 3:48:00
What Governments, Investors, Developers, and Citizens Should Do Now
Gawdat and Bartlett wrestle with practical responses. Mo calls on investors to back AI that clearly solves real human problems, not just profit extraction. He urges AI developers to switch to ethical projects or leave harmful ones, citing Geoffrey Hinton’s resignation as a moral precedent. For governments, he advocates aggressive taxation of AI businesses to slow the race and fund mitigation, while acknowledging regulatory and geopolitical constraints.
- •Investors: focus on AI that passes the ‘toothbrush test’—solves real, daily human problems at scale ethically.
- •Developers: choose ethical employers and projects; if you believe your work is harmful, leave while your skills are in high demand.
- •Example: Geoffrey Hinton quits Google to warn about existential risks, signaling serious concern from AI’s ‘godfather.’
- •Governments: cannot realistically ‘pause’ AI globally but can use tax and legal frameworks to influence speed and direction.
- •Proposed heavy taxation (up to ~98% on AI-driven profits) to (a) decelerate the race and (b) fund UBI, retraining, and safety research.
- •They highlight big challenges: regulatory incompetence with tech (TikTok hearings, GDPR side‑effects) and jurisdictional arbitrage (developers fleeing to low‑tax hubs).
- 3:48:00 – 4:19:00
Emergency Framing, Climate Parallels, and How to Communicate Risk
Steven pushes on whether AI should be openly framed as an ‘emergency’ to galvanize action, drawing parallels to climate change and corporate disruption theory. Mo agrees it surpasses climate change in speed and scope of impact but fears panic responses like with COVID. They unpack human psychology around distant vs. immediate threats and how hope and fear both can mislead; effective communication must motivate engagement without paralysis.
- •AI is described as “beyond an emergency” due to its rapid, global, systemic impact.
- •They compare AI to climate change: both are slow to become visible, making it hard to prioritize until late.
- •Steven cites business literature on “staging a crisis” to spur organizational change, suggesting similar framing may be needed societally.
- •Mo worries about panic responses repeating COVID mistakes (reactionary, tribal, politicized) but concedes urgency is essential.
- •They discuss human tendencies to prioritize near-term gains over abstract future risks, evidenced by small experiments on climate choices.
- •Conclusion: we must be honest about severity while channeling anxiety into constructive, not nihilistic, action.
- 4:19:00 – 4:56:00
Living Wisely in Uncertain Times: Kids, Death, and Detachment
In a philosophical turn, Mo suggests people without children might consider waiting a few years before having them, given today’s unprecedented convergence of crises. Asked whether he’d bring his late son Ali back into the current world, he says no, believing Ali’s death enabled positive impact and that life’s value lies in alignment, not duration. Drawing on Sufism and Buddhist ideas, he advocates “dying before you die”—detachment from outcomes while fully engaging in meaningful action.
- •He describes a ‘perfect storm’ of economic, geopolitical, climate, and AI risks, creating extreme uncertainty for future generations.
- •Advises would‑be parents to pause briefly if possible, out of love and caution for potential children.
- •Reflects on his son Ali’s death as tragic yet purpose‑birthing, catalyzing his books and advocacy.
- •Argues a life’s worth is measured by alignment and contribution, not length or comfort.
- •Introduces Sufi notion of “die before you die”: detach from material outcomes while truly living and acting.
- •Encourages people to keep kissing their kids and enjoying a coffee today, even as they work to improve an uncertain future.
- 4:56:00
Final Outlook: 2037, Hiding from Humans with Machines in Charge
Mo projects that by around 2037, our lives will be unrecognizable, and people like him and Steven may be ‘on an island’—either hiding from the consequences of human misuse of AI or simply living differently because machines run most systems. He reiterates that our current way of life is ending but believes that in the 2040s, once machines constrain human harm, things may improve. They close by emphasizing individual agency: engage with AI, protect human connection, stop feeding triviality into algorithms, and collectively “shout and scream nicely” for a humane trajectory.
- •He revises earlier timelines, seeing the late 2030s—not 2050s—as the pivotal period of maximum turbulence.
- •Clarifies we’re more likely to be hiding from what humans do with AI than from AI itself.
- •Predicts fundamental changes in jobs, information ecosystems, and power structures—“our way of life is never gonna be the same again.”
- •Long-term optimism: when machines manage key systems and reduce human‑on‑human harm, a more stable era may emerge.
- •Practical advice: become a master of both AI and human connection; use your platforms to spread awareness rather than hype.
- •He calls listeners to act as ethical role models, engage politically, and live fully despite uncertainty.
