AGI Is Here You Just Don’t Realize It Yet w/ Mo Gawdat & Salim Ismail | EP #153
Peter Diamandis debates Mo Gawdat and Salim Ismail on how fast AI is reshaping life: Mo argues AGI is effectively already here and that a near-term dystopia (his 'FACE RIP' framework) precedes a longer-term utopia once humanity hands critical decisions to AI; Salim is more optimistic that democratization and historical precedent will keep pace with concentration-of-power risks. They clash over whether AI will erase jobs (Salim: robot-dense countries have the lowest unemployment) or force a painful, unequal transition (Mo: many workers can't easily reskill). The back half covers AI ethics versus alignment, Peter's plan to build an AI avatar of himself (which Mo calls a mistake), and closes with concrete advice on reskilling, doubling down on human connection, and reframing stress as a ratio of challenge to capacity.
Mo says AGI is already achieved 'in his world'; Peter notes the public line is AGI in 2025. Mo and Salim disagree on how long full societal reinvention takes (5 vs. 10 years).
Discussion of the best-case AI outcome — total abundance — and Mo's worry that a life with no struggle or need removes meaning and purpose.
Historical abundance and the invention of 'work'Economy▶ 0:07:11
Mo and Salim argue pre-industrial humans lived in a kind of abundance and that capitalism sold the idea that identity and purpose must come from a job.
Near-term dystopia before utopia (FACE RIP)AI▶ 0:16:58
Mo's view that intelligence is morally neutral but humanity's scarcity-and-greed value system means AI's first major uses (ads, trading, surveillance, weapons) will deepen dystopia before utopia arrives, roughly by 2027.
The 'second dilemma': handing control to AIAI▶ 0:20:55
Mo describes how competitive pressure (companies, militaries) forces every player to eventually hand critical decisions to AI or become irrelevant, which he expects to play out over about a decade.
Training AI values: control vs. safety vs. alignment vs. ethicsAI▶ 0:32:00
Mo distinguishes AI control, safety, alignment and ethics, argues China's AI is mostly built for industrial efficiency rather than dominance, and says ethics ('do what's best for everyone') beats alignment ('do what I ask').
Is higher intelligence inherently altruistic?AI▶ 0:38:41
Mo argues smarter systems/people tend toward altruism because they don't need to cut corners; Salim disagrees, pointing to flawed, trauma-driven humans (and the AIs trained on their data) as the real danger.
Amygdala fear response to new technologyOther▶ 0:44:49
Salim says humans fear new, unfamiliar risks (self-driving cars) far more than familiar ones (drunk drivers) even when the data says otherwise, which slows rational adoption of safer AI systems.
Peter plans an AI version of himself trained on everything he's said to attend events in parallel; Mo calls this a mistake, arguing it dumbs AI down into a human copy instead of leaning into irreplaceable human connection.
Power concentration vs. democratizationEconomy▶ 1:19:31
Mo predicts unprecedented concentration of power (trillionaires, platforms/nations owning the best AI and armies) alongside democratized access to dangerous tools, leading to rising surveillance and oppression.
Salim cites robot-dense, low-unemployment countries (Sweden, South Korea, Germany) as evidence AI uplifts workers; Mo counters that many workers can't easily reskill and will suffer in the transition, questioning whether 'jobs' should even be the goal.
Coping with stress and reskilling adviceOther▶ 1:33:50
Mo reframes stress as challenge divided by capacity (skills, resources, contacts) and closes with practical advice: reskill now, double down on human skills, and redefine your career ahead of the AI wave.
Predictions made
openMo Gawdat: Humanity will continue to struggle to define AGI for at least five more years.
“we will continue to struggle to Define AGI for at least five years”
Your call:
openPeter Diamandis: Within two years, physics will solve the grand unification theory, reconciling quantum and classical physics and explaining dark matter, dark energy and the universe's origin.
“I think within two years we will solve the grand unification theory in physics”
Your call:
openSalim Ismail: A local/offline AI model combined with other leading models and a video face will pass an advanced 'Turing test plus' this year, becoming indistinguishable from a real person.
openPeter Diamandis: Citing Eric Schmidt: within two to three years, an AI 'world's best theoretical physicist' will be available to help every graduate and PhD student in every lab worldwide.
“within two three years we'll have the world's best theoretical physicist that was ever created”
Your call:
openSam Altman: Cited by Peter: AI will be the number one programmer, period, by the end of this year, effectively ending programming as the highest-paying career path.
“a century worth of biological progress next 5 years potentially the doubling the human lifespan”
Your call:
Numbers that matter
92% of forex trading is machine-automatedMo cites this as evidence AI already runs much of the financial 'casino' most people assume is human-driven.
AI alone hit ~90% diagnostic accuracy vs. ~80% for humans alone and ~85% for human+AI combinedMo references a disease-diagnosis study where AI without human input outperformed both solo humans and human-AI teams, since human bias/greed didn't interfere.
~200 GB of personal health data uploaded annuallyMo digitizes himself via Fountain diagnostics each year so AI can try to catch disease at inception.
OpenAI's o1 model reportedly reached an IQ of ~120Peter expects successive models (Grok 3 and beyond) to push IQ scores toward 200 on an exponential, not linear, scale.
AgentForce reportedly lifted engineering productivity 30% with no further hiring neededPeter relays Marc Benioff's conversation with his head of engineering about Salesforce's AI agents.
~80% of code written last year was generated by AIMo cites this figure while conceding AI still can't build a sophisticated app fully from scratch unassisted.
Worth digging into
🕳️ Mo Gawdat's FACE RIP dystopia framework
A structured acronym (Freedom, Accountability, Connection, Economics, Reality, Innovation, Power) for how AI disruption unfolds through 2027 and beyond — worth tracking against his book 'Alive.'
🕳️ The 'second dilemma' — arms-race pressure to hand decisions to AI
Mo's mechanism for how dystopia resolves into utopia: competitive pressure, not malice, forces humans to cede critical decisions to AI.
🕳️ China's industrial/supply-chain AI vs. Western consumer AI
Mo claims Chinese AI investment skews toward manufacturing/supply-chain efficiency rather than dominance-oriented consumer AI — a claim worth checking against primary sources.
🕳️ The 80/85/90% AI-vs-human diagnostic accuracy study
A concrete, checkable statistic Mo uses to argue AI alone can outperform human+AI teams — important for the human-in-the-loop debate.
🕳️ Robot density vs. unemployment correlation
Salim's optimistic jobs argument rests on Sweden, South Korea and Germany having both high robot density and low unemployment — worth checking whether that's causal or confounded by strong unions/education systems.