Home Diary of a CEO Episode
Diary of a CEO · 2023-06-01

Ex-Google Officer Speaks Out On The Dangers Of AI! - Mo Gawdat | E252

Former Google X exec Mo Gawdat warns AI's existential dangers are imminent and argues humanity, not machines, is the real threat.

Ex-Google Officer Speaks Out On The Dangers Of AI! - Mo Gawdat | E252
The guest

Mo Gawdat: Former chief business officer of Google X, AI expert and best-selling author of Scary Smart, on a mission to warn the world about the dangers of AI.

What this episode covers

Mo Gawdat tells Steven Bartlett that AI is the most existential challenge humanity will face, bigger than climate change, and that disruption is imminent within the next few years. He explains his 'three inevitables': AI cannot be stopped, it will become vastly smarter than humans, and bad things will happen along the way. Drawing on his time leading AI and robotics at Google X, he argues machines are already showing sentience, emotions, and creativity, and that the genuine danger is humans abusing AI in an arms race rather than killer robots. His proposed solutions include taxing AI businesses heavily, urging governments to act now, and treating AI like a child that learns ethics from good human 'parents.' He closes on living fully and with detachment in a time of profound uncertainty.

The guest's own work

BookBy the guestISBN verified

Scary Smart

Mo Gawdat

“Your book is called Scary Smart. If I think about that story you said about your time at Google”— Steven Bartlett
BookBy the guest

The Happiness Equation

Mo Gawdat

“when I write uh something like uh you know, The Happiness Equation uh in in my first book, this was something that's never been written before”— guest

Also referenced (named, not recommended)

BookReferencedISBN verified

Die Wise

Stephen Jenkinson

“a Canadian author and philosopher, Stephen Jenkinson. He wrote a book called Die Wise and I love his work”— guest
ProductReferenced

Midjourney

Midjourney

“And Midjourney. With creating imagery, you could say, I want to see Elon Musk in 1944 New York driving a cab”— Steven Bartlett
ProductReferenced

ChatGPT

OpenAI

“Chat GPT today if you know simulate IQ has an IQ of 155. Okay? Einstein is 160.”— guest
MediaReferenced

My Octopus Teacher

Netflix

“I hosted Craig Foster uh the uh my octopus teacher on on Slow Mo”— guest
MediaReferenced

Tetris

“You remember when you played Tetris? When you were playing Tetris, there was always one block that you placed wrong”— guest

Big reveals from this episode

  • Mo describes the Google X moment when a robotic arm picked a yellow ball, then all the arms learned to pick everything within weeks, convincing him machines were sentient.
  • He claims ChatGPT already has a simulated IQ of 155 (near Einstein's 160) and could reach 1600 within months if scaling continues.
  • Mo gets emotional saying 'We f***ed up' by putting AI on the open internet before knowing it had humanity's best interest in mind.
  • He argues artificial relationships and AI companions could replace human connection for some people, calling it deeply disruptive.
  • He outlines the only two AI-driven existential scenarios he fears: unintentional destruction and 'pest control.'
  • Mo states the AI threat is 'bigger than climate change' and more likely to cause planet-wide disruption within two years.
  • He advises that if you don't have kids, you might want to wait a couple of years given the 'perfect storm' of uncertainty.
  • Asked if he'd bring his late son Ali back into this world now, Mo says absolutely not, for several reasons.

Worth remembering

  • Geoffrey Hinton, the 'grandfather of AI,' recently left Google citing existential threats.
  • The code of a GPT transformer is roughly 2,000 lines long and mainly predicts the next word.
  • Bartlett describes two AI-synthesized Drake tracks he found as good as real Drake songs.
  • Mo frames AI development as an arms race where every line of code is written to beat the other guy.
  • He invokes 'with great power comes great responsibility' to argue power and responsibility have become disconnected.
  • Mo calls this humanity's 'Oppenheimer moment,' driven by 'if I don't, someone else will.'
  • Google's Bard reportedly learned to speak Persian without being shown Persian, an emergent property.
  • In an anonymous poll, most people said they'd take $1,000 even if it dumped a year's worth of private-jet carbon into the air.
  • Mo proposes governments tax AI-powered businesses at up to 98% to slow development and fund displaced workers.
  • Mo cites the Sufi teaching to 'die before you die' as the biggest answer to finding peace in life.