The people warning you that AI is too smart have it exactly backwards. The real danger right now is that AI is too stupid, and you have already handed it the keys.
Pedro Domingos is professor emeritus of computer science at the University of Washington, winner of two of the highest awards in AI and data science, and author of The Master Algorithm, which sold over 300,000 copies, was called essential reading by Bill Gates, and was spotted on Xi Jinping’s bookshelf. A decade later, while much of the AI world chases ever larger language models, Pedro argues that path will not deliver AGI. In this conversation he dismantles both the doomers and the boomers, explains why we should worry about who controls AI rather than AI itself, and makes the case that AI’s true purpose is not to amplify your individual intelligence but to transform our collective intelligence as a species.
What You’ll Discover:
🤖 The Real AI Danger Nobody Is Talking About
- Why incompetent AI making consequential decisions today is a bigger threat than any hypothetical superintelligence
- The homunculus fallacy: why even AI experts keep projecting human qualities onto machines that have none
- Pedro’s reframe: do not worry about AI manipulating you; worry about the people manipulating you via AI
⚖️ Why Regulating AI Is Like Regulating Mathematics
- The category error behind the thousand AI bills currently working through legislatures
- Why most AI regulation should simply be regulation of the underlying applications, invariant to whether a human or machine made the decision
- The lessons Congress learned about the internet in the 1990s, and why they need to apply them again now
🎯 The Real Bottleneck Founders Should Solve
- Why the people who win in each industry will be domain experts who pick up AI, not AI experts who pick up domains
- The three levels of AI literacy, and how deep a non-technical founder needs to go
- Why ‘copilot’ is a flawed analogy that leads founders to misunderstand what they are actually working with
📈 The Job Apocalypse That Never Arrives
- Google Translate was supposed to end translation as a profession; there are now more translators than ever
- How automation creates more demand once it makes the underlying work cheaper
- The question every founder should ask: not ‘will AI replace me’ but ‘who will use AI to replace me’
🧠 The Killer App: Collective Intelligence
- Why Pedro thinks the emergence of AI is on the scale of the emergence of homo sapiens itself
- How AI could finally fix the dinosaur problem inside large companies, where signals take weeks to travel from foot to brain
- The mechanism design challenge: how to build social systems that surface the wisdom of crowds rather than the madness
Key Insights:
“People worry that computers will get too smart and they’ll take over the world, but the problem is that they’re too stupid and they’ve already taken over the world.”
“The real danger for you doing your job is not that AI is going to replace you. It is that somebody is going to do your job better with AI.”
“AI is a new form of intelligence on earth. The best analogy we have is the emergence of homo sapiens out of the African savanna.”
About Pedro Domingos:
Pedro Domingos is professor emeritus of computer science at the University of Washington and a leading voice in machine learning. He is the author of The Master Algorithm, a global bestseller translated into more than a dozen languages, and the satirical novel 2040, which imagines an AI running for president of the United States. He has won the SIGKDD Innovation Award and the IJCAI John McCarthy Award, two of the highest honours in AI and data science.
🎯 Perfect for: AI founders who feel out of step with both the doom and the hype, builders trying to figure out how deep into the technical stack they really need to go, leaders thinking about AI integration inside large organisations, and anyone who wants a clear-eyed view of what AI actually is from someone who has spent his career building it.
⏰ Timestamps:
00:00 – Cold open
00:09 – Introduction: The man who saw it coming
01:27 – Why incompetent AI is a bigger danger than super-intelligent AI
03:35 – The homunculus fallacy: why we anthropomorphise machines
07:15 – Domesticated wolves and why machine learning will not turn evil
09:01 – Why humans are vastly better at manipulation than AI
12:23 – Regulating AI is like regulating mathematics
18:19 – The doomers and the boomers: both are getting it wrong
22:06 – Every time, people say this time is different. Every time, they are wrong.
27:32 – ATMs and bank tellers: the lesson nobody learns
28:24 – Man with machine versus man without
31:17 – How deep into AI does a non-technical founder need to go?
35:07 – Why ‘copilot’ is the wrong analogy for AI
38:01 – The killer app of AI: amplifying collective intelligence, not individual
44:47 – Responding to the attention crisis
52:52 – Why mature fields have small textbooks again
Subscribe for more conversations that matter, and let us know in the comments: Are you more worried about AI being too smart, or too stupid?
#PedroDomingos #MachineLearning #AGI #AIRegulation #FounderMindset #CollectiveIntelligence #MasterAlgorithm #ArtificialHappiness