Podcast Summary: The Last Invention — EP 3: Playing the Wrong Game

Date: October 9, 2025
Host: Gregory Warner, Longview
Main Guests/Voices: Liv Barry, Karen Howe, Yoshua Bengio, Geoffrey Hinton, Jasmine Sun, Keech Hagee, various commentators

Main Theme / Episode Overview

This episode explores the long, winding road from chess computers to today’s artificial intelligence revolution. It challenges the narrative that mastering games like chess marks true AI progress and details the contrarian visionaries who insisted that “real” intelligence—akin to that of a child, not a chess master—would have to follow entirely different approaches. The story travels from IBM’s Deep Blue to Google’s game-changing acquisition of DeepMind, revealing how failed expectations, ignored researchers, and obscure gaming competitions fuel today’s AI race and existential debates.

Key Discussion Points & Insights

The Drama of Man vs. Machine: Deep Blue vs. Kasparov

• Historic Chess Match (00:00–04:47):

  • The 1997 Deep Blue–Kasparov chess match grabbed global attention as a test of human versus machine intelligence.
  • The IBM supercomputer Deep Blue, capable of analyzing 200 million chess positions per second, defeated the world champion Kasparov in a dramatic, public match.
  • Despite fears, the match didn’t mark a true leap in AI intelligence. Chess and humans thrived afterward. “While some of us looked at this and shrugged or perhaps breathed a sigh of relief, there were those who looked at this moment and thought we were playing the wrong game.” (Gregory Warner, 03:31)

• Contrarian Takeaways:

  • Some saw Deep Blue’s victory as brute-force calculation, not thinking. “Deep Blue, even though it was a pinnacle of AI at the time, it did not seem intelligent that way.” (Karen Howe, 04:08)
  • “That Deep Blue beat Kasparov was nothing like how human chess players do it.” (Yoshua Bengio, 04:19)

Games as AI’s Laboratory (04:47–11:49)

• Games as AI Benchmarks:

  • AI’s milestones tracked progress with games: tic-tac-toe in the UK (“knots and crosses”), checkers, backgammon, chess, Jeopardy.
  • “Being able to solve a game is being able to understand a particular environment… as artificial intelligence gets better at game theory… they are getting better at navigating the world.” (Liv Barry, 05:43)

• The Limits of Game AI:

  • Watson’s Jeopardy win in 2011 was hyped as transformative, yet “the strategy was very good at winning a complicated game, but it failed to live up to the hype. It failed to lead to anything very useful outside of the world of that game.” (Narrator, 11:35)
  • Games were useful for testing but skills didn’t translate to real-world intelligence.

The Contrarians: Connectionists vs. Symbolists

• Why Didn’t Game-Winning AI Become Truly Intelligent? (12:16–19:57)

  • Symbolic AI dominated: “hand-coding” lots of rules and databases enabled victory, but these systems were brittle and narrow.
  • “Nothing much, because we knew that it was just brute force search, which is a classical AI technique that is very unlike human intelligence.” (Yoshua Bengio, 13:12)
  • Outsiders like Geoffrey Hinton and Yoshua Bengio believed the way to real intelligence was to model brains, not just code rules.

• Neural Networks: The Wild Idea

  • Hinton persisted despite ridicule: “For a long time this was regarded as crazy. And more recently it's turned out it works much better than symbolic AI.” (Geoffrey Hinton, 14:14)
  • “He was interested in [AI] from the perspective of if we successfully create intelligent systems and computers that will enable us to better understand our own intelligence.” (Keech Hagee, 15:50)
  • AI researchers faced technical barriers: limited computing power, data scarcity, and the “local optimum” problem which limited learning.

The Technical Breakthrough: Backpropagation and Big Data (21:19–31:08)

• The Local Optimum Problem:

  • Neural nets needed to learn from mistakes; early systems would get “stuck” and couldn’t retrace steps to improve. “It's like a mountain range where you get trapped on a peak and you can't get to the higher peaks because you have to go downhill to get to the higher peaks.” (Geoffrey Hinton, 22:05)
  • The invention of backpropagation allowed neural nets to correct errors and keep improving—if only the computational resources were sufficient.

• Why the Timing Was Right in 2012:

  • Explosion of “big data” from the Internet, and access to powerful GPUs developed for video gaming enabled neural nets to finally shine (28:11–29:47).
  • “An industry of chip makers emerged that made these GPUs. The most important one being this little company called Nvidia.” (Narrator, 29:11)

ImageNet: The Quiet Revolution (31:08–32:33)

• The Game That Mattered:

  • Hinton’s team entered (and won) the 2012 ImageNet competition—a massive contest to build AI that could accurately recognize objects in images, a child’s task but—until 2012—one machines couldn’t do well.
  • “It turns out that this is something that is really easy for humans to do, even children, but actually very difficult for machines.” (Narrator, 25:39)
  • Hinton’s model “blows everything else out of the water with how suddenly accurate it is.” (Unknown AI Expert, 30:39)
  • “Some of the best researchers in the field…immediately switched their opinion. They said, this is amazing. We're going to start doing that.” (Geoffrey Hinton, 31:08)

• Aftermath and Shift:

  • Major tech companies scrambled to acquire neural net talent. Hinton joined Google.
  • The shift from hand-coded expert systems to “learning” AI began to sweep all of technology: image search, translation, recommendations.

Neural Nets, the Mystery, and the Trade-off (32:33–35:12)

• Power vs. Transparency:

  • Neural networks were powerful but opaque.
  • “If you go with the model of the AI toddler over the model of the AI expert, a trade off you have to make is mystery, essentially an understanding of exactly how it works.” (Narrator, 33:55)
  • “We'd like to know, but the reality is that if you let go of that requirement, then you can get much more powerful systems.” (Yoshua Bengio, 33:55)

• AI as a Mirror for Ourselves:

  • “They weren't trying to make some like robot that would do stuff for you. They were trying to actually understand ourselves. And we don't know how the human mind works. So it's a mirror in some ways.” (Jasmine Sun, 34:37)

The Next Frontier: Demis Hassabis and DeepMind (39:17–50:38)

New Contrarians: The Gamer’s Dream of AGI

• Demis Hassabis—The New Prodigy (40:01–41:56):

  • Chess prodigy, Pentamind champion, computer scientist, and neuroscientist. Haunted by the question: Why can he, a human, switch between games so skillfully? (42:00–42:24)
  • Founds DeepMind with Shane Leg to pursue artificial general intelligence (AGI), but “almost no one in Silicon Valley thinks that AGI is going to be possible anytime in the near future.” (Narrator, 43:55)
  • First major investment from Peter Thiel, who supports the AGI dream.

• DeepMind’s Atari Breakthrough (47:13–49:40):

  • Inspired by Hinton’s success, DeepMind sets out to build an AI that learns to win at Atari games—without any preset rules, playing like a child with the joystick for the first time.
  • “It basically showed this AI agent teaching itself in real time how to play a vintage Atari game, and then became better at it than any human in the world.” (Jasmine Sun, 47:13)
  • “This could be anything. This could be climate data, this could be economics data, stock market data, anything that has temporal sequences of data, which is most things these days.” (Karen Howe, 49:22)

• The AI Gold Rush:

  • DeepMind’s demo triggers a tech industry race—Google quickly buys DeepMind. “They are putting their seal of approval, they're putting their money and their resources behind this previously wild pie in the sky idea of creating a true digital supermind.” (Narrator, 49:55)

Notable Quotes & Memorable Moments (with Timestamps)

• "It did not seem intelligent that way."
  — Karen Howe on Deep Blue (04:08)

• "That Deep Blue beat Kasparov was nothing like how human chess players do it."
  — Yoshua Bengio (04:19)

• "AI’s milestones tracked through games, but skills didn’t translate to real-world intelligence."
  — Liv Barry (05:43–06:29)

• "If you want to be honest, you can't be sure, but still you have this strong feeling. These emotions is what allowed us to go through the times when it was maybe difficult to work on these topics."
  — Yoshua Bengio on scientific conviction (18:04)

• "It's like a mountain range where you get trapped on a peak and you can't get to the higher peaks because you have to go downhill to get to the higher peaks."
  — Geoffrey Hinton on the local optimum problem (22:05)

• "Many of the things that we do effortlessly, like recognize objects or recognize the words when somebody's talking, are actually very difficult computational tasks that require huge amounts of computation."
  — Geoffrey Hinton (25:47)

• "Some of the best researchers…immediately switched their opinion. They said, this is amazing. We're going to start doing that."
  — Geoffrey Hinton on the immediate impact of his 2012 ImageNet win (31:08)

• "They weren't trying to make some like robot that would do stuff for you. They were trying to actually understand ourselves. And we don't know how the human mind works. So it's a mirror in some ways."
  — Jasmine Sun (34:37)

• "If you go with the model of the AI toddler over the model of the AI expert, a trade off you have to make is mystery…you will never fully understand how they know, what they know, how they work."
  — Narrator (33:55)

• "Through games, through these competitions, we inch a little closer to building a mind unlike any that has come before."
  — Paraphrased from the episode’s closing arc

Chronological Timeline / Important Segments

• 00:00–04:47: Deep Blue vs. Kasparov and the disappointment of “brute force” victory.
• 04:47–11:49: Games as milestones for AI, from tic-tac-toe to Jeopardy.
• 11:49–19:25: Enter the contrarians—neural networks seen as fringe, philosophical divisions in AI.
• 19:25–25:47: Early technical obstacles of neural networks, the backpropagation breakthrough.
• 25:47–31:08: The ImageNet game—Hinton’s team changes everything.
• 31:08–34:51: The tradeoff between transparency and performance; AI as a mirror for the mind.
• 39:17–50:38: The DeepMind story—Demis Hassabis, Atari games, the Google acquisition, and the AI race’s new phase.

Final Notes & Tone

The episode’s tone is thoughtful, dramatic, and often reflective, blending historical narrative with personal stories of scientific perseverance. There’s a clear sense of awe for those who “stuck with the crazy idea” and also a growing concern over the implications—technical, philosophical, and social—of AI that can generalize far beyond games.

For listeners or readers:
This episode is crucial for understanding why “playing the wrong game” for decades in AI set back the field—and how a small group of outsiders, radical new algorithms, and unforeseen jumps in hardware created today’s AI revolution. The lines between games and reality are blurrier than ever, and the stakes are rising.

Next time: The existential AI risk debate and the start of the AGI safety arms race.