Home Lex Fridman Notes
Lex Fridman · 2020-06-22 · 4h 08m

Ben Goertzel: Artificial General Intelligence | Lex Fridman Podcast #103

AGI pioneer Ben Goertzel lays out his decentralized, compassion-first path to beneficial superintelligence and immortality.

Ben Goertzel: Artificial General Intelligence | Lex Fridman Podcast #103
The guest

Ben Goertzel — AI researcher who popularized the term AGI, founder of SingularityNET, designer of the OpenCog framework, and former chief scientist at Hanson Robotics (creators of the Sophia robot).

The gist

Ben Goertzel traces his lifelong obsession with artificial general intelligence from childhood Star Trek and science-fiction inspirations through his mathematics career to his current AGI projects. He explains the OpenCog cognitive architecture built on a weighted labeled hypergraph, why deep neural nets alone will never reach AGI, and how SingularityNET aims to create a decentralized, blockchain-based network of cooperating AIs. He discusses Sophia the robot as a 'walking skeleton' for embedding human values into AI, defends its theatrical illusion against critics like Yann LeCun, and applies AI to COVID and longevity research. He closes with a transhumanist vision of abolishing involuntary death and pain, naming joy, growth, and choice as life's core values.

Big reveals

  • Goertzel argues a benevolent AGI solves all other problems because it can rapidly become massively smarter than humans.
  • He recounts originating and 'launching the term AGI upon the world' while editing a book, without ever loving the term.
  • He flatly says GPT-3 understands nothing: 'he's a complete idiot but brilliant idiot.'
  • He claims current deep neural net architectures will never get anywhere near AGI.
  • He reveals Hanson Robotics deliberately keeps people in Sophia's illusion because telling them how it works doesn't change their belief.
  • He describes a teleoperated Philip K. Dick robot experiment where people still believed 'the spirit of Phil' was animating it.
  • He states death is simply bad and his goal is to abolish 'the plague of involuntary death.'
  • He envisions making multiple copies of himself, including one that fuses with superhuman AGI and one kept human.

Things worth remembering

  • His friend Hugo de Garis proposed SIPI: that super-intelligences shrink to subatomic scale, so quantum randomness may be their thoughts.
  • Goertzel named his child 'Qorxi' (Quantum Organized Reward eXpanding Intelligence), beating Elon Musk to a strange baby name.
  • Shane Legg, co-founder of DeepMind, got 'his head full of AGI' working at Goertzel's company Webmind in New Zealand.
  • New Zealand AI grant funding went toward predicting the menstrual cycles of cows rather than military uses.
  • Goertzel notes Vladimir Putin and corporate CEOs now invoke AGI, a term once reserved for 'crazed Mavericks.'
  • He holds 'extraordinary negative respect for Ayn Rand,' having kept long-lived fruit flies at an Ayn Rand elementary school.
  • He ran an experiment having Sophia's chatbot and Facebook's Transformer chatbot converse for weeks to generate training data.
  • He warns 23andMe licensed its pooled DNA database exclusively to GlaxoSmithKline.
  • Goertzel describes himself as a panpsychist who views even a light bulb or quark as having consciousness.
  • He ends with his own words: 'our language for describing emotions is very crude, that's what music is for.'