If we speculate about how we might get to AGI from what we see from narrow AI today, do we learn anything? On this show we mentioned the following:

• Google
• Self-driving cars
• Embodiment (robots)
• Whole brain emulation
• Digital assistants
• Brain-computer interface (augmentation path)

and then went deeper on these:

• AlphaGo
  • Input -> Output function
    • I believe Deep Mind calls this Deep Learning. This is neural networks adjusting the strength of various connections, similar to the way human brains adjust neural pathways over time.
  • Reinforcement learning
  • How might embodiment make a difference here?
• Many problem-specific narrow AIs that are rewarded by the marketplace

Conclusion: Tools are a dead end with respect to creating AGI. Future episodes will take on robots, digital assistants, etc.

See also:

• 0050: Paths to AGI #3: Personal Assistants

 

• 0047: Paths to AGI #2: Robots