Comment by dguest

Our current economic model around AI is going to teach us more about psychology than fundamental physics. I expect we'll become more manipulative but otherwise not a lot smarter.

Funny thing is, AI also provides good models for where this is going. Years ago I saw a CNN + RL agent that explored an old-school 2d maze rendered in 3d. They found it got stuck in fewer loops if they gave it a novelty-seeking loss function. But then they stuck a "TV" which showed random images in the maze. The agent just plunked down and watched TV, forever.

Healthy humans have countermeasures around these things, but breaking them down is a now a bullion dollar industry. With where this money is going, there's good reason to think the first unarguably transcendent AGI (if it ever emerges) will mostly transcend our ability to manipulate.