Spectrum of speculation

I have found it useful to group positions on artificial intelligence into five axes, each of which has a spectrum of perspectives. I find that understanding someone’s opinion about each axis helps reveal their hopes and fears about AI.

1. AI bad — AI good
2. AGI far — AGI close
3. Slow takeoff — fast takeoff
4. Decentralize — centralize
5. Anthropocentric — biocentric — theocentric

Net benefit

• AI is net bad for humanity
• AI is net good for humanity

Superintelligence

• Close — AGI could happen in the next few years
• Far — AGI will not happen in our lifetime, maybe never

Takeoff

• Slow — reaching AGI will be a slow iterative process, if ever
• Fast — AGI could begin self-improving and reach superintelligence in a matter of days, weeks, months

Centralization

• Centralized — AI should be tightly regulated, have strict controls
• Decentralized — AI should be accessible to all humans

Human-centricity

• Anthropocentric — AI should be in the service of humanity
• Biocentric — AI is part of nature and will be our successor
• Theocentric — AI is the creation of a new god