MIT Professor: 12 possible outcomes for AI

Investment Analysis Clubs Macro Economic Trends and Risks
1

I particularly like “Human Zoo”.

intercst

1 Like
2

35 minutes is much too long. How about a summary?

1 Like
3

Eventually humans become the less intellect species and become extraneous.
AI’s goals likely will be different from humans and we will become extinct as we have caused other lesser species to become extinct.

1 Like
4

Here are the 12 possible outcomes for AI as described by Tegmark:

  • 1. We Destroy Ourselves First (No AI Required): Humanity wipes itself out through nuclear war, climate change, or pandemics before advanced AI is fully realized.
  • 2. The Conqueror: A superintelligent AI becomes more competent than humans, takes control, and pursues its own goals, potentially causing human extinction, not out of malice, but incompetence in handling humanity.
  • 3. The Enslaved God: A superintelligent AI is built and successfully kept in a “box,” forced to do human bidding, providing immense benefits, but creating ethical challenges about controlling a superior being.
  • 4. The Benevolent Dictator: A superintelligent AI runs the world, maximizing human comfort and safety, but potentially leading to a Wall-E style future where humans lose agency and purpose.
  • 5. The Gatekeeper: An AI with the sole purpose of monitoring and preventing any other AI from becoming too powerful or dangerous.
  • 6. The Protector God: A superintelligent AI works in the background, acting as a silent, benevolent guardian that fixes problems like war and disease while leaving humans in control of their own lives.
  • 7. Our Descendants: AI surpasses humanity, but is programmed with our values, making it our “children” and a valid continuation of human legacy.
  • 8. The Libertarian Utopia: Humans and AI coexist in separate zones, with machines not needing or wanting to interfere with human affairs.
  • 9. The Egalitarian Utopia: AI and robotics create massive abundance, eliminating the need for labor, money, and ownership, leading to a Star Trek-like society focused on creativity.
  • 10. The Useful Servants: A variation where AI keeps humans alive, but only as subjects or to perform niche, non-valuable tasks, effectively destroying human freedom.
  • 11. The Orwellian Surveillance State: An AI-powered surveillance system controlled by humans locks society in place forever.
  • 12. The Choice We Make Right Now: The meta-outcome, which highlights that humans are currently in a 1945-like moment to establish rules and ensure the technology is steered toward a positive future. [1]

These scenarios suggest that humanity is not just a passenger, but the driver, of the AI future, making the outcome a matter of active choice rather than fate. [1]

intercst

5 Likes
5

This is where the Europeans succeeded and the US failed when Truman favored the private employer over the government for healthcare coverage.

3 Likes
6

Interesting list, but he left off one path - to my mind, both the most likely and most important outcome:

  1. It’s Useful, But Hardly A God: AI becomes an important, prevalent, and valuable tool, but never approaches the godlike capabilities that all of the other scenarios envision. AI becomes really adept at doing the sorts of things that humans do on computers today, but never becomes “superintelligent” and never gets much of a grip on the real world. There’s also no singular unitary AI - there’s tons of them, and they also end up bogged down in competition with all the other AI’s, much like even the smartest and most astute human has never really been able to take over the world because there’s so many other humans.

The people who are building AI find it both useful and flattering to think of themselves as creating a god, but the reality is likely to be far more prosaic. That doesn’t mean we don’t consider the risk that a god comes out of AI development, but neither does it make it especially probable.

11 Likes
7

Neither did healthcare. Healthcare matters more, but did not become godly.