We now 1 can see that it is a matter of time before AI can do any work that humans can do on a computer.

It is also a matter of time before AI can do any work that humans can do.

There are three things I worry about currently.

I do hope to figure this out, at least for myself.

Financial security

In 10 years, what will we still pay humans to do?

I find it really hard to think of examples that will stand the test of time.

Here are some examples I could think of that might still be true in 10 years.

  • You still probably cannot sue model providers for the faults in arbitrary model outputs. There might still be humans who are responsible for understanding work done by superintelligent AI, and making decisions on how to allocate resources.
  • Airplanes already have autopilot. The pilot might just need to press a few buttons to fly a plane from one location to another. However, we still need two pilots to fly every plane, and we seem likely to continue requiring this.

Given that you might not earn an income in 10 years, there might be some things that you want to give up.

This may include buying a house 2, starting a family, or traveling the world.

Sense of purpose

You may already have something like one million dollars, which is probably enough to live comfortably.

It is possible that you can retire somewhere low-cost and play (AI-generated) video games for the rest of your life. Then, what is the purpose in life?

Do you really need to continue to create a positive impact on the world? 3

Existential risks

Money and a sense of purpose do not matter if humanity ends.

This is what the labs worry about. 4

The most conceivable way for this to happen is that someone engineers a virus that wipes out humanity 5.

Covid-19 killed 7 million people. If there were a virus more infectious than Covid-19, with a longer incubation period and a higher mortality rate, it could kill a lot more people.

Conclusion

The world will continue to spin.

While anything now feels possible, there are some things we know are impossible. We still live under the same laws of physics. 6

I do hope to figure this out, at least for myself.

Footnotes

  1. This is written shortly after AI solved some Millennium Problems. I have been confident that AI will improve, but I am still surprised that it has improved this quickly. ↩

  2. I am not sure if the world has priced in the potential drop in housing prices. ↩

  3. If you have done something that you are proud of, I imagine that you could be less worried about your legacy. ↩

  4. One thing I had to realize is that the anxiety I experience might not fully overlap with that of lab employees. I do strongly agree that someone should think really hard about this problem, similar to how we think about terrorism. However, I do not have the means to influence the trajectory, as I currently do not work for a lab. ↩

  5. There is a much better exposition on LessWrong ↩

  6. I tweeted this sentence almost two years ago. I was expressing hope that AI could be performing much better, given that humans, despite their very inefficient brains and chain of thought, have been accomplishing a lot more. It reads really differently now. The same laws of physics are what prevent AI from destroying humanity. ↩