If you controlled the first agentic AGI, what would you set as its first task(s)?

post by sweenesm · 2024-03-03T14:16:49.708Z · LW · GW · 2 comments

This is a question post.

Contents

  (If you work for a company that’s trying to develop AGI, I suggest you don’t publicly answer this question lest the media get ahold of it.)
None
  Answers
    1 Ustice
None
2 comments

(If you work for a company that’s trying to develop AGI, I suggest you don’t publicly answer this question lest the media get ahold of it.)

(Let’s assume you’ve “aligned” this AGI and done significant sandbox testing before you let it loose with its first task(s). If you’d like to change or add to these assumptions for your answer, please spell out how.)

Possible answers:

  1. Figure out how to perform a unilateral “pivotal act” to keep any other AGI from coming online, and then get my approval before doing it unless there’s no time to, then just do it
  2. Figure out how to get people/governments on board to perform a multilateral “pivotal act,” and then do it
  3. Do (2) first, then (1) if (2) isn’t possible
  4. Prepare for having to police the world against potential “bad” AGI’s as they come online - do this in a lawful way
  5. Prepare for having to police the world against potential “bad” AGI’s as they come online - do this by illegal means if necessary, including illegal mass surveillance
  6. Figure out how to align ASI’s
  7. Figure out how to get to ASI as quickly as possible
  8. Figure out how to stop aging
  9. Figure out how to save as many human life-years as possible (such as by curing malaria, COVID, cancer, etc.)
  10. Figure out fusion energy
  11. Room temperature superconductors, if you please
  12. Solve water scarcity
  13. Figure out and make humanoid robots so we don’t have to work anymore
  14. Figure out how to raise people’s ethics on a massive scale
  15. Figure out how to help on a massive scale with mental health
  16. Figure out how to connect our brains to computers for intelligence augmentation/mind uploading
  17. Figure out how to make us an interplanetary species
  18. Make the world robust against engineered pandemics
  19. Make the world robust against nuclear war
  20. Figure out how to put an end to factory farming as quickly as possible
  21. Other? _______________

 

If you think things should be done concurrently, your answer should be in the form of, for example: “(1) 90%, (6) 9%, (10) 1%.”

If you want things done sequentially and concurrently, an example answer would be: “(1) 100%, then (8) 100%, then (9) 50% and (21) 50% (Other: "help me win my favorite video game").”

You can also give answers such as “do (8) first unless it looks like it’ll take more than a year, then do (9) first until I say switch to something else.” I’d suggest, however, to not get too too crazy detailed/complicated with your answers - I’m not going to hold you to them!

 

There’s a somewhat similar question I found on Reddit to possibly give you some other ideas.

Answers

answer by Ustice · 2024-03-03T15:13:51.908Z · LW(p) · GW(p)

First? Swing low, see how it performs, especially with a long-term project. Something low-stakes. Maybe something like a populated immersive game world. See what comes from there. Is it stable? Is it sane? Does it keep to its original parameters? What are the costs of running the agent/system? Can it solve social alignment problems?

Heck, test out some theories for some of your other answers in there.

↑ comment by sweenesm · 2024-03-03T16:22:02.361Z · LW(p) · GW(p)

Thank you for the comment. I think all of what you said is reasonable. I see now that I probably should’ve been more precise in defining my assumptions, as I would put much of what you said under “…done significant sandbox testing before you let it loose.”

Replies from: Ustice
↑ comment by Ustice · 2024-03-21T20:44:20.835Z · LW(p) · GW(p)

I kind of think of this as more than sandbox testing. There is a big difference between how a system works in laboratory conditions, and how it works when encountering the real world. There are always things that we can't foresee.  As a software engineer, I have seen system that work perfectly fine in testing, but once you add a million users, then the wheels start to fall off.

I expect that AI agents will be similar. As a result, I think that it would be important to start small. Unintended consequences are the default. I would much rather have an AGI system try to solve small local problems before moving on to bigger ones that are harder to accomplish. Maybe find a way to address the affordable housing problem here. If it does well, then consider scaling up.

2 comments

Comments sorted by top scores.

comment by Dagon · 2024-03-03T15:53:09.280Z · LW(p) · GW(p)

This depends entirely on context and specifics.  How did I get such control (and what does "control" even mean, for something agentic)?  How do I know it's the first, and how far ahead of second is it?  What can this agent do that my human collaborators or employees can't?

In the sci-fi version, where it's super-powerful and able to plan and execute, but only has goals that I somehow verbalize, I think Eliezer's genie description (from https://www.lesswrong.com/posts/4ARaTpNX62uaL86j6/the-hidden-complexity-of-wishes) [LW · GW] fits: 'There are three kinds of genies:  Genies to whom you can safely say "I wish for you to do what I should wish for"; genies for which no wish is safe; and genies that aren't very powerful or intelligent'.  Which one is this?

A more interesting framing of a similar question is: for the people working to bring about agentic powerful AI, what goals are you trying to imbue into it's agency?

Replies from: sweenesm
↑ comment by sweenesm · 2024-03-03T16:40:24.750Z · LW(p) · GW(p)

Thanks for the comment. I agree that context and specifics are key. This is what I was trying to get at with “If you’d like to change or add to these assumptions for your answer, please spell out how.”

By “controlled,” I basically mean it does what I actually want it to do, filling in the unspecified blanks at least as well as a human would to follow as closely as it can to my true meaning/desire.

Thanks for your “more interesting framing” version. Part of the point of this post was to give AGI developers food for thought about what they might want to prioritize for their first AGI to do.