Matthew Gray ⏸️

Vaniver lots of places; I mostly post to LessWrong.

[Matthew Gray ⏸️ @Vaniver]

@tracewoodgrains Surely you know artists who would be able to draw you a quokka using a machine gun

(Turnaround time is slower than the robot, and it'll be more expensive, but it's not like it's not available)

[Matthew Gray ⏸️ @Vaniver]

@EpistemicHope @QiaochuYuan I thought I got it from EY

[Matthew Gray ⏸️ @Vaniver]

@goblinodds party-list proportional: https://t.co/nQOTn87Afs

elect Congress nationally

[Matthew Gray ⏸️ @Vaniver]

@So8res @bryan_caplan I think this does actually work for things like mortgages (where the bank has a very different time value of money than you do), especially if you structure it as an interest-only loan where you sell the house at the end.

[Matthew Gray ⏸️ @Vaniver]

@SarahTheHaider Most likely reason we're still around in 10 years, IMO, is that AI doesn't continue advancing, for one reason or another.

[Matthew Gray ⏸️ @Vaniver]

@SarahTheHaider 3) Returns to intelligence much lower than anticipated, such that we get continued moderate growth (and so a decade in we still don't have AI scientists / generals / politicians etc. that have pulled a Napoleon on human civilization)

[Matthew Gray ⏸️ @Vaniver]

@atrembath Nope

[Matthew Gray ⏸️ @Vaniver]

@benlandautaylor (Giving those parts extra power, of course, doesn't make sense.)

[Matthew Gray ⏸️ @Vaniver]

@benlandautaylor Hmmm. I don't think the "But that's probably" part is true b/c people developed Harberger taxes I think specifically as a structural solution to corruption.

Also, I think looking for the parts of society that are the most corrupt to figure out what to reform makes sense!

[Matthew Gray ⏸️ @Vaniver]

@ZyMazza Short answer yes. Long answer: there are lots of pieces of "AI doom" which have _different_ "ok that was a false alarm" points, and so a complete answer is long and made of lots of partial invalidations.

[Matthew Gray ⏸️ @Vaniver]

@slatestarcodex @robbensinger I think it's quite possible that OpenAI without its alignment team would have been much less successful; for example I think they might not have gotten a commercially successful chatbot without them (or would have much later).

[Matthew Gray ⏸️ @Vaniver]

@segyges @allTheYud @MatriceJacobine @Simon248 Nope, my argument doesn't rely on this! Try rereading this with the version where all advancements are vectors in the positive quadrant.

[Matthew Gray ⏸️ @Vaniver]

@segyges @allTheYud @MatriceJacobine @Simon248 Actually, sorry, that's the wrong question. Do you see why, if I believe in the 'two progress bars' model, the claim that every alignment advance is a capabilities advance _isn't_ evidence against the two progress bars model?

[Matthew Gray ⏸️ @Vaniver]

@segyges @allTheYud @MatriceJacobine @Simon248 Like, do you see why the argument that "every alignment advance is a capabilities advance" doesn't constrain how much it's a capabilities advance, and that the ratio might matter?

[Matthew Gray ⏸️ @Vaniver]

@segyges @allTheYud @MatriceJacobine @Simon248 That is, yes, aligning models generally makes them more capable. But does making the model more capable align it? Can you find things that do both, instead of just making the models more capable?

That's alignment research. I don't think it's impossible in principle.

[Matthew Gray ⏸️ @Vaniver]

@segyges @allTheYud @MatriceJacobine @Simon248 in the first tweet I brought up "differential advancement." I don't think I've seen that in any of your tweets that followed; I think "differential advancement" works pretty well as a reply to all of them. That's the thing I was hoping Opus might be able to walk you thru.

[Matthew Gray ⏸️ @Vaniver]

@segyges @allTheYud @MatriceJacobine @Simon248 ok; maybe you should see if Claude can explain to you my points? I apparently didn't do a very good job of it.

[Matthew Gray ⏸️ @Vaniver]

@segyges @allTheYud @MatriceJacobine @Simon248 But this can be basically independent of the capabilities of those models, or the domains that they operate in, or so on. (Of course it's easier to write utility functions for smaller worlds, but that's distinct from the tech breakthrus required to be able to make that agent.)

[Matthew Gray ⏸️ @Vaniver]

@segyges @allTheYud @MatriceJacobine @Simon248 Hmm, I don't think we mean the same thing by 'corrigibility'. An agent which has a clearly identifiable utility function / reward system, and also somehow is not incentivized to prevent a particular source from modifying it, is more corrigible than an opaque policy model.