𝕮𝖎𝖓𝖊𝖗𝖆 (Parody)

Public alt of @CineraVerinia

'we should be gods professionally' — @anderssansberg

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@teortaxesTex Okay, I didn't expect you to be blue.

Like this is surprising to me!

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

What does it mean for a decision theory to be true?

I don't know what "CDT is true" means.

I don't think decision theories are the type of thing that has an objective truth value.

[IntellectualDemon @wifiejwn38]

I also think we have to say these correlation-based decision theories like FDT are not necessarily true, maybe CDT is true and in that case it is purely irrational to select blue unless you value other lives more than your own at like a 200 million to 1 premium (the probability of you being the tiebreaking vote).

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@MonasukoUHC @NathanpmYoung No, it's just that like giving me an extra $10 is not very meaningful?

My regret for not getting the $10 if blue loses is tiny compared to the expected altruistic utility I get from everyone in the world getting $10?

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@S0hny @NathanpmYoung That's what I said!

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@Romy_Holland @prerat This doesn't work because humans have approximately logarithmic utilities of money and are frequently risk averse.

A guaranteed $500K is worth more to many people than a very high chance of $1M.

People would (IMO rationally) two box for $500K simply for the guaranteed payout.

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@kitten_beloved @arctotherium42 Newcomb's problem doesn't rely on a perfect predictor.

It's not meant to be a physically impossible thought experiment.

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@arctotherium42 There are versions where Omega leaves Box B empty if it predicts you'll randomise your actions.

But the monetary differential between Box A and a full Box B is so large that randomisation should only reduce your expected utility (assume Omega probability matches).

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@methodolog73944 @Authw8 But utility of money is logarithmic/humans are risk averse.

If you want one boxing to be the expected utility maximising choice for most humans, then payoffs need to grow superlinearly with reduced predictor accuracy.

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@Authw8 If you're reducing the AI's accuracy to 90%, I would not also reduce the payouts of one boxing by a factor of 10!

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@agrippa_dr @ptrschmdtnlsn @RokoMijic Hello.

So here's my reasoning.

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

One shot prisoners dilemma with standard payoffs.

The other player is a stranger who always cooperates with everyone.

The game is private and your choice here has no effects on public reputation, etc. (any higher order effects from your choice are internal).

You

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@IsaacKing314 @eshear @JPobserver I'm going to see if LLMs can generate better charitable interpretations.

But I don't have high expectations.

General relativity is just not the kind of thing that could be about Kolmogorov complexity.

[But I know zero re: GR, just that it's about gravity/spacetime curvature.]

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@IsaacKing314 @eshear @JPobserver

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@IsaacKing314 @ApriiSR @CineraVerinia @an_interstice No, but I had been convinced that my original idea was wrong which was satisfactory for me.

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@IsaacKing314 @ApriiSR @CineraVerinia @an_interstice Jumping one point maintains it's distance from the point you jumped over.

Then I thought of how you'd jump the other points to get a new square and we'd end up with the same side lengths and area.

But I was thinking of like reflection operations and similar.

I was just wrong.

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@IsaacKing314 Yeah, but that won't create a square.

I was thinking of ways to create new squares from the original.

But my thoughts were mistaken.

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@princess_worms @ApriiSR @CineraVerinia @IsaacKing314 @an_interstice You didn't produce a new square though, but I do think that my original guess was wrong.

https://t.co/yOReNJa7vv

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@ApriiSR @CineraVerinia @IsaacKing314 @an_interstice GPT 5-Thinking's response:

https://t.co/yOReNJa7vv

[𝕮𝖎𝖓𝖊𝖗𝖆 (Parody) @CineraVerinia_2]

@ApriiSR @CineraVerinia @IsaacKing314 @an_interstice And the ways I sequenced these moves to create new squares in my head had new points at same original distances from each other and hence same area.