TAG's Shortform

post by TAG · 2020-08-13T09:30:22.058Z · LW · GW · 10 comments

10 comments

Comments sorted by top scores.

comment by TAG · 2020-08-13T09:30:22.526Z · LW(p) · GW(p)
  1. It’s impossible to reach that world from ours, it’s entirely causally disconnected, and
  1. That world “really exists

That's exactly what decoherent many worlds asserts!

Replies from: Dagon
comment by Dagon · 2020-08-13T17:02:26.933Z · LW(p) · GW(p)

"really exists" is insufficient to affect decisions, though. Two more assertions must be made

3. That world is a valid moral target, which I care about. (Debatable. Stipulated for this purpose.)

4. That world is sufficiently different from my causally-reachable light-cone that I should behave differently than I would if I only cared about "my" world.

#4 is the tricky one IMO. Maybe it is, but how could I know how it differs or what decisions I should make differently?

Replies from: TAG
comment by TAG · 2020-08-13T17:37:37.694Z · LW(p) · GW(p)

Many worlds strongly implies that you make all possible decisions, so it undermines the very basis of decision theory -- you have no freedom to perform one act while refraining from another.

Replies from: Viliam, mikkel-wilson
comment by Viliam · 2020-08-17T15:37:56.876Z · LW(p) · GW(p)

Saying "in many worlds you make all possible decisions" is technically true, but it is important to add that some actions happen with large probabilities, and some of them happen with microscopic probabilities. That doesn't undermine decision theory; you can still use it to do the right thing with 99.9999% probability.

(Unless you believe in a version of "many worlds" that says that everything happens with the same measure, which is not what physicists believe. If that was true, then quantum computers would be useless for any purpose other than generating perfectly random numbers.)

(And if your objection is that you can't make "decisions" when the atoms in your brain are following the laws of physics, then 1. this is unrelated to quantum or classical physics, and 2. yes you can, this is exactly how evolution designed us by selecting for configurations of atoms that were more likely than random to do the right thing.)

Replies from: TAG
comment by TAG · 2020-08-17T16:06:53.134Z · LW(p) · GW(p)

That doesn’t undermine decision theory; you can still use it to do the right thing with 99.9999% probability.

If it's possible to use decision theory in a deterministic universe, then MWI doesnt make things worse except by removing refraining. However, the role of decision theory in a deterministic universe is pretty unclear, since you can't freely decide to use it to make a better decision than the one you would have made anyway.

(And if your objection is that you can’t make “decisions” when the atoms in your brain are following the laws of physics, then 1. this is unrelated to quantum or classical physics,

Deterministic physics excluded free choice. Physics doesn't.

Replies from: max-kaye
comment by Max Kaye (max-kaye) · 2020-08-31T04:35:23.708Z · LW(p) · GW(p)

If it's possible to use decision theory in a deterministic universe, then MWI doesnt make things worse except by removing refraining. However, the role of decision theory in a deterministic universe is pretty unclear, since you can't freely decide to use it to make a better decision than the one you would have made anyway.

[...]

Deterministic physics excluded free choice. Physics doesn't.

MWI is deterministic over the multiverse, not per-universe.

comment by MikkW (mikkel-wilson) · 2020-08-14T03:19:06.587Z · LW(p) · GW(p)

I feel this misses the mark

The probability density of an agent (e.g. person) operating under a given decision theory (realistically, with a given set of internal causal gears) making a certain decision varies from potential decision to potenial decision. So, although every possible decision has non-zero probability density of happening, there are meaningful differences in frequency between various actions

Replies from: TAG
comment by TAG · 2020-08-14T10:05:51.631Z · LW(p) · GW(p)

But MW is deterministic. Under single universe determinism, you can't actually choose to perform an action you would not have performed. (Decision theory is either a step in the process or irrelevant). What MWI adds to that inability is a further inability to refrain. If MWI allowed freely willed agents to change the measures of their actions, then they could lower the measures of the less favoured ones...but it doesn't.

Replies from: mikkel-wilson
comment by MikkW (mikkel-wilson) · 2020-08-14T22:10:57.588Z · LW(p) · GW(p)

Probability is still a well-defined concept, even in a deterministic many worlds model

Replies from: TAG
comment by TAG · 2020-08-15T10:01:56.673Z · LW(p) · GW(p)

And you still can't refrain.