A Sketch of Answers for Physicalists

post by Chris_Leong · 2020-03-14T02:27:13.196Z · LW · GW · 15 comments

Contents

  Indexicality:
  Pre-Reduction References
  Epistemic Status of Physics
  Anthropics:
  Functionalism:
  Causality vs. Logic:
  Final Thoughts
None
15 comments

Epistemic Status: Just quickly sketched out some rough solutions

Jessicata recently wrote a post Puzzles for Physicalists [LW · GW]. I'm not a physicalist, but I'll attempt answer these questions from this perspective anyway.

Indexicality:

Jessicata argues that indexicality is fundamental (without defining the exact sense she means fundamental) because our attempts to define things objectively don't actually succeed. For example, the indexical "my phone" could be expanded to the objective-looking "Chris Leong's phone". However, this still actually needs to be defined relatively because there could be an exact clone of me somewhere in the universe.

I agree that the only way to refer to features of the world without encountering these non-uniqueness problems is relatively. So we never really encounter anything objectively. However, everyone already kind of knows the we can't definitely show the existence of any objective reality behind our observations and that we can only posit it. This isn't exactly news.

Pre-Reduction References

Jessicata argues that even though we often define water as H2O, if the chemical composition of the substance with the properties we observe from H20 had been XYZ, we would have defined water as that instead. She then argues that a) before we knew the science we still had an initial pre-reduction definition of water b) a philosophical account of science should contain these pre-reduction definitions so we can describe how they get attached to scientific definitions.

Why are these hard to define? Consider water. For simplicity we'll pretend that it's pre-reduction definition consists of "feels wet", "transparent" and "behaves like a liquid". Let's zoom in on "transparent" as if we could explain one, we will most likely be able to explain the others too. Just like water, "transparent" has a high-level and low-level definition and to fully understand it we need both. The high-level may contain thing like "see", "object" and "light" as primitives so that we can define an object as transparent if we can see another on the other side of it. The low-level definition describes the exact physics from the object to the light in your brain. The high-level is just a description of logical relations while the low-level is physical. So both will occur in a physicalist account.

Epistemic Status of Physics

The argument here is that in order to explain our justification of physics we need a concept of agents. Otherwise we we'll be able to talk about running or observing experiments. Fair enough, but as to whether observations imply agents depends on what is meant by "observation". If we imagine observations have a qualiatic element, then we can't model them with science. On the other hand, if they don't then we can apply the same double model trick. The high level relational model is something to do with informational processing; again we can define this as a logical relation. The low level contains a physical description of object observed, transmission of light to the observer, mechanism of the eyes and a description of the observers mind. This would explain the entanglement of the object and the observation.

Anthropics:

Argues that physics on its own doesn't have the language to say what an observer is. If qualia existed then we could just say an observer is someone who experiences qualia, but without this we need to find a definition of observer. I agree that just given a physical system on its own we can't define what counts as an observer. Observers are an information processing concept so in order to define them we need an interpretation scheme so we can understand how a physical system is processing information.

So given a physical system containing a human, how do we figure out which interpretation scheme to use? Well, humans need to interact with the world. We could hypothetically look at a bunch of humans, see how they interact with the world, scan their brains and then define an interpretation scheme of the information in their brains. Given such a scheme, we could then define information update operations that accurately match the real world as observation and observers as systems that are efficient at making observations.

Functionalism:

The argument is that functions like the hammer function are relative to observers. A hammer for humans isn't the same as one for an octopus. The argument is that similarly the functions of a mind are relative to a user and so what counts as planning, observation, ect. is relative. I'm not a fan of functionalism at all, but I already explained about how to define observations without making it relative to each human. It's not so hard to generalise other mental operations in the exact same way.

Causality vs. Logic:

I've broken down this conflict here [LW · GW]. We really shouldn't expect counterfactual "coulds" and logical "coulds" to be the same just because it's the same word.

Final Thoughts

It didn't actually turn out that challenging in the end to produce plausible solutions for all of these problems. I'm not definitely claiming that any of these solutions are correct, indeed they'd probably all need work since I just quickly sketched them out, but the point is that it isn't at all clear that any of these issues are really problematic.

15 comments

Comments sorted by top scores.

comment by justinpombrio · 2020-03-14T14:18:29.475Z · LW(p) · GW(p)

Would you and Jessicata mind clarifying what you mean by "physicalism"? Is it the same or different than Yudkowski's definition of "reductionism", for which he said:

Reductionism is not a positive belief, but rather, a disbelief that the higher levels of simplified multilevel models are out there in the territory.

For example, I'd like to separate:

Physics-y (i.e., low level) maps are always better than high-level maps.

from:

Physics-y (i.e., low level) maps always make predictions that are at least as accurate as high-level maps, given sufficient information and computation.

I'm suspicious that Jessicata may be attacking a straw version of physicalism while you're defending a steel version, but it's hard to tell. (And even if not, it's nice to know exactly what's under discussion.)

Replies from: Chris_Leong
comment by Chris_Leong · 2020-03-15T14:10:15.737Z · LW(p) · GW(p)

I'd define it as the argument that nothing non-material exists (except possibility logic)

comment by Pattern · 2020-03-19T22:40:39.308Z · LW(p) · GW(p)
The argument here is that in order to explain our justification of physics we need a concept of agents. Otherwise we [won't] be able to talk about running or observing experiments.

I think it might be possible to create an account of when an information processing system can interact with the world and develop an accurate understanding/model for it (even if the system is deterministic, etc.).

comment by jessicata (jessica.liu.taylor) · 2020-03-14T17:22:16.389Z · LW(p) · GW(p)

Some comments:

However, everyone already kind of knows the we can’t definitely show the existence of any objective reality behind our observations and that we can only posit it. This isn’t exactly news.

Un-referenceable objective reality goes rather beyond un-knowable objective reality. The second doesn't collapse into absurdity, while the first does (note that "un-referenceable objective reality" is a reference!).

The high-level may contain thing like “see”, “object” and “light” as primitives so that we can define an object as transparent if we can see another on the other side of it.

I'm not objecting to an account of physics that takes these things as primitives for definition. The way you're defining these is very much compatible with how I would, and incompatible with the sort of physicalism that thinks it isn't meaningful to talk about "seeing" (e.g. consciousness) independent of a physical definition.

We could hypothetically look at a bunch of humans, see how they interact with the world, scan their brains and then define an interpretation scheme of the information in their brains.

"Define an interpretation scheme" is incredibly vague. If it's a functionalist interpretation scheme then the functionalism section implies.

Replies from: TAG, Chris_Leong
comment by TAG · 2020-03-14T19:46:58.194Z · LW(p) · GW(p)

note that “un-referenceable objective reality” is a reference!).

Says who? It's a piece of ostensibly interpretable English. Does that make it a reference? .Fine then it's a reference and there is no paradox. Or does a reference have to latch on to some definite bunch of atoms, somessubset of physical reality? (Ie not be a "dangling pointer") Then it isn't a reference,and there is still no problem.

If you try to have it both ways, by insisting that any comprehensible English noun phrase must be a reference (even talk of unicorns or noumena),whilst simultaneously insisting that a reference must be defined narrowly as non dangling pointer that latches on to a physical object,then you have created a paradox. A paradox you didn't have to create.

This was solved over a century by Frege. There's Sense and there's Reference. Even if you grasp the sense,its up to the universe to decide whether an appropriate referent exists.

Replies from: Chris_Leong
comment by Chris_Leong · 2020-03-14T20:14:47.580Z · LW(p) · GW(p)

I can follow your argument, but could you clarify what you mean by Sense and Reference?

Replies from: TAG
comment by TAG · 2020-03-14T21:43:49.479Z · LW(p) · GW(p)

It's not something I invented,it's the foundation stone of the Analytical philosophy project.

https://en.m.wikipedia.org/wiki/Sense_and_reference

PS having taken a look at the wiki article,it's crap...or I have a different take on the subject.

comment by Chris_Leong · 2020-03-14T20:11:19.657Z · LW(p) · GW(p)
Un-referenceable objective reality goes rather beyond un-knowable objective reality. The second doesn't collapse into absurdity, while the first does (note that "un-referenceable objective reality" is a reference!).

We can construct a model were we (the external observers) can reference things that not observer in the model (the internal observers) can reference. Here's an analogy - we can't prove an unprovable theorem, but we might be able to prove a theorem unprovable.

Incompatible with the sort of physicalism that thinks it isn't meaningful to talk about "seeing" (e.g. consciousness) independent of a physical definition.

I'm not familiar with that strain of thought, but I can posit why some people might find that compelling

"Define an interpretation scheme" is incredibly vague

Yeah, as I said, this is just a sketch. There's a lot more that would need to be said it order to actually do this

Replies from: jessica.liu.taylor
comment by jessicata (jessica.liu.taylor) · 2020-03-15T05:12:20.283Z · LW(p) · GW(p)

We can construct a model were we (the external observers) can reference things that not observer in the model (the internal observers) can reference.

An account would have to be given of how we, as humans embedded in the universe, can speak as any kind of "external observer". I have not so far seen a coherent account of this. It seems that it would correspond to a view-from-nowhere.

I am not sure how the analogy to theorems applies; this seems to be a case of a given perspective forming an argument that it cannot know some fact it can still define, which doesn't have anything to do with an "outside" perspective external to the initial one.

Replies from: Chris_Leong
comment by Chris_Leong · 2020-03-15T14:19:46.020Z · LW(p) · GW(p)

Maybe I'll write a post on this sometime.

"An account would have to be given of how we, as humans embedded in the universe, can speak as any kind of "external observer"" - If we construct a model that doesn't contain us, then we are an external observer of that model. We can then be analogy posit the existence of an agent that exists in that relation to us.

Re the analogy: We can't have an entity that is both internally referenceable and internally unreferenceable. However we can have an external reference to an unreferenceable entity. Okay, maybe the analogy wasn't quite as direct as I was thinking.

Replies from: jessica.liu.taylor
comment by jessicata (jessica.liu.taylor) · 2020-03-15T18:17:41.060Z · LW(p) · GW(p)

As an "external observer", we still have a standpoint, e.g. getting visual input from some camera. (This is apparent in movies)

We can look at ourselves from the outside, e.g. from the perspective of a camera placed in a room. However, this is a particular outside (that can be placed relative to us). If the outside could not be placed relative to us, we could not know anything about how we would appear to that outside. It would be kind of like getting lost in a random parallel universe and seeing phenomena totally unrelated to ourselves, if this thought experiment is even conceivable (I'm not convinced it is).

Our ability to imagine data about us being received by some perspective, depends on placing that perspective relative to our own.

Replies from: Chris_Leong
comment by Chris_Leong · 2020-03-15T19:37:40.942Z · LW(p) · GW(p)

Maybe this will help. Consider the characters in a video game. We are an external observer as we can see what is happening in the game, but they can't see us. The point isn't that we can see ourselves from the outside, but that we can imagine what it would be like to be seen from the outside.

"Our ability to imagine data about us being received by some perspective, depends on placing that perspective relative to our own" - Yes, there are limits to what we can say about the outside perspective as we can't reference it directly. We can only discuss it by analogy.

Replies from: jessica.liu.taylor
comment by jessicata (jessica.liu.taylor) · 2020-03-15T20:25:51.431Z · LW(p) · GW(p)

That doesn't at all contradict what I said. The camera in a video game such as Super Mario Bros is placed at particular space-time coordinates, and for the viewer to see Mario, the camera has to be close to Mario, and in fact the game logic makes sure the camera does not drift from Mario's position. It isn't a view from nowhere, it's a view from somewhere pretty close to Mario's locality.

The camera Mario gets when (hypothetically) he imagines himself from outside is different than Luigi's if they are at significantly different locations (hence, multi player games such as Maro Kart have multiple screens even in a third-person view).

The outside perspective is outside but it is not observer-independent.

Replies from: Chris_Leong
comment by Chris_Leong · 2020-03-15T20:49:24.479Z · LW(p) · GW(p)

I guess I should have been more precise. Imagine a game where we can see all the information, but some characters inside only have access to limited info.

"The outside perspective is outside but it is not observer-independent"

Sure, but it's not subject to the world-internal observer effects

Replies from: jessica.liu.taylor
comment by jessicata (jessica.liu.taylor) · 2020-03-15T21:09:42.039Z · LW(p) · GW(p)

See this thread [LW(p) · GW(p)] on Game of Life.

The simulator's perspective is outside "our universe" but not outside the totality; there are multiple possible simulable universes, like there are different video games. Mario's (hypothetical) notion of "a view outside this world" refers to a view of the world of Super Mario Bros, and this differs depending on the video game character. Additionally, the video game players / simulators live in their own world, which is part of the totality.

Any given perspective can imagine zooming out by a given "distance" (in terms of space, time, simulation level, multiversal branch, perhaps others). This yields a sequence of views, each of which is dependent on the initial perspective. Perhaps the "view from nowhere" may be considered as the limit of this process. I am not convinced this limit may be coherently reified as a referenceable thing, however. In addition, such a view would be infinitely far from our own, and it would take an infinite time to zoom in from there to our actual here-and-now location.