Anthropic Compression Principle (ACP)

[ste...@ridgway.com]

Hi,

 

I have written a draft manuscript developing a theory I call the Anthropic Compression Principle (ACP), and I thought some people here might find it interesting.

 

The core idea is that observer-moments are fundamental, that their measure should be weighted by algorithmic simplicity / generative support, and that the lawful physical world we observe is best understood as the simplest generative compression of experience rather than as an unexplained base from which consciousness somehow later emerges.

 

The project sits somewhere between anthropic reasoning, algorithmic information theory, philosophy of mind, and foundations of physics. It draws on ideas related to Bostrom’s SSSA, Solomonoff induction, UDASSA, Tegmark-style mathematical ontology, and some of the implementation / continuity issues explored in Greg Egan, though it is not identical to any of those.

 

This is a substantial draft rather than a finished formal theory, and I would be very interested in critical feedback, especially on:

• whether the central formal picture is coherent
• whether counting detector complexity really helps with arbitrary implementation worries
• whether the treatment of observer-moments and measure avoids naive copy-counting
• whether the relation to existing views is stated fairly
• whether the theory seems genuinely scientific in schematic form, or still too underdeveloped

 

The manuscript is attached below. I would be very grateful for any serious comments.

 

Regards – Steven Ridgway

[Alastair]

Hi Steven

Some thoughts/problems on an initial look at your impressive book:

- Observer-moments as states of consciousness require an explanation - they are too complex just to be acceptable as 'brute facts' (especially if 'add-ons' are included).

- I can't believe that Nature had a sole proclivity towards directly creating these elements of consciousness, or indeed any such proclivity at all (IOW one shouldn't expect it to be inherently human-centered).

- I am not sure how communication between OMs (or their highest-weighted sequences) in different minds would work - aren't we in danger of solipsism here - would we be talking to a real other mind or just a zombie mind of the law-based product of our own current OM?

Also, do we have a 'consciousness/OM -> laws/compression -> evolved-brains -> physical-consciousness' loop here? (Back to the Hard Problem?).

But I will need to study the book more carefully - I've probably missed something.

Alastair

[ste...@ridgway.com]

Hi Alastair,

 

Thanks, this is very helpful. These are exactly the sorts of questions the view needs to face.

 

On your first point:

 

• “Observer-moments as states of consciousness require an explanation - they are too complex just to be acceptable as ‘brute facts’ (especially if ‘add-ons’ are included).”

 

I agree that this is a real pressure point. My response would be that every worldview has to stop somewhere. The standard alternative is to take the whole physical universe, together with its laws and initial conditions, as a brute fact. That is not obviously a cheaper stopping point than observer-moments. And even if one accepts the whole physical universe as brute, one still has no explanation at all of why any observer-moment should be experienced. One simply has laws, matter, and then an unexplained consciousness relation. So ACP is not claiming that observer-moments are simple in the everyday sense, only that they may be the least bad starting point. I do agree, though, that if too much is packed into them at the primitive level, the theory risks winning too cheaply.

 

On your second point:

 

• “I can't believe that Nature had a sole proclivity towards directly creating these elements of consciousness, or indeed any such proclivity at all (IOW one shouldn't expect it to be inherently human-centered).”

 

I agree that it should not be human-centred. “Observer-moment” is meant in a fully general sense, not a specifically human one. We begin from a human observer-moment only because that is the evidence we happen to have, not because the theory is supposed to privilege human minds.

 

On your third point:

 

• “I am not sure how communication between OMs (or their highest-weighted sequences) in different minds would work - aren't we in danger of solipsism here - would we be talking to a real other mind or just a zombie mind of the law-based product of our own current OM?”

 

I do not think ACP is committed to solipsism, though I agree this needs careful handling. The claim is not that only my current observer-moment is real. The claim is that my current observer-moment is the starting evidence from which explanation proceeds. Other minds are then part of the simplest lawful world-model explaining that experience, just as physics itself is inferred rather than primitively given. So, the aim is not to collapse reality into a private theatre, but to explain why experience presents itself as a lawful shared world containing other minds.

 

And on your last point:

 

• “Also, do we have a ‘consciousness/OM -> laws/compression -> evolved-brains -> physical-consciousness’ loop here? (Back to the Hard Problem?).”

 

This is probably the deepest issue. My intention is not a vicious loop, but a reversal of explanatory order. The standard picture says: laws first, then matter, then brains, then somehow consciousness. ACP says: observer-moments are fundamental, laws are inferred as the best compression of their structure, and brains are then understood as lawful structures within that inferred world that support further observer-moments. That does not remove every residual difficulty, but I think it avoids the usual hard-problem structure better than the standard matter-first view does.

 

In short, ACP certainly does not remove all brute fact. The hope is only that it gives a deeper and more unified stopping point than “these laws, these initial conditions, and somehow consciousness appears”.

 

Thanks again for your feedback.

 

Regards - Steven Ridgway

--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/everything-list/d8f9c42d-e5de-4a7c-a271-434ff34506dbn%40googlegroups.com.

[Alastair]

Hi Steven

Having had a further look at your book, I have to say that I am struggling to compatibilize two of ACP's basic features - those of fundamental observer-moments, and their law-based content. A thought experiment may serve to help illustrate this.  

In this thought experiment an 'Examiner'(E) is able to inspect any microscopic part of the brain (of 'Subject'(S)) that E chooses (say by using some form of reflected electromagnetic radiation from the relevant neuron(s) in S's brain), whilst S is asked to look at a blue object and ponder its 'blueness'. In this way E can in principle trace a causal physical pathway from the activation of neurons/cells in the retina (which can be considered as part of the brain for these purposes), through the various brain regions concerned with setting up cortical maps, recognizing the object etc, and further on to S's thoughts about its perceived colour.

Now as I understand it, all this falls under ACP's acceptance of physical laws for sequences of OMs for S. Of course this 'pathway' will involve the simultaneous activation and interaction of many different neurons as well as possibly relevant intra-neuron events, all of which E will want to at least sample - and there will also be other physical inputs from various forms of memory and elsewhere. But the overall point is that there is no separate causal role for consciousness or OMs at this microscopic level (unless one is willing to invoke non-physical causal processes for which there is an absence of significant evidence and also a violation of Ockham's Razor); their role is confined to that of being a higher level description only (though these descriptions can still be useful for many if not most purposes). It follows that no fundamental theory based on these physical laws need include consciousness or OMs in order to be complete at the observed microscopic level, though they (consciousness and OMs) should in principle still be derivable from this theory once they are fully defined. In short, consciousness-based OMs cannot be a more fundamental description of reality than a physical-law based one if they incorporate those laws as ACP seeks to do; hence the incompatibility.

One might argue that consciousness and OMs, as the basis for our thinking, have to be fundamental; but it seems that an AI is able to reason without needing either of these (though for the case of philosophical reasoning it would have to be able to work with them as abstract concepts/components in some of that reasoning).      

There are also the separate issues of identifying the particular stage in the growth of a baby, or in the evolution of mammals, at which consciousness should be ascribed; and also the apparent incongruity of the existence of the Big Bang being inferred by induction from an OM that includes the (indirect) experience of a red-shifted photon originating from near it (so one Big Bang for each OM???).

As far as I can see the key parts of all of the above would be unaffected by considerations of simplicity-weighted preference for certain OMs (assuming this was going to be a valid approach - if it is indeed valid we would be dealing with the resultant set of OMs that we find ourselves part of (and yet still be based on the kind of actual micro-physical law-based events that E saw in the thought experiment), and it is the seeming incompatibility that started this post that I am concerned about here, not the details of the selection and the compression aspects).

I am conscious that I could still be missing something, but this is where I currently am at.

Alastair