Friday, January 22, 2021

Do Conscious States Depend on History?

I’ve had a few additional thoughts further to my recent post on counting conscious states, particularly on the extent to which a given conscious state is history-dependent (i.e., depends on its history of prior conscious states) and whether a particular conscious state can be created de novo (i.e., from scratch, without the person experiencing that state having actually experienced previous conscious states).

Imagine that a person has actually experienced a particular series of conscious states (which of course depend, at least in part, on the stimuli sensed).  For the sake of simplicity, I’ll just assume that there’s a conscious state for each stimulus “frame,” and for ≈10 distinct frames/second, there are about 300 million stimulus frames per year.  I’m 43 now, and not sure whether we should start counting conscious frames from birth or sometime later, but let’s say that I’m just about to experience my ten billionth conscious state.  In my last post, I gave a (very, very) rough estimate for the minimum number of information bits necessary to specify such a state.  That number may be large – say, on the order of a trillion bits – but it’s not ridiculous and is less information capacity than many people have on their mobile phones.  Whatever that number happens to be – that is, the minimum number (B) of bits necessary to specify a particular conscious state – the point is this: By assumption, the instantiation of those B bits in the configuration necessary to create conscious state C1 will indeed create that state C1.  (For the following argument, it doesn’t matter whether the mere existence/instantiation of that particular configuration of bits is adequate, or whether that configuration must be executed on some general-purpose computer/machine.)

In other words, by assumption, some conscious state C1 is sufficiently encoded by some series of B bits that may look like: 0011010100110111110... (trillions of bits later)... 10001001111100011010.  There may be a lot of bits, but the idea is that if physicalism is true and the information content of any given volume is finite, then any particular conscious state must be encoded by some string of bits.  If this seems odd to you, it’s definitely the majority opinion among physicists and computer scientists who actually think about this kind of stuff.  For example, Scott Aaronson characterizes the situation this way:

“Look—I don’t know if any of you are like me, and have ever gotten depressed by reflecting that all of your life experiences, all your joys and sorrows and loves and losses, every itch and flick of your finger, could in principle be encoded by a huge but finite string of bits, and therefore by a single positive integer. (Really? No one else gets depressed about that?)”

For the record, I don’t get depressed about that because I don’t believe it’s true, although I’m still trying to formulate my reasoning for why.  OK, so let’s assume that I have in fact experienced ten billion conscious states.  The state I am currently experiencing is C10,000,000,000 (let’s call it CT), and a tenth of a second ago I experienced conscious state C9,999,999,999 (let’s call it CT-1), and a tenth of a second before that I experienced C9,999,999,998 (let’s call it CT-2), and so on back.  Again, by assumption, each of these states is encoded by a particular string of bits.  So here’s my question: is it possible to just recreate state CT  de novo without, in the process, also producing state CT-1?

Here’s another way of phrasing the question.  Is the person who is experiencing conscious state CT someone who actually experienced CT-1 (and CT-2 and so on back), or someone who just thinks/believes that he experienced CT-1?  Is there a way to produce someone in state CT without first producing someone in state CT-1?  I don’t think so; I think that state CT is history dependent and literally cannot be experienced unless and until preceding state CT-1 is experienced.  After all, if conscious states are indeed history independent, then the experience of CT is precisely the same no matter what precedes it, and that could lead to some odd situations.  Imagine this series of conscious experiences:

Series #1

C1000: sees alligator in the distance

C1008: gets chomped by alligator

C1045: puts tourniquet on chomped arm

C2000: eats own birthday cake

C3000: rides on small plane to experience skydiving

C3090: jumps out of airplane to experience thrilling freefall

C3114: pulls rip cord of parachute

C3205: lands safely on ground

 

If conscious states are history independent, then the person’s experience at C3205 is precisely the same even if the physical evolution of the world actually caused the following ordering of conscious states:

Series #2

C1045: puts tourniquet on chomped arm

C3000: rides on small plane to experience skydiving

C1008: gets chomped by alligator

C3090: jumps out of airplane to experience thrilling freefall

C2000: eats own birthday cake

C3114: pulls rip cord of parachute

C1000: sees alligator in the distance

C3205: lands safely on ground

 

I can’t see how the above series would make any sense, but more importantly I can’t see how, even if it did make sense, the experience of conscious state C3205 could possibly be the same in both cases.  If I’m right, it’s because conscious states are history dependent and state C3205 actually cannot be experienced immediately after C1000.

I’m not sure where I’m going with this.  If conscious states are history dependent (which is what I’ve suspected all along, as in this paper) then lots of interesting implications follow, such as that conscious states cannot be copied, consciousness is not algorithmic, etc.  (I believe I've already independently shown these implications in this paper.)  The above analysis certainly suggests history dependence but is not a proof.  Maybe the way to prove it is by first assuming that conscious states are independent of history – in which case conscious state C3205, for example, can be created de novo without first creating conscious state C3204 (which can be created de novo with first creating conscious state C3203, etc.) – and then see whether that assumption conflicts with observations and facts about the world.

Remember that, by assumption, state C3205 is just instantiated by a (very long but finite) string of bits, say 0011010100110111110...  So imagine if we start with a long series of on-off switches all initially switched off.  We turn some of them on until eventually we have instantiated the correct series (0011010100110111110...), which encodes state C3205.  But it does not (and cannot) matter the order in which we flip those switches.  I have to think more about the mathematics, but I suspect that in guaranteeing that C3205 is independent of history, so that it and every preceding conscious state can be instantiated independently of its own history, we will end up needing far, far more bits than my original estimate of N^T.  I suspect that even the most conservative estimate will show that if conscious states are history independent, then consciousness will require far more information storage than is currently believed to reside in the brain. 

Then again, I really don’t know.  This is still just the initial seed of a thought. 

No comments:

Post a Comment

All comments are moderated. After submitting your comment, please give me 24 hours to approve. Thanks!