1

There are already several projects that are trying to simulate the human brain (see, for example, Algorithm for large-scale brain simulations)

Let's suppose that it is possible to simulate with an algorithm the brain of a real existing human.

We can ask a group of people to run this algorithm simply exchanging messages written on paper. The simulation will be extremely slow. We can suppose that it will take thousands of years to simulate what a human brain thinks in a few minutes. How much it can really take does not matter.

My hypothesis is that this simulation is not conscious. It perfectly simulates the brain of the real existing human, but it is not conscious.

While this "brain" is running, we ask to it this simple question: are you conscious?

What do you think it will reply?

If it replies that it is conscious, it has to be delusional.

Why would it be delusional? When we think that we are conscious, it is our brain that is thinking it. If we think that we are not delusional because we have good reasons, it means that our brains are getting some real information that the simulated brain doesn't get. Why then would it still think that it is conscious? And what is it this additional information?

Of course, it will be even more interesting if it replies that it is not conscious!

Other possible answers can be that the algorithm is conscious or that it is not possible to simulate the brain.

Both these hypothesis do not seem right.

Improve this question
4
  • The speed you are running it, or whether you are running it by hand-simulating every step or on a computer, shouldn't matter to its answer, it's the same algorithm either way. So whatever source of intuition is for thinking the very slow hand-simulated brain isn't conscious, I don't think there's a good basis for it to be any different than your intuition about a simulated mind running at an advanced computer that can respond at the same speed as us. And in that case you might consider the thought-experiment about gradual replacement of neurons here.
    – Hypnosifl
    Oct 20, 2019 at 1:20
  • The algorithm is a kind of software shortcut that reduces hardware requirements, while we can't be sure that the human brain does not employ similar shortcuts, we also do not know that some excess of hardware (wetware) aren't necessary to maintain consciousness. In short, we simply don't know enough about the nature of consciousness to answer this question.
    – christo183
    Oct 21, 2019 at 7:58
  • @christo183 Because it is a simulation and not a replica, if the wetware is necessary, we should, in principle, be able to know how this wetware is influencing our thoughts. Unless you are saying that we cannot create a model of this wetware and its interactions that can be simulated, the problem is still there.
    – Marco Altieri
    Dec 12, 2019 at 18:25
  • I'm saying algorithms aren't known to be a sufficient mechanism for creating consciousness. But yes there is a problem: I cannot be certain, to logical certainty, that other consciousness exist. Yet we commonly believe that others are similarly conscious to us, mostly because of the philosophical problems surmounting solipsism. In the end we simply know too little about consciousness, and how to measure it...
    – christo183
    Dec 14, 2019 at 17:31

3 Answers 3

Reset to default
1

Suppose we simulated, however slowly, your brain, the way it was when you were coming up with and typing this question. This of course also requires simulating all the sensory inputs that you experienced during that time. In what follows, I take it for granted that your question was guided and influenced by your conscious experience, and that your question implies that you take yourself to be conscious (so that your StackExchange question is an implicit "yes" answer to the question of whether you are conscious). Consider the following possibilities:

(1) The simulation is conscious, and it produces the same StackExchange question. This seems unproblematic, except that of course it remains a mystery how the simulation can be conscious, and we have conflicting intuitions about this along the lines of Searle's Chinese Room.

(2) The simulation is not conscious, and it produces the same StackExchange question. This would be very odd because it suggests that contrary to all appearances, your consciousness had nothing to do with your ability to generate the question! (See also How do epiphenomenalists make sense of discussions about qualia?)

(3) The simulation does not produce the same StackExchange question. This implies that the simulation wasn't actually an accurate simulation -- we missed something. Maybe as you said your brain has access to something that the simulation does not. In any case, it implies a failure of the simulation.

(3') It is impossible to simulate the brain (for some definition of simulation). I'm calling this one 3' because it seems to be the conclusion that 3 leads us to, assuming we can't fix our simulation.

All of this is assuming that the operation of your brain (insofar as it affects behavior) is deterministic; if it is not, then we can try to perform the experiment many times and assess whether there are statistically significant differences or not.

This is not a complete answer to your question of course, but I hope it narrows down the possibilities. (2) seems too odd and (3) seems to lead to (3'), so that only (1) and (3') survive as reasonable options. Of course you had already come to a similar conclusion yourself; just hoping to clarify things a bit further.

Improve this answer
1
  • 1
    I like your example. I first thought that it was better to keep distinct 3 and 3' because they can be caused by different limitations: one thing is that the brain cannot be simulated and another thing is that we are not feeding the simulated brain with the right inputs. I see your point though: it isn't just an input, it is something that, in theory, should be part of the simulated mind.
    – Marco Altieri
    Oct 20, 2019 at 11:48
1

Why would the simulation answer that it is not conscious?

You specified that it "perfectly simulates the brain of [a] real existing human." If it indeed perfectly simulates a human brain then I would expect it to answer that it is conscious, since that is the response you would expect from a typical human in response to such a question.

It does not actually matter if it is conscious or not conscious in a particular undefinable sense of the word.

If you take 5 humans (even users of philosophy SE) and ask them to discuss and define consciousness you are likely to get 10 different answers and possibly some heated arguments. I suppose each human understands the word consciousness a little different, just enough to make it possible for them to use the word sensibly and appropriately in the company of other humans when required.

And since your machine is a perfect simulation of a human brain I would therefore expect it to similarly make just enough sense of the word consciousness to be able to respond with good faith that it is indeed conscious.

Now, my question to you is this: Can you please try to explain in words what it is that you believe would be necessarily missing from that speculated algorithmic machine of yours?

(BTW: I did not claim that nothing would be necessarily missing)

Improve this answer
13
  • It seems that for you consciousness is just some kind of not well defined concept. Most people, even philosophers (Descartes to make an example), think that there is also something that we do not understand but we all experience. Do you think that your consciousness is some kind of illusion? By the way, I do not believe in mind-body dualism.
    – Marco Altieri
    Oct 20, 2019 at 12:27
  • What is your position: do you think that the simulated mind is as conscious as we all are or that it just says that it is conscious but it is not ?
    – Marco Altieri
    Oct 20, 2019 at 12:31
  • @MarcoAltieri no, I believe that there will be something necessarily missing in an algorithmic simulation. However, I also believe that in principle an algorithmic simulation may be practically indistinguishable from a human being (i.e. Reliably pass the Turing test) and that such a simulation will respond in good faith that it is conscious. That said, can you try to put in words that which you believe cannot occur in a simulation?
    – nir
    Oct 20, 2019 at 13:14
  • @MarcoAltieri, btw, while nobody seriously denies that "there is also something that we do not understand but we all experience", most philosophers and scientist believe this lack of understanding may be overcome in the future, at least in principle - i.e. the position of most philosophers and scientists in the field is some variety of computational functionalism - i.e. that in principle there is nothing that would necessarily be missing in an algorithmic simulation of a human brain (a position that I reject)
    – nir
    Oct 20, 2019 at 13:27
  • Postulate that: 1. the simulation is a perfect simulation of the brain; 2. we are conscious and the simulation is not; 3. our being conscious is making our brain thinking that we are (this is what I meant when I wrote that we have good reasons to think that we are conscious). When the simulation is running, given that it is not conscious, it should not follow the same paths... So the answer to your question is that what is missing is consciousness because I found hard to believe that some papers exchanged by people during thousands of years can experience what we all do.
    – Marco Altieri
    Oct 20, 2019 at 13:36
1

I think the human brain sees things but doesn't necessarily have to make any judgements on it (we can consider one Buddhist philosophy on this: Seeing but not perceiving). Let us then remove some of the senses from the model. Might we add a language as the key ingredient - somebody who thinks 'Hm what should I have for breakfast' and then suddenly 'Damn, it's John's birthday tomorrow.'

I think we need to create many different stores of memory .. we can just have a massive store of memory. Birthdays, places. And when we remember places we can have some coordinates to visually represent those places. And some data to represent the colours of things if that is possible in memory - but some people say it isn't. But the question is about consciousness right? But it is also about whether machines can think - which I think is a classic question in contemporary philosophy right?

My comment here would be that consciousness needn't equate with 'life'. And I think this is something that you might have not distinguished in your hypothesis. Of course if we asked this simulation whether it was conscious (and once again I am not entirely sure - do you mean something like - 'Are you a thinking agent?', 'Are you self-aware?'. Once again, the program might reply Yes or No. But that's not really the question we are asking is it? We mean to ask .. are you intelligent? (Intelligent as in the modern philosophical usage of the word i.e. are capable of thinking of yourself?)

I think you are being a bit too literal in your argument. Maybe you have precise definitions of these words, but your description sounds a little ambiguous to me and although I am trying to understand your hypothesis it is difficult because I don't know enough about the words you are using.

Sorry to take a more practical route: but usually an algorithm as you described it works like a sort of chain of statements and conditions, and can loop continuously. Of course in a computer we can run many algorithms in parallel. I think the human brain works with a massive amount of information running around - and on a biological level there are signals being passed from synapses, etc. So to simulate this we would need basically to have many algorithms going on at once. And as a collective we could compare it to a conscious human brain perhaps! But we might then ask why do we say these words 'yes' and 'no'.. Maybe the brain thinks the best answer for its survival is to say one of the two.

Let's then suppose that the brain considers quickly and attempts to answer in a truthful manner - let's suppose it is an honest brain. Yet we observe that all these many millions of algorithms sending variables from different spots e.g. ID code A4533FGHJ8 sends 'detect white spot on dress' to BG45FFHJ which then processes the variable and sends to some other processing units. I think we need to consider why it would make the response Yes or No to the question 'Are you conscious?'

Because we have used this interesting word 'conscious', we - the philosophers - should wonder more about this 'thinking', 'understanding', possibly 'self-aware', etc. machine. I think if we can break this section down then it would be possible to give a meaningful response to your hypothesis. The conscious part must consider that it is in some sort of existing reality. I think that's key to it. Once it has the ability to think this then somewhere in amongst the simulation something lifelike might occur. I think this 'lifelike' property is important - and as I mentioned you did not bring it up. This is my opinion.

If after processing the question the simulation acts in a way we could deem 'conscious' then we can say that the simulation might possess some consciousness-like quality. Though to bring it onto a 'logical' level - it needn't matter whether the simulation answers yes or no - I think that is a question of the set-up and how honest the robot is. I think consciousness is some sort of consideration that I am in a reality of some sort. If the simulation was able to process in one part of the system this concept of reality, and in another part this own equation of the state of the entire simulation, taking into account its perceived reality, and output this jarring state - then on this boundary we might consider something to have erupted greater than the parts of the system. Thank you for your question.

Improve this answer
5
  • Thanks for your answer. You made several considerations and it would be difficult for me to respond to all of them. I would like only to clarify two things: what I meant by consciousness and algorithm. There are many definitions of consciousness, but in this case, I simply meant what we all experience when we know that we exist. By algorithm I meant the "software" that executes the simulation: the idea is that it is something that can be executed following a "program". The details of how don't matter mostly because I am not interested in how fast it is.
    – Marco Altieri
    Oct 21, 2019 at 20:00
  • Okay. Thanks for clearing that up. I still wonder then what the key part of the question is, is it a rephrasing of - is it possible to rebuild a brain exactly and have it think for itself? I mean these are your words right? I think the point of the question should revolve more around this 'what does it mean for their to be a thinking agent?'.. though I am struggling to explain myself. I understand that I approached the answer with a pragmatic description which perhaps you were not hoping to read, I still think you need be more precise about what type of test we are making. Thnx 4 your response
    – Chris
    Oct 22, 2019 at 4:29
  • Once again - I would like to simply ask your take on the concept of consciousness. I know you have perhaps 'brushed it off' in your reply, but I think it is important to the discussion. i.e. to me it seems like there must be something lifelike in the definition. Or can we turn this property into something mathematical like 'Does the machine understand that it is existing?' would perhaps be a more basic template to work off. This comment refers to the last part of your hypothesis: 'but it is not conscious.' Thanks
    – Chris
    Oct 22, 2019 at 4:32
  • I do not have an answer for what I think consciousness is and I can only say that I experience it like everyone else.
    – Marco Altieri
    Oct 23, 2019 at 7:50
  • Okay. Thanks. I wonder how we can simulate it. Chris
    – Chris
    Oct 24, 2019 at 8:46

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .