Philosophy Now Forum

Paul Austin Murphy computes the probabilities.

https://philosophynow.org/issues/124/Is_Everything_A_Computer

Paul Austin Murphy article wrote: ↑Mon Jan 29, 2018 1:00 pm Where does the idea that the brain is a computer come from? Firstly, there are strong links often made between brains, mathematical models and computers. Workers in artificial intelligence are keen to tell us that physicists have created accurate models of all aspects of physical reality, and that these models are essentially mathematical in nature. Thus it’s only one step on from there to say that they’re also computable. Thus a computer can model and compute the whole of physical reality, including the brain. Some go so far as to say that mathematics is synonymous with computation, and through maths we can model all of reality (or at least each bit separately), including the brain. The argument here is, very roughly, that once we have mathematically described all the workings of the brain, then a computer could model brain processes. This makes the brain a computer, they say. Other people talk about ‘simulating’ physical systems rather than modelling them. One such person (Aaron Roth) concludes, “if the brain is a purely physical object, which is the only option consistent with our understanding of how the universe works, there is no reason it cannot be simulated.”

The logic in either approach is simple:

(i) All physical objects or systems can be mathematically simulated/modelled.

(ii) The brain is a physical object or system.

(iii) Therefore the brain can be mathematically simulated/modelled.

(iv) Therefore the brain is a computer.

The problem is the slide from x being computable to x being a computer. Even if the brain or its workings were computable, that wouldn’t necessarily make it a computer. Searle’s wall (or window) is digitally computable, and some subset of its behaviour is the behaviour of a computer; but that doesn’t make either the wall or window a computer. Sure, we can define ‘computer’ in such a way as to stipulate, for example that If the brain is computable, then it’s a computer; and do the same for Searle’s wall or window. If that wall (window, etc.) is computable, then it’s a computer… At this rate, almost everything physical is a computer. But, on the other hand, a proper computer must be able to systematically process input to create output. So a computer mustn’t only be computable, it must also be a computer!

The major stumbling block is the actual senses that a sentient intelligence experiences. These are not simply inputs 'computed' by a computer, they are actually something extremely hard to define.
A robot and a monkey both put their hand on the table in front of me. I strike the hands of each with a hammer. Which of the two is going to truly sense pain? At this stage in robotics, all the robot has received as input is perhaps the pressure that had been applied to its hand...
if pressure > 10
...retract hand
...act like am in pain!

It is one thing for a computer to model the entire physical reality, including our brains, but is a huge stretch for it to truly be considered conscious, having the sentient ability to sense. If everything can be modeled mathematically, as some believe, then how can sentient sense be modeled?

Computers are just hybrid machines.

The major stumbling block is the actual senses that a sentient intelligence experiences. These are not simply inputs 'computed' by a computer, they are actually something extremely hard to define.

I'm not sure if many of those involved in artificial intelligence (or others) are always - or primarily - concerned with "sentience" or "experiences" at this moment in time. In terms of what's discussed in the piece, computers and computations are only tangentially related to these issues. However, it is indeed also the case that some people do trace a route not only from computations/computers to intelligence or mind; but also to sentience and experience.

As it is, this is the Big Problem for AI. Is it insurmountable? I don't really know.

In a sense, because of the import of the piece, some would say that sentience and experience are "simply inputs computed by a computer". Or at least that's a possibility. It may also be the case that, in the future, sentience will be simulated or actually created because of computations - if used in the right way, with the right hardware, etc.

A robot and a monkey both put their hand on the table in front of me. I strike the hands of each with a hammer. Which of the two is going to truly sense pain? At this stage in robotics, all the robot has received as input is perhaps the pressure that had been applied to its hand...

I don't think that anyone - not even an extreme adherent of Strong Artificial Intelligence - believes that, at this moment in time, we have a "robot" that can "sense pain". Nonetheless, it could happen in the future... Or could it?

It is one thing for a computer to model the entire physical reality, including our brains, but is a huge stretch for it to truly be considered conscious, having the sentient ability to sense.

I agree. I also have a problem with the notion of the entire universe being modelled or computed. How would that work? I suppose that if you get down to the fundamentals of physics, then those fundamentals can be modelled and therefore the entire universe can be modelled. Though there's a mountain of ifs and buts here. What won't be modelled are the indefinite number of holistic or emergent features of the universe - that's if scientists or philosophers accept such things in the first place!

However, the modelling/computing of the universe (or a brain) isn't itself meant to bring about (or constitute) consciousness. The modelling/computing is a different thing - though related.

My statement "sentience will be simulated.... because of computations" (above) doesn't make sense. I'm not sure how sentience can be simulated at all. Intelligence can be; though not experience/consciousness/sentience itself... Indeed there's an argument that if intelligence is simulated (by a computer, robot, etc.), then there's no simulation about it! Behaving intelligently is actually being intelligent... (That, of course, has a behaviourist ring to it.) However, all this doesn't also apply to sentience/experience/consciousness....

A human brain functions biomechanically to form the senses but interpretation is not biomechanical. It's cognitive. Besides a person can function without words or instructions. Words and instructions follow cognition. A computer requires words, tables, definitions and instructions. Even if the results can be modeled, the functions are different.

Moreover humans can be motivated. Computers must be programmed.

it may all be in the programming...

but not all humans behave intelligently...

can desire be programmed?

-Imp

A bit of a problem with the assertions of what "is a computer" in the paper. For one, a Turing machine is a specific formal architecture, and DNA does not qualify. The state machine requires the tape to be written among other things. Despite being able to correspond various biological parts to the parts of a Turing machine, they do not function similarly. That's OK, since nobody is claiming that there is a Turing machine going on anywhere in a biological system. A brain may well be a computer, but it doesn't operate via a tape or some instruction set.

This is the part with which I have the problem:

Paul Austin Murphy wrote:The argument here is, very roughly, that once we have mathematically described all the workings of the brain, then a computer could model brain processes. This makes the brain a computer, they say.

If 'they' say this, they're wrong. I can simulate snow melting with a computer, but that doesn't mean melting snow is a computer, it means the process can be computed, or simulated. The last step of the argument seem to rely on this false assertion.

(iii) Therefore the brain can be mathematically simulated/modelled.
(iv) Therefore the brain is a computer.

IV simply doesn't follow from III. I might agree that the brain is a computer of a sort, but not because of this reasoning. The function of a brain can be simulated via computer, therefore a computer can act as a brain, just like a computer can act as melting snow. It makes no statement as to the nature of what the simulated thing is doing.

A computer follows a program. A person follows his/her mind. The program is fixed. The mind is not.

Hi Noax,

For one, a Turing machine is a specific formal architecture, and DNA does not qualify.

The article doesn't really state what a computer is. It states what some philosophers and others take it to be. Perhaps that's what you meant.

DNA does have a "specific formal architecture". Whether it's the same as a Turing machine, it's hard to say. In the most abstract sense, surely the answer is that it can be seen to have such an architecture. DNA takes in "information" and produces "output". We can question these metaphorical uses of the words "information" and "output". Then again, we can do the same with most Turing machines. That is, a DNA molecule doesn't know what information is or how to use it. And slightly less abstractly, it doesn't really produce output either. There are causal and physical consequences of DNA, but is that output? In a loose sense, I suppose it is. I'm less sure about input as "information"; which can also be seen in simple causal and physical/biological terms.

The state machine requires the tape to be written among other things. Despite being able to correspond various biological parts to the parts of a Turing machine, they do not function similarly. That's OK, since nobody is claiming that there is a Turing machine going on anywhere in a biological system. A brain may well be a computer, but it doesn't operate via a tape or some instruction set.

Some people do think that Turing machines are "going on" in "biological systems". That's the case because of the unique abstractness (or vagueness) of a Turing machine.

No defender of AI has ever denied that there are "other things" which differentiate Turing machines from biological systems. Nor do they demand complete similarity. However, all these other things and differences simply aren't important when it comes to whether or not something is a computer, a Turing machine, or even a mind. I'm not saying that this is my own position; though it's certainly a position many people hold.

Why does it matter that a brain doesn't "operate via tape"? Remember that Alan Turing's first Turing machine was a complete abstraction. He had no idea how it would be physically implemented. It did come to be physically implemented; though, in theory, it could be implemented in an indefinite amount of ways. (This is also what functionalists - in the philosophy of mind - believe about mental states and their various and many physical bases or implementations.)

I can simulate snow melting with a computer, but that doesn't mean melting snow is a computer, it means the process can be computed, or simulated.

I don't think the argument is that simulation alone shows us that what's simulated is a computer. The argument needs more than that. It must mean that if something can be simulated, then that at least implies that it's a computer or has elements that work as computers. That is, the simulation wouldn't work if that which is simulated weren't a computer.

Having said all that, I agree, I don't think it makes much sense. It's not my own position.

The last step of the argument seem to rely on this false assertion.
(iii) Therefore the brain can be mathematically simulated/modelled.
(iv) Therefore the brain is a computer.
IV simply doesn't follow from III.

That's not my own position. It's a position which I've heard and read advanced. Perhaps not so simply as I've put it. That may be the problem - they haven't put it that simply. I don't think that iv) does follow from iii). However, that seems to be the position when it's boiled down to basics. I might well have created a straw target here; though I don't think that I have.

The function of a brain can be simulated via computer, therefore a computer can act as a brain, just like a computer can act as melting snow. It makes no statement as to the nature of what the simulated thing is doing.

Certain functions (in the plural) can be simulated; though not the brain as a whole. And because of that, no computer (at present) "can act as a brain". However, if a computer or machine could/did simulate (or replicate) the workings of an entire brain, then there would be no simulation at all - it would be a brain. The main difference would only be what the two things are made of different physical materials. Then again, John Searle strongly suggests that biology matters very much.

Does a brain or a computer have an intent? I don't think so. Talking about the computer and the brain ignores intent which is a key characteristic of any person. Comparing an intentionless brain to an intentionless computer sways the discussion to an irrelevant point.

Paul Austin Murphy wrote: ↑Thu Feb 01, 2018 5:38 pm Hi Noax,

Thank you for responding to my post. Most authors seem to never bother.

For one, a Turing machine is a specific formal architecture, and DNA does not qualify.

The article doesn't really state what a computer is. It states what some philosophers and others take it to be. Perhaps that's what you meant.

I meant a Turing machine, which was specifically likened to the workings of DNA. Yes, a computer in general can be any number of things, and instruction sets are not necessary. I myself have experimented with instructionless analog computers and finite asynchronous state machines which bear little resemblance to the classic Von-Neumann architecture of which one often first thinks.

DNA does have a "specific formal architecture". Whether it's the same as a Turing machine, it's hard to say.

No, that was my first (minor) point. The Turing machine is a very specific abstract architecture, and the working of DNA is not it. It is a mathematical abstraction, to be sure, not meant as a blueprint for a workably efficient computing engine. The abstraction can be proved to be able to implement any computable function including said Von Neumann machine above.

DNA takes in "information" and produces "output".

DNA IS information, but it doesn't take it in. The DNA does not change in normal operation, and Turing machine tape does. Hey, I'm not arguing that DNA operation is not a mechanism of information processing. I think it is, but it isn't a Turing machine is all.

We can question these metaphorical uses of the words "information" and "output". Then again, we can do the same with most Turing machines. That is, a DNA molecule doesn't know what information is or how to use it. And slightly less abstractly, it doesn't really produce output either. There are causal and physical consequences of DNA, but is that output? In a loose sense, I suppose it is. I'm less sure about input as "information"; which can also be seen in simple causal and physical/biological terms.

All depends on your definitions of knowing, information, and output and such. I'm pretty liberal about it and would probably be inclined to say that DNA does indeed know information, but has no need to know 'what information is', so no, it doesn't know that. Similarly I would be inclined to say it produces output, since it would not serve its function if it didn't.

Some people do think that Turing machines are "going on" in "biological systems". That's the case because of the unique abstractness (or vagueness) of a Turing machine.

Turing machine definition is hardly vague. The number of states and the next-state table is not fixed, but otherwise the definition is pretty exact. No natural Turing machine has ever been discovered, and for good reason. They're hideously inefficient.

Remember that Alan Turing's first Turing machine was a complete abstraction. He had no idea how it would be physically implemented.

Implementing one is pretty trivial. I cannot see how Alan Turing would find it difficult to build one, but not sure what purpose would be served by it. I've seen it done with Tinker Toys, and Scientific American had an article showing an implementation using only a battery-run toy train and a lot of track.

Why does it matter that a brain doesn't "operate via tape"?

No talk of brain operation up to this point. The discussion was of DNA chemistry acting as a Turing machine.


My more primary point:

I can simulate snow melting with a computer, but that doesn't mean melting snow is a computer, it means the process can be computed, or simulated.

I don't think the argument is that simulation alone shows us that what's simulated is a computer. The argument needs more than that. It must mean that if something can be simulated, then that at least implies that it's a computer or has elements that work as computers. That is, the simulation wouldn't work if that which is simulated weren't a computer.

You're saying I can't simulate snow melting because melting snow is not a computer? That's clearly false. The weather forecasting guys depend on snow melt as part of their simulations.

Having said all that, I agree, I don't think it makes much sense. It's not my own position.

OK then. I disagree with the position where iv follows from iii. Melting snow can be simulated, but it takes a very loose definition of 'computer' to qualify snow melting as one.

Certain functions (in the plural) can be simulated; though not the brain as a whole.

Seems like an unbacked assertion. If it is physics, it can hypothetically be simulated in totality, even if no hardware today is up to the task. It would have to account for all the chemical effects as well since there is more chemistry than circuitry going on. It would also need a feed of input and output, a world with which it can interact. Lots of reasons like these are why you're not going to see such a simulation in practice. AI might outsmart us some day, but not by simulating biology. That you would deem such a thing to be conscious or not seems only to be a matter of definition of the word.

There are no turing machines in the physical universe because the physical universe does not utilize information. Machines that utilize information do so because humans made them that way. Melting snow is not a computer (that's silly) because melting snow does not utilize information. Computers utilize information because humans made them that way. They didn't emerge from natural information and the information they produce has no effect on the physical computer. I suspect that genes have a bio-chemical effect which is physical not informational. To have an informational affect the cells in the body would need to have interpreters. If physical is the case, then the information associated with genes is not causative and only seems so to people who see computers everywhere.