On Consciousness

[Piggy]

It's as if you didn't read my refutation of Massimo's comparison of consciousness to sugar, or you didn't understand it.

Photosynthesis outputs sugar.

What does the brain output?

How is consciousness like sugar?

How is sugar like hauling brush to the dump?

It's not.

But neither can be accomplished by logic alone... in fact, no real-world phenomenon can be.

There's no reason to believe -- and every reason not to believe -- that colors and sounds and smells and pain and pleasure and all the other conscious behaviors can be performed by logic alone without real work.

 

[Piggy]

They wired up some artificial neural networks according to a hypothesized pathway that they thought might be part of how the mamallian brain controls motor movements via imagination, hooked the simulation up to a robot arm with a camera sensor, and it worked. The robot imagined that if it turned a certain way it would see an object of a color that it "liked" and because of that imagined reward it actually turned.

Great post. I don't have time to respond now, but I will when I get back later today.

I think we can actually use this post to point out where some of our differences aren't, and where they are.

More later.

 

[rocketdodger]

"Information" is a term we can use in a variety of ways, as you know.

In the Wolfram sense, it's totally equivalent to physical activity and properties.

If you use the term to refer to degrees of entropy, then the equivalence is not complete.

If you use the term in such a way that it allows an informational "world of the simulation" to exist which can (and must) be described in radically different ways from the physical activity of the simulator -- as y'all do -- then you (not I) have divorced the information from the physical behavior in a radical way.

If we use the term "information" in the sense of something that can be encoded and decoded, then that information becomes independent of physical reality in very important ways.

I can send the same message to you -- say, "Help!" -- using the motion of flags, or an arrangement of rocks on a beach, or by flashing lights in certain patterns, or by causing air to bounce inside your ears in a particular way. That means that there must be a "difference between information and physical behavior" when we're talking about this kind of information, or else all those different physical set-ups couldn't carry the same kind of information.

However, there's a caveat when talking about this last sort of information, because it requires an encoder and decoder.

Someone who doesn't speak the same language as me (whether that's English or semaphore) won't get the message, unless they use other shared information such as our human understanding of what a frantic cry sounds like, which is another kind of encoding and decoding.

In other words, there actually is no "information" in the rocks or the lights or the flags or the air, just like there's no pain in the knife.

So we can talk about the brain objectively (without inserting any encoder and decoder) in the Wolfram sense, and we can speak metaphorically about "information in the brain" when we're really referring to the brute physical activity, and we can talk about information exchanged in a system that contains more than one brain.

But to talk about this last kind of information existing objectively in the isolated brain, well, that makes very little sense.

I'm just so tired of having to wade through walls of text in order to make sense of your responses to what should be extremely simple questions. You insist on smothering people with words, typically when you are just wrong about something.

Piggy, you can't divorce information from physical behavior because everything is physical behavior.

You are as wrong as wrong can be. "We" don't consider the world of the simulation to be different from the physical activity of the simulator, I have no idea why you keep saying this even though we keep saying we don't mean it. How stubborn does someone need to be to repeatedly tell multiple people that they mean something other than what they explicitly say they mean?

 

[rocketdodger]

But neither can be accomplished by logic alone... in fact, no real-world phenomenon can be.

There's no reason to believe -- and every reason not to believe -- that colors and sounds and smells and pain and pleasure and all the other conscious behaviors can be performed by logic alone without real work.

Here you go again -- what the heck are you talking about? Who has ever said that logic can be performed without real work?

The behavior of making absurd stuff up and then refuting it just so you can say "see, I'm right" is less than useful.

Here, let me try it: There's no reason to believe that you can get green paint by mixing red and black paint.

By saying that, I implicitly suggest that you are stupid enough to think that the opposite is true, eh? Why else would I have said it, if it didn't need to be said, eh? And now anyone reading the posts might think I am a little bit smarter than you, because something I said in a post is obviously true. I obviously have a grasp on reality because I said something obviously true ( never mind that nobody ever said the opposite ). Hooray for me!!

 

[dlorde]

If you're talking about the IP sense of "compute" then the outcome of any computation is symbolic, and requires a reader.

What, exactly, do you mean by a 'reader'? One can hook up a symbol manipulating computer to physical effectors and have its computations cause physical actions. I'm happy to accept that the interface to an effector is a 'reader' (although 'interpreter' seems more precise), but it leads me to query why you say the outcome of a computation requires a reader. Computations directed to controlling effectors require those effectors in order to cause an effect, if you remove the effector interface the computation can still be done, although to no useful effect.

In the case of consciousness, who or what is the 'reader'? It seems clear, as evidenced by 'locked-in' syndrome, that consciousness doesn't require physical effectors, but seems able to be maintained effectively in isolation. I can only think that by your requirement above, consciousness must be reading itself - a clear expression of SRIP

 

[rocketdodger]

I've never said we know nothing about it, of course.

You have said we don't know much, in a number of ways, a number of different times.

And although there are some hypotheses being tossed around, the focus at the moment is just to continue investigating NCCs and see where it goes, which is pretty raw stuff.

Forgive me if I don't take your word for it.

Besides, I quoted them directly so the viewpoints expressed are theirs.

No offense, but I certainly think you are guilty of quote-mining in almost every single case. I absolutely cannot stand quote mining ...

Anyway let's see if the folks I cited have done any actual research....

You realize that it is a double edged sword, and if you are not good at swordplay ( I don't think you are ), you are more likely to hurt yourself, right?

Christof Koch of Caltech, author of Biophysics of Computation: Information Processing in Single Neurons -- his research focuses on visual systems, information theory, and questions of machine sentience.

Ok, he definitely sounds like he would know a thing or two. I stand corrected on that front.

On the flipside, just a cursory glance at his research page should convince anyone capable of reading English that he feels the current state of research is beyond "we don't know much," as you claim he would say. He certainly doesn't seem to be just slogging through NCC data with little clue as to how to put it all together.

Ned Block from NYU -- you might be interested in his panel appearance along with Koch and info-theorist Giulio Tononi in the "Consciousness and Intelligence" seminar of last year's "Brains, Minds, and Machines" symposium at MIT, which btw I think demonstrates the actual current intersection of neurobiology and info theory.

I disagree with you on this one. This guy is neither a programmer nor a biologist. He is a philosopher and psychologist. Him saying "we don't know much" carries about as much weight with me as Romney saying he could fix global warming.

Bernard Baars of Berkeley who's done all kinds of things.

If you think I don't know about Baars then you really are crazy, especially since the global workspace model is used all over the place in the world of machine consciousness, in fact I am trying to get my colleagues to agree to experiment using it for game AI.

But here is the thing, piggy -- he hasn't actually written a paper since 2002, according to that cv page of his. And here is another thing, piggy -- he is a really smart guy whose position changes over time, you can see that in the history of his work.

So are you going to sit there and honestly tell me you can find quotes of his to the extent of "we don't know much" from anytime in the near past? Even a quote more than 5 years old is irrelevant, given the speed at which research moves forward today.

It sort of boggles my brain that you would consider Bernard Baars as a supporter of the notion that we are still just slogging through NCC data trying to make heads or tails out of it.

V.S. Ramachandran of UC San Diego's Center for Brain and Cognition, who has done some research.

Yeah, let's talk about quote mining piggy. Quote mining is when you look through something someone has written just to find isolated quotes that could be interpreted as supporting your position. It isn't inherently bad, except that often ( as is the case with you ) the context of the quote is completely changed. Case in point:

Ramachandran's site says "It is ironic that although we now have a vast amount of factual information about the brain, even the most basic questions about the human mind remain unanswered."

Whoah, it looks like you might be correct!! Hmmm.. wait a second though, I should keep reading...

"Why do we laugh, i.e., make a rhythmic sound and bob our heads in certain situations? Why do we cry? Why does a salty liquid flow down our cheeks when sad? How does the human brain create and respond to art? Why do we enjoy music? What causes us to dance? What makes some of us so amazingly creative in mathematics, science, and poetry? How are metaphors represented in the brain? What is "body image" and why does it get distorted in anorexia nervosa? How did language evolve?"

Ohhhh, I see what he means. He means what humans would consider the most basic questions, not the fundamental questions. I think it would be slightly dishonest to attribute the position of "we don't know much" to someone like Ramachandran based on a quote taken so out of context, don't you? ( not sure why I'm asking you, since you did it !! ) Let's keep reading though....

"Then there are more basic questions. How do we see color? Why can we pay attention to only one thing at a time? How do we recognize faces so effortlessly?"

Ahh, those are the kinds of things I had in mind. So we don't know much about them??

"Neuroscientists and psychologists have, in the past, shied away from such questions, but our center has become well known for tackling questions such as these experimentally, questions that have traditionally been the preoccupation of philosophers."

Hmmm....so they are actually doing research on these things? Call me crazy, but that doesn't sound like a lab full of people who think "we don't know much."

Michael Gazzaniga who was on the original split-brain study team, led the continuing research, and is the current president of the Cognitive Neuroscience Institute.

I don't even need to touch this one. Please cite a recent quote ( not mined, please keep it in context ) where he says anything remotely similar to "we don't know much."

Steven Pinker of Harvard, advocate of the computational model of mind

Same. Please cite a recent quote ( not mined, please keep it in context ) where he says anything remotely similar to "we don't know much."


So I admit I was partially wrong -- you do actually reference people who know, at least partially, what they are talking about. However I would strongly argue that you are taking their sentiments completely out of context to support your own position. Prove me wrong. Start using recent statements and link to the full articles where you got them. Avoid quote-mining.

EDIT -- what I am trying to figure out is why you are a proclaimed materialist yet you seem to have a desire to "dumb down" the current state of research, almost as if you want to convince people that we know far less than we really know. It is peculiar ...

 

[Belz...]

What do you mean by "data"?

If I consider food "data", then my body can accept it, store it, and process it according to a predetermined set of rules.

And if I consider cows to be tables, I can put napkins, set up dinner and and eat on one.

 

[Belz...]

What, exactly, do you mean by a 'reader'? One can hook up a symbol manipulating computer to physical effectors and have its computations cause physical actions.

An better answer than mine.


So nobody has anything to say about my biological truck ?

 

[Mr. Scott]

I can't see a computer, with the rules it must be assembled with, coming up with:
"It was a dark and stormy night"... or "Call me Ismael"... or "In the beginning..",
particularly the last which has what a computer would recognize as physical absurdities, much less plots that have been pre-figured in the author's mind to go in a specific direction, which may or may not be possible in a physical world.
A computer that might do that would have so few constraints its output would resemble that herd of monkeys more often than not.
Dawkins got his program to find a particular phrase in Shakespeare, but he had to direct the computer to that goal, external to the programming of the phrase generator.

Ah, the old "fine arts" argument against consciousness as data processing. It's a fallacy class: argument from ignorance and argument from personal incredulity. It follows the form:

"I don't know (or can't imagine) how computers could write beautiful prose, so computers can't write beautiful prose, and human brains must do it with some form of magic (outside the known laws of physics)."

The argument unfortunately implies that people who don't produce fine art are not fully conscious.

There are some poem generators online. Here's a nice Haiku Generator I found. I don't doubt that some of its poems would challenge one to sense they were computer-generated.

 

[dlorde]

So nobody has anything to say about my biological truck ?

Apart from the gratuitous surgical tweaking of neurons and the novel method of locomotion, I couldn't really see much difference, in principle, between it and pack animals used for transport, e.g. donkeys, horses, elephants, camels, bullocks, etc.

It did bring to mind Boston Dynamics' 'Big Dog' - should we call that a truck?

 

[Mr. Scott]

I just had an idea for an analogy and I'm not sure if it'll work. Here goes...

A truck is a metal frame over an axle and four wheels, transports stuff and goes places, right ?

Let's do the consciousness/computer thing in reverse.

Let's say I build an organism in a lab. It's a vertebrate with a tough skin with hard plates. It has flabs on either side that you can open to get inside its frontal cavity, in which there are ganglions you can push or pull to trigger certain functions of its nervous system. In front are two white bioluminescent organs. Same thing for the rear, but red. There is another cavity at the back. You can input a slush-like food in an orifice to the side, which feeds the organism, and comes out as fart from the rear. Under it are four legs arranged in a way that they can spin on themselves without damage, so they turn like wheels. Using these attributes, you can shove stuff in its back cavity, hop inside, press a few ganglions and go see your father-in-law in Jersey.

Is it a truck ?

I'd say you'd have a number of truck enthousiasts, amongst others, claiming it can never be a truck because it's organic, etc.

Yes, a truck is substrate-independent. If it's above a certain size and licensed to transport stuff on the roads, it's a truck, even if it's made of wet, fleshy carbon-based and DNA-directed materials.

The brain processes data. It would, in theory, process data the same way, generating fine arts and reports of internal subjective experiences, if it were made of metal and ran on gasoline, I maintain.

 

[Mr. Scott]

Thermodynamic consequence of the qualia hypothesis

If it's true that the brain "performs" red (and qualia in general) then there must be a consequence in thermodynamics.

Some energy has to be expended performing red, and some of this would have to be recovered (with losses) in the process of appreciating the performance.

Wouldn't this have to be happening if the arguments of qualiaphiles were correct? Is there any evidence for the metabolism needed for qualia generation and appreciation, beyond what's involved in action potentials and neurotransmitter activity?

 

[Fudbucker]

Ah, the old "fine arts" argument against consciousness as data processing. It's a fallacy class: argument from ignorance and argument from personal incredulity. It follows the form:

"I don't know (or can't imagine) how computers could write beautiful prose, so computers can't write beautiful prose, and human brains must does it with some form of magic (outside the know laws of physics)."

The argument unfortunately implies that people who don't produce fine art are not fully conscious.

There are some poem generators online. Here's a nice Haiku Generator I found. I don't doubt that some of its poems would challenge one to determine if it was computer-generated.

To be charitible, I think Iratant was talking about computers as they currently are.

A couple Haikus from the haiku generator:

fisherwomen climb
giddily, moose flying plum
lying conjurers

pumpkins recover
wistfully, laughter festers
rosemary frolics

blazing strut gravestone
dwindles, frivolous lustrous
dunes pulsate, mealy

I'm not too impressed
Haiku program is lacking
Hail new machine lords!

(OK, that last was my own)

So computers aren't at the "dark and stormy night" level yet. However, there's no reason to think they won't be. Maybe fairly soon.

 

[Piggy]

Piggy, you can't divorce information from physical behavior because everything is physical behavior.

You are as wrong as wrong can be. "We" don't consider the world of the simulation to be different from the physical activity of the simulator, I have no idea why you keep saying this even though we keep saying we don't mean it. How stubborn does someone need to be to repeatedly tell multiple people that they mean something other than what they explicitly say they mean?

You keep saying you don't mean it, and then you make arguments which presume it.

That's what troubles me.

 

[Piggy]

Here you go again -- what the heck are you talking about? Who has ever said that logic can be performed without real work?

You're not seeing what I have plainly said.

Of course logic must be performed with real work.

But if you only demand (or have) enough real work in the system to run the logic, then all you have done is run the logic.

If you want to run the logic and do something like make enzymes or move objects or play sounds, then you need enough work to run the logic and enough work to make the enzyme or move the object or play the sound.

Nobody disagrees with that, of course.

Where you and I differ is that you believe that no additional work is required for conscious behavior like colors and sounds and textures to be performed, above and beyond what's necessary to "run the logic" because these phenomena result directly (and exclusively) from the logic, which is not entirely dependent on substrate -- in other words, you can run the logic with a brain or a computer or whatever, as long as it runs, that's all you need.

I've called this the "pure programming" point of view, for short, on other threads.

If this is accurate, then machines can be programmed to be conscious, they need not be otherwise specially built to be conscious. Creating conscious machines becomes a question entirely of computer engineering rather than mechanical-electrical engineering (except as the latter serves the needs of the former).

The neurobiological approach sees things a bit differently, and this is what I've been saying about the requirements for doing work....

Per the NB approach, the neural correlates of consciousness (NCCs) aren't so much seen as a substrate supporting logic which in turn enables conscious behavior to occur, but as the direct mechanism itself -- one which is much often simpler to talk about in terms of information, but which nevertheless must be plain old physics at the end of the day.

From your point of view, comparing consciousness to photosynthesis or hauling trash to the dump makes no sense.

Because from your point of view, unlike photosynthesis or hauling trash, consciousness is a pure-programming problem.

Which means that running a digital simulation of the brain will create consciousness -- with no extra work required for doing so, beyond what's needed to run the simulation -- even though a digital simulation of a leaf will never make sugar and a digital simulation of a truck will never haul your trash to the dump.

From a neurbio point of view, the challenge to building conscious machines is a bit different, because we assume that electro-mechanical engineering will be crucially important over and above its contribution to supporting the logic.

In other words, the machine brain will have to behave in real spacetime in some ways like the biological brain to make these behaviors actually occur, and this is the bit of work (despite your dubbing it a "magic bean") which I'm referring to.

So the simulation will not actually create an instance of blue or pain or the smell of coffee or anything else, unless it's got the hardware to make that happen in real spacetime.

That's the difference.

Sorry if it takes a while to explain it.

 

[Piggy]

They wired up some artificial neural networks according to a hypothesized pathway that they thought might be part of how the mamallian brain controls motor movements via imagination, hooked the simulation up to a robot arm with a camera sensor, and it worked. The robot imagined that if it turned a certain way it would see an object of a color that it "liked" and because of that imagined reward it actually turned.

Sounds good.

I don't think it is necessarily directly relevant to the search for phenomenal consciousness. Rather, I think it is relevant to all of the arguments you have made against the computational model, for starters.

You have explicitly stated that simulating neurons can't lead to real world results, that what happens in simulations is just ones and zeros that only humans can make sense of. Well, you are wrong about that, because these guys hooked up a robot to simulated neurons, and it did something on its own.

No, that's not what I've said, explicitly or otherwise.

Of course they can lead to real-world results -- happens every day. It makes as much sense as being able to send letters.

What I've said is that when you do insert a simulation into part of a process, it has to be the case that you need signal-in and signal-out.

The trouble comes if you need any actual physical mechanism to do real work in spacetime anywhere along that part of the system that's been swapped with a simulator.

If you do, then your system breaks because the simulated stuff is in informationland and can't do any work... the only work being done is the simulator being run.

So yeah, there's no problem with simulated neurons "leading to real-world results".

But there's a problem with simulators allegedly performing work that they're not performing in spacetime (which from your point of view isn't a problem, I know).

You have explicitly stated your distaste for AI research because it goes on in some sort of vacuum from brain research. Well, you are wrong about that, because these guys made an artificial neural network based on known brain pathways, in fact the paper goes into a good amount of detail describing such pathways and even provides citations of its own for the "brain research," as you call it, that the researchers educated themselves with.

My objections aren't with AI research, never have been.

I've consistently said it's interesting, important, and useful.

My objection is to its application on this forum to the question of consciousness, that's all.

I object to irrelevant discussions based on AI research that doesn't have any bearing (or precious little) on consciousness research.

And I haven't claimed that AI researchers ignore brain research. I've complained that the comp.lit camp on this forum much prefers to discuss research on non-conscious machines than to discuss current neurobiology, and I'm dead right about that, as far as I can see.

Those are two very different things.

You have explicitly stated that computing is somehow not "real time" and thus a program is incapable of emulating what is necessary for consciousness. Well, you are wrong about that, because these guys simulated the neurons in software, and those neurons didn't have any problem interacting with the world in "real time."

I don't recally anyone saying that computing is not real time.

I do recall discussions about whether a conscious machine could stay conscious at any arbitrarily slow operating speed.

In any case, nothing in this experiment is shocking to me, it doesn't go against anything I've said so far, as far as I'm aware.

You have explicitly stated that there is no indication by any research thus far that the functions of the brain necessary for consciousness can be divorced from its "real-spacetime function" which you have explicitly stated is surely more than what known computers can provide. Well, you are wrong about that, because these guys seem to have emulated at least a few of the functions required for that thing we call "imagination" in a computer.

That's not quite what I've said, but close enough for our purposes here anyway.

Yeah, I've said that running logic alone won't get consciousness done, just like logic alone won't make a machine sing or run or perform any other behavior.

That's why I always talk about conscious machines rather than conscious computers.

But this only means that, for the particular function of consciousness, you're going to need properly set up hardware.

What these guys are doing really has nothing to do with that question, I'm afraid. It's not that what they're doing isn't damn interesting and useful -- it certainly is -- but it doesn't have any bearing on the question of the hardware component of consciousness, I'm afraid.

Whether or not the research explains much in terms of consciousness for you isn't really the point. The point is that the research clearly shows that just about every single one of your counterarguments isn't valid. That's why I brought it up. Turns out there are dozens of projects like this, just go to the aisb home site and look at their symposiums over the last decade.

And that's fine, and very cool, but my objections have never been what you seem to think they are. It's a matter of relevance, nothing else.

So my question would be, given that there is actual research going on where people are looking at the mamallian brain and emulating its functions with artificial computers and getting behavioral results that confirm their hypotheses, why do you think that simply making post after post trying to argue to the contrary is productive? Once you move past this silly need to control the conversation and "win" a debate at any cost maybe you will start to learn a few things, eh?

I'm not arguing against that fact.

I know there are folks doing this research.

Some of the people I cite are involved in it, and have written about it.

Yes, this is "actual research" but it is not particularly relevant actual research. And I wonder why it ends up dominating every thread about consciousness to the exclusion of much more relevant research on the human brain itself, which is indeed focused on the question of consciousness.

Once we crack the big questions about consciousness neuobiologically, then we'll be able to design similar machine tests in that area.

I sincerely hope I'm dead before that happens, though.

 

[PixyMisa]

You're not seeing what I have plainly said.

Of course logic must be performed with real work.

But if you only demand (or have) enough real work in the system to run the logic, then all you have done is run the logic.

Yep..

Where you and I differ is that you believe that no additional work is required for conscious behavior like colors and sounds and textures to be performed, above and beyond what's necessary to "run the logic" because these phenomena result directly (and exclusively) from the logic, which is not entirely dependent on substrate -- in other words, you can run the logic with a brain or a computer or whatever, as long as it runs, that's all you need.

I've called this the "pure programming" point of view, for short, on other threads.

If this is accurate, then machines can be programmed to be conscious, they need not be otherwise specially built to be conscious. Creating conscious machines becomes a question entirely of computer engineering rather than mechanical-electrical engineering (except as the latter serves the needs of the former).

Yep.

The neurobiological approach sees things a bit differently

Nope.

That's the problem. There is no "differently". Either at some level it's computational, or it's magic.

 

[tensordyne]

That's the problem. There is no "differently". Either at some level it's computational, or it's magic.

So I guess then trucks and leaves are magical. I simply did not know.

 

[tensordyne]

If it's true that the brain "performs" red (and qualia in general) then there must be a consequence in thermodynamics.

Some energy has to be expended performing red, and some of this would have to be recovered (with losses) in the process of appreciating the performance.

Wouldn't this have to be happening if the arguments of qualiaphiles were correct? Is there any evidence for the metabolism needed for qualia generation and appreciation, beyond what's involved in action potentials and neurotransmitter activity?

Interesting idea, but I do have to note that if the neurobio is approach is correct, qualia generation might be the physical result of action potentials and neurotransmitter activity without any need of bringing in the computational role of the same. Of course, I would like to add in that qualia generation as per thermodynamics and CEMI might come in the form of energy generated to create the EM field.

 

[Piggy]

Any system that is Turing complete^WP. That's what anyone who understands computers means by the term.


The brain is Turing complete. This was understood by everyone in the all related fields the minute we came up with the term "Turing complete".

Forgive me if I don't just take your word for that last bit about the brain.

And it's funny that your claim about the brain isn't mentioned in the article you cite, but my claim about the whole universe is.

And the folks over at Stanford don't seem to agree with you.

A myth seems to have arisen concerning Turing's paper of 1936, namely that he there gave a treatment of the limits of mechanism and established a fundamental result to the effect that the universal Turing machine can simulate the behaviour of any machine. The myth has passed into the philosophy of mind, generally to pernicious effect.

In fact, their treatment of this "myth" sounds like a pretty good paraphrase of many of your claims in this forum.

So unless you can demonstrate that this is so, and unless you can cite leaders in the field agreeing with you, I'm betting that the Stanford guys are right and you're wrong.