On Consciousness

[quarky]

I know, though I asked you why consciousness studies in particular seemed to enrage you. What I hear from you is it's just one of many areas of scientific inquiry you'd prioritize lower than others, like feeding the hungry.

You may want to dis some areas of inquiry as conforming to a religious dogma of science. I'd argue that the yield of science, including past work where no yield was anticipated, resulted in unanticipated benefits. The track record of pure science is awesome. Compare that to the track record of tradition religions. I think it's the comparison is weak. Science works.

A conscious food distribution network might be very effective. Let the AI research continue. New tools will be used for good and bad purposes. It's always been so. (I'm not afraid of conscious sex robots, though unconscious sex robots don't seem that shabby. joking!

I totally dig your concerns and observations. For me and my various rage, its never been a question of science vs religion. its always been about science, and how can it become more sane and ethical.

of course, there are movements within science that lean this way.
Some of the most respected scientists would have very little issue with my spewage on this matter. Marrying ethics to science is a slippery slope, but the opposite (marrying science to business) turns out a lot of useless crap.
Or worse, harmful crap.

But fear not, my blows against the empire are pathetically ineffective.
My personal history, of course, flavors my attitude. I realize such anecdotes don't belong here, but here's one anyway:

My son in law, whom is a brilliant and highly sought after chemical engineer, was actually working for Bechtel, when they decided to privatize water in Bolivia. He met my daughter when she was working for those exploited peasants. She's not a scientist. He was brilliant enough to catch the wave of her innate wisdom.

 

[Modified]

No, his advice was not about how to be successful but how to be fulfilled.

And for that purpose it is a failure as well.

 

[rocketdodger]

Heard about this before, got a link? Conscious behaviours in what sense?

The presentations of page 26 and 50 in this symposium are some fairly simple ones from over half a decade ago: http://sacral.c.u-tokyo.ac.jp/pdf/Ikegami_MachineConsciousness_2005.pdf

Conscious behaviors in the sense of rehearsing possible actions in their mind using the same neural network pathways as actual perception, AKA imagination.

 

[Zeuzzz]

That makes no sense. Where does this consciousness come from ?

Science shows us that it's the other way around, by the way.

Prove its the other way round then, i'm all ears.

 

[Zeuzzz]

The presentations of page 26 and 50 in this symposium are some fairly simple ones from over half a decade ago: http://sacral.c.u-tokyo.ac.jp/pdf/Ikegami_MachineConsciousness_2005.pdf

Conscious behaviors in the sense of rehearsing possible actions in their mind using the same neural network pathways as actual perception, AKA imagination.

"4 Experimental Results
The implemented system currently runs on a
2.5 GHz Pentium 4 machine"

[Bolding added]

Really?

That first paper (p26) was well worded and written, theoretically fine, but totally lacked any sort of in depth analysis for me to comment on. They didn't even include any of the source code they used for the 'robot' in question for me to analyse. Again all I see with these artificial neural network based algorithms are people trying to model real biological neural networks with abstract models of information processing, which although may prove useful in a scientific sense still lack any sort of consciousness, any more than a cleverly coded super computer does.

 

[Zeuzzz]

There's not much point pursuing this line of reasoning. You could work him down to string theory and he could still argue there's demons plucking them.

Why do you say demons? I never mentioned anything even remotely relevant to that reference.

If you want to discuss the problems with string theory that's fine, but try to keep it to a relevant thread and not hijack this one. Or just buy this http://www.amazon.com/The-Trouble-With-Physics-Science/dp/0618551050

 

[quarky]

Its kind of cool how threads about consciousness are so easily hi-jacked and de-railed.
Its pretty hard to be off-topic.

Which is why I'd like to see kittens.

 

[blobru]

...from a recent Kitten Symposium on Unconsciousness:

 

[Beelzebuddy]

Why do you say demons? I never mentioned anything even remotely relevant to that reference.

You never mentioned anything at all. But since an empirical viewpoint would lead one to conclude that consciousness (for most definitions of this terrible, horrible term) is a product of the brain's activity, and not the other way around, you must have something in mind.

 

[quarky]

You never mentioned anything at all. But since an empirical viewpoint would lead one to conclude that consciousness (for most definitions of this terrible, horrible term) is a product of the brain's activity, and not the other way around, you must have something in mind.

funny about language...

What do you mean by "...in mind."?

Its hard to discuss a background sort of consciousness without sounding all wooed-out or religious fundamentalist. I'm neither; more of a fun, mental sort...

I'd try to give a go at it, though its a very strenuous hypothesis, and past attempts have garnered some mockery and nastiness.

But, I'm still here. My 'single quark" hypothesis is exhausting to my brain.
Maybe I'll give it a shot, after chores.

 

[rocketdodger]

"4 Experimental Results
The implemented system currently runs on a
2.5 GHz Pentium 4 machine"

[Bolding added]

Really?

That first paper (p26) was well worded and written, theoretically fine, but totally lacked any sort of in depth analysis for me to comment on. They didn't even include any of the source code they used for the 'robot' in question for me to analyse. Again all I see with these artificial neural network based algorithms are people trying to model real biological neural networks with abstract models of information processing, which although may prove useful in a scientific sense still lack any sort of consciousness, any more than a cleverly coded super computer does.

Why do you think this? Let me phrase the question in another way:

Why do you think the causal sequences of node activation in an artificial neural network different than the causal sequences of node activation in a biological neural network?

The essential property of a neural network is that one neuron's output leads to a change in the behavior of neurons downstream. If a conscious behavior arises due to the way a network functions, what difference does it make how or where that network is implemented?

As for the first presentation, you don't need source code. It wouldn't make sense even if you saw it, because there is no specific programming done that is relevant to the robot. That isn't how neural networks work. They trained the robot so that its goal is focusing on blue, and they trained the robot that when it turned one way it saw blue, and not blue if it turned the other way. They did *not* tell the robot to turn -- ever.

What the robot did was imagine the act of turning in either direction, and imagining a left turn led to the imagination of a blue percept, which caused the robot to then *want* to turn left -- it effectively decided to turn left because it imagined that if it turned left it would see blue, and seeing blue is its goal. And this was all done with trained neural networks -- no hard coding of any behavior.

This is true imagination, by any possible definition of the term. Genuine, real, authentic imagination. And imagination is one of the behaviors we normally attribute to consciousness. Yeah the robot can't write poetry, or even play Jeopardy as well as Watson, but Watson doesn't *imagine* things like we do. This robot did.

 

[quarky]

...from a recent Kitten Symposium on Unconsciousness:

[qimg]http://i298.photobucket.com/albums/mm270/blobru/Decorated%20images/kittens-sleeping.jpg[/qimg]

You are a sick puppy.
That was over the top of the cuteness charts, and I must now wash off the gay.

 

[Mr. Scott]

I was just thinking about how the first electronic computer, the ENIAC, was developed to feel the future, and it did so successfully.

It was developed to calculate artillery trajectories. In other words, given a specific weight and size of bullet, amount of gunpowder, the gun angle, wind speed and angle, air temperature and humidity, where will the bullet land?

Feeling the future is a fundamental application of mathematics and computers. Great things are happening as computers do it more and more like our brains do.

 

[Mr. Scott]

Oh dear you missed the whole point but confirmed the inspiration for the video.
The video was not about copying the successful outlier's history mechanically and unconsciously, but about living consciously through creativity in the now in an unpredictable world.
You falsely assume that because Jobs was successful his advice had something to do with it. No, his advice was not about how to be successful but how to be fulfilled.

Success at fulfillment perhaps?

"Follow your heart (gut, feelings, do what you love, etc.)" is too often bad advice for being successful in fulfillment or any other endeavor, because the "heart" (emotional part of the brain that purportedly feels the future) evolved through chaotic evolutionary processes only guaranteed, in the past, to have endowed enough success to proliferation of the genes responsible.

 

[rocketdodger]

This is pretty interesting:

http://www.iis.ee.ic.ac.uk/yiannis/DemirisJohnson03.pdf

Although this is almost 10 years old, it talks about a very important mechanism in biological brains -- re-using the same circuitry for both action and planning, and in some cases observation and learning ( both of those are supersets of planning, though ).

The basic idea is that the circuitry of the motor cortex is used not only for controlling muscles and decoding muscle position, but also for simulating the control of muscles and the effects of that control, I.E. imagining movement. And furthermore, that the imagining of movement is used during learning I.E. "if I do this, and my arm then moves up like so, I will be in the right position."

In this case the researchers use a sequence something like this:

1) robot observes goal configuration of arm, on another robot
2) the code modules that plan the robots arm movement are re-routed to internal locations ( they no longer control the arm, rather their output goes back into the robot brain )
3) those modules then control simulations of the arm I.E. if one would "raise" the arm the arm isn't actually "raised" yet portions of the robot brain are activated as if it was ( for example, imagine raising your arm -- you can also imagine what your arm feels like in the raised position )
4) the results of those simulations are evaluated to see if any of them bring the arm closer to the goal configuration
5) the simulated movements that are rated the best are reinforced, and more iterations of imagination are performed
6) eventually a sequence of movements that the robot imagined would put it in the goal configuration is found, and the arm control modules are re-routed back to the real arms
7) the action is performed

This seems very convoluted, but it is important to realize that this is the exact mechanism by which animals not only plan movements but also learn movements from observing others. In this case a neural network was not used, but the high level information flow is nevertheless the same ( or at least very similar ).

In a previous post I linked to some research that is like this but in that other case they actually *did* use neural networks.

Who said we don't know much about consciousness?

 

[Mr. Scott]

This is pretty interesting:

http://www.iis.ee.ic.ac.uk/yiannis/DemirisJohnson03.pdf

Although this is almost 10 years old, it talks about a very important mechanism in biological brains -- re-using the same circuitry for both action and planning, and in some cases observation and learning ( both of those are supersets of planning, though ).

The basic idea is that the circuitry of the motor cortex is used not only for controlling muscles and decoding muscle position, but also for simulating the control of muscles and the effects of that control, I.E. imagining movement. And furthermore, that the imagining of movement is used during learning I.E. "if I do this, and my arm then moves up like so, I will be in the right position."

In this case the researchers use a sequence something like this:

1) robot observes goal configuration of arm, on another robot
2) the code modules that plan the robots arm movement are re-routed to internal locations ( they no longer control the arm, rather their output goes back into the robot brain )
3) those modules then control simulations of the arm I.E. if one would "raise" the arm the arm isn't actually "raised" yet portions of the robot brain are activated as if it was ( for example, imagine raising your arm -- you can also imagine what your arm feels like in the raised position )
4) the results of those simulations are evaluated to see if any of them bring the arm closer to the goal configuration
5) the simulated movements that are rated the best are reinforced, and more iterations of imagination are performed
6) eventually a sequence of movements that the robot imagined would put it in the goal configuration is found, and the arm control modules are re-routed back to the real arms
7) the action is performed

This seems very convoluted, but it is important to realize that this is the exact mechanism by which animals not only plan movements but also learn movements from observing others. In this case a neural network was not used, but the high level information flow is nevertheless the same ( or at least very similar ).

In a previous post I linked to some research that is like this but in that other case they actually *did* use neural networks.

Who said we don't know much about consciousness?

Yes, we know a LOT about consciousness already, and what you're describing is a type of "feeling the future."

What's cool is that when we rehearse movements in our minds, our minds are actually moving the limbs, but inhibitory impulses prevent the muscles from physically moving. When I think really deeply about playing piano, sometimes my fingers come to life and start to weakly play the notes in the air. Maybe it's because the inhibitory neurons become exhausted.

 

[Mr. Scott]

The Chinese Room Thought Experiment

When I first heard this thought experiment I was really intrigued. It seemed persuasive and made the hard problem of consciousness very tangible.

Now, I find the Chinese Room idea stupid. Searle is a smart guy, so why does he (and so many others) find it so compelling? Would someone explain to me why it's important or persuasive?

Chinese Room on Wiki

Video demo of the Chinese Room starts at 16:45 in this cool BBC program, "The Hunt for AI."

 

[rocketdodger]

Yes, we know a LOT about consciousness already, and what you're describing is a type of "feeling the future."

What's cool is that when we rehearse movements in our minds, our minds are actually moving the limbs, but inhibitory impulses prevent the muscles from physically moving. When I think really deeply about playing piano, sometimes my fingers come to life and start to weakly play the notes in the air. Maybe it's because the inhibitory neurons become exhausted.

Yeah I have known about that first pathway for awhile. What I realized just recently, which is mentioned in the research, is the idea that not only are the outgoing motor impulses inhibited, but they also lead to the same downstream effects as incoming sensory percepts.

Apparently the motor networks are always recurrent and there is always a model of the results being generated from any outgoing movement signals, we just don't notice it because usually the real thing happens and the sensory percepts from actually moving a limb trump those from imagining moving a limb. Only when the real thing is inhibited do the simulated results become apparent.

This also nicely explains why a deviation from the expected is such an attention-getter for a conscious animal -- if the results of the model and the results of reality don't match up it would be trivial for a network to see it, especially since both results will arrive in the same location at approximately the same time.

Fascinating.

I wonder if we can start a sticky thread about "consciousness: the facts"

 

[rocketdodger]

Would someone explain to me why it's important or persuasive?

No, because it isn't.

Searle formulated it in just about the stupidest way possible. He did that on purpose. He doesn't want people actually thinking about the issue, he wants them to be blinded with emotion and just give up.

Case in point -- why Chinese and not English? Why a man in the room, and not a robot? Why a room, and not the brain of a giant?

The whole thing is absurd.

 

[Beelzebuddy]

Now, I find the Chinese Room idea stupid. Searle is a smart guy, so why does he (and so many others) find it so compelling? Would someone explain to me why it's important or persuasive?

Natural language translation is a hard problem. We're talking "hard" with a capital Nobel.

But like many hard problems, it's theoretically possible to brute force it. To just make a giant-ass lookup table covering every possible circumstance. That's all Google does today, really. You'd be surprised how few unique phrases there actually are.

Incidentally, computer scientists around the time of this argument (1980) were all really excited about the possibility of making giant-ass lookup tables for absolutely everything. They called it "expert systems."

These computer scientists argued that a computer armed with enough of these lookup tables was intelligent. Not "indistinguishable from," not "might as well be considered," was. A computer with a sufficiently large Chinese-English dictionary would know how to translate between them.

But hold on, Searle said. Let's give this giant-ass lookup table to some jackass in a room instead. He don't know Chinese. He ain't gonna learn Chinese, not when he just looks up sentence indexes. He doesn't understand what you're asking him. Look at him, he gets paid to sit in a dark room and do whatever was the 1980 equivalent of filling out captchas all day.

So whatever we're looking for with this whole "intelligence" thing, whatever Derpy McBlackbox over there has that my pocket calculator don't, the dictionary alone doesn't have it either. Moreover, this is a general problem. Just because you have a big enough index to answer every question doesn't mean you can call it "thinking."