what makes this robot different from any other animitronic robot?

[GabrielsThoughts]

this puzzles me, I see it , and I know it  supposed to be this fascinating new innovation, but I see no real difference between the Actroid and any other animitronic  figure at disneyworld.


http://www.youtube.com/watch?v=lJZTWwy6eUw

http://www.youtube.com/watch?v=MY8-sJS0W1I

[techmaster-glitch]

Well, obviously, it is more advanced than a theme park animatronic. It looks more lifelike, has fluid movements, and is (supposedly) able to carry on conversations with people.

Though, I severely look down on most robot anthropomorphization attempts (that doesn't mean I can't appreciate good tech when I see it). Take this from a robot geek, I don't want robots to look like humans. It's just not right to me. Their bodies should always be optimized for their functions. If a robot's single purpose is human interaction (like animatronics), I'll just meh. But if you go down the route of that movie I, Robot (the one with Will Smith. It was a good movie, by the way, but I am really tired of the machines ALWAYS being portrayed as the bad guys), that's where I get peturbed. All the robots are humanoid (save for the big demo-bot), and that bothers me. I've been told that the reason some people want their robots looking like humans is so that the humans are more comfortable around them. I most certainly am not.
The Freefall comic, on the other hand, presents to me what I think is a quite accurate depiction of robots in the far future. Few are actually "humanoid", but the big thing is that the general behavior patterns of the robots are what I would have expected from robots endowed with true sentience.

[Zedd]

http://www.youtube.com/watch?v=CyIHzCsbA_w

http://www.youtube.com/watch?v=iaRAf1PLRjo

KILL IT WITH FIRE!!!

Its trying to speak!!!

[Fuyudenki]

the first one's a few months old.  I was unimpressed then, and I'm unimpressed now, it basically is just like the automations at Disney World.

Zedd's robot, however, is new to me, and I wish I understood what they were saying about it.  Looks somewhat interesting.

[Valynth]

To create a robot like us we must first know what we are.  That alone makes sentient robots something of the VERY, VERY far future.  Then we just have to use electronic analogs to simulate that.  That's along the lines of translating War and Peace into grammatically correct Russian, Japanese, and German 5 times a second in terms of difficulty.  And that's not including all the pressure sensors and regulatory functions robots would have to process before it even moves.

In short it will probably take hundreds of years to know how we know and then three times that long to get computers to know in the way we know.  Savy?

[llearch n'n'daCorna]

I disagree, Valley. Oh, not about the end result in kind - just in degree. Making sentient robots is, indeed, complex. However, I don't see it taking that long to make them - not by half.

Replicating something you understand in code is relatively easy. Replicating something you do -not- understand, on the other hand, is not. Electronic analogs to simulate human thought patterns is nigh-on impossible... UNTIL we understand those patterns. Once we do, it's comparatively easy to feed that into a computer.

Getting the computer down to a size that you can fit it into a human form is simply a matter of time - and not really that much time, either. Moore's Law is still in effect. Same goes for all the other bits and pieces we put together - right now, we could probably build a human robot (you get my meaning) but not at human size - and the size we'd be forced to build it at, structural integrity would be an issue; people don't grow to house sizes because their bones would end up thicker than their arms, etc etc...

I'm not sure we'll see "human robots" by the time I die. I -am- pretty sure that we'll see at least the beginnings by the time my son dies. Unless, of course, the human race collapses under it's own weight in the meantime...

[Valynth]

llearch, I don't think you're grasping the scope of how complex the mere process for transmitting data from one neuron to another is.  For one thing a nerve doesn't simply go on and off like a computer processor (all coding is simply an evolved form of binary that is easier for humans to understand), it has three "degrees" of on (based simply on the three chemicals found when a nerve fires) and we have yet to truely map out exactly what those chemicals tell the nerve or muscle cells to do.  Yes we can get a general idea of what they are telling the cells to do, but we don't understand -precisely- how they do it or the exact movements they entail.  In otherwords unless we can make a processor and learn to code it with three degrees of "on" with the degree "off" it could be awhile till we create a being who can think in analog and therefore be considered truely "sentient" as analog thinking allows us to be creative on our own rather than having to be told everything by an outside source.

For example a camera will only tell you what it can see, if you put it in a box it can no longer see the room and therefore the room does not exist anymore to the camera since it is pure logic (I do not see it therefore it does not exist).  An analog thinker, however, knows that the room still exists even though he cannot see it and has not been told to assume it still exists by its programmer.

Frankly I say screw making sentient robots since they'll just be humans with mechanical parts (which entails problems all its own) and screwing with biology to get better results is by far easier since we can just skip the coding.

[techmaster-glitch]

I agree with llearch. The other thing that is not being taken into account here is how fast we are getting closer to understanding our own brains. Every year, we take a leap closer to said goal. And I think we're closer than you think, because I think it can't be that difficult for nerotransmitters to transmit, therefore it's not that hard to understand. Y'wanna know why i think that? A few years ago, some scientists took a single rat neuron brain cell, a microchip of about the same size, and fused them together. When I say fused, I mean that they were actually exchanging electrical impulses with each other! I know that's not almighty proof that we're already there, I'm saying that it is a big step in the very right direction. If nerotransmitters really are as comlicated as you say, then by that logic, it should never have been compatible wth something as simple as a microchip.

Anyways, I'm just saying that I have high hopes for artificial sentience and intelligence. It will happen. Maybe not in the next century, or two centuries, or three, but it will happen someday. As for humanoid robots, I give a big fat 'no' to all that crap.

[Valynth]

I'm not saying it's impossible to register the electrical pulses, I'm saying that that is only a tiny sliver of the iceberg.  If registering electrical pulses was all there was to a neuron, we'd already have perfect prosthetics.  As it stands we only have about 3-4 degrees of movement registered for arm/hand prosthetics and the newer ones have about 4-5 and they're far from completely accurate or fluid.  What we need for a completely operable arm/hand is about 25-27 degrees of movement.

Whats the major hold up you ask?  It certainly isn't the mechanics of the hand, we've already mapped that.  It's translating the nerve's pulses into a form the computer understands and can then translate the command into physical action.

You also have to account for the fact that the arms/limbs are in a completely logical setting, in other words the individual cells that make the arm only know what the brain tells them.  The arm doesn't feel pain, the BRAIN does and the logical step the brain takes is to halt the pain by moving the arm away from the pain.  We, however, possess an ability to enter painful situations against this logic so there must clearly be something illogical that takes place in our perceptions.

[llearch n'n'daCorna]

llearch, I don't think you're grasping the scope of how complex the mere process for transmitting data from one neuron to another is.  For one thing a nerve doesn't simply go on and off like a computer processor (all coding is simply an evolved form of binary that is easier for humans to understand), it has three "degrees" of on (based simply on the three chemicals found when a nerve fires) and we have yet to truely map out exactly what those chemicals tell the nerve or muscle cells to do.  Yes we can get a general idea of what they are telling the cells to do, but we don't understand -precisely- how they do it or the exact movements they entail.  In otherwords unless we can make a processor and learn to code it with three degrees of "on" with the degree "off" it could be awhile till we create a being who can think in analog and therefore be considered truely "sentient" as analog thinking allows us to be creative on our own rather than having to be told everything by an outside source.

Heh. The problem is, you're presuming too much. ;-]

Rather than trying to make a brain that will fit into the standard "human" infrastructure, is it not more intelligent to make a brain that fits into the robot architecture? Making a trinary brain isn't -entirely- impossible now. Making it as complex as people, well, that's much harder.

I reiterate - you're looking at people at too low a level. And you're still in a "we don't understand" state. While you don't know precisely what the problem is, you're unable to effectively plan a solution - and that's a general you, I'm in precisely the same place. Well.. perhaps not -exactly-.

I suspect Alondro might have more idea than either of us - for now, I'm shooting the breeze based on hearsay, rather than exact detail. However, notwithstanding that, we -can- do the physical movements of the brain fairly easily.

Which is why we have people working on Turing-capable computer brains. And, since there's no real -immediately- useful result out of such a thing (as distinct from long-term benefits) there's nobody pushing to get it solved. It's within our reach, though.

Heck. In the last ten years, computer power has become -insanely- compact. How much more compact does it have to be before we can build a computer that has enough power, and house it in a room? There's no -immediate- need to write one that works on tiny hardware, and if you work on rooms full of kit, you can expect Moore's Law to solve your miniaturisation issues over time. Given that, the task will get solved.


Honestly, while I agree about some of the complexity, I think you're looking at the stiff end of the task, and not the way it will eventually be accomplished. While trinary computing is doable, why not stick with binary hardware, and implement trinary in software? It's just as easy, you get to use commodity hardware, which means much much cheaper, and the standards are all the same. All you need to do is fiddle the overhead, and we've been doing that for 30+ years.


(Do bear in mind that I'm not criticising, and am, in fact, enjoying the conversation, even if we -do- disagree

Edit:
Almost forgot:

Frankly I say screw making sentient robots since they'll just be humans with mechanical parts (which entails problems all its own) and screwing with biology to get better results is by far easier since we can just skip the coding.

Heh. You and me both, boyo. In fact, Asimov had that in The Bicentennial Man (the book, not the movie, which, although good, didn't have all the depth of the book.) - mention of the fact that US Robotics (Asimov's version, not RL) moved from human-shaped robots to industrial types, purely in reaction to Andrew - a subtle move, but it fits with how your average person reacts to people-shaped devices. If it's human shaped, behaves human, etc, we get upset when it gets turned off. Not to mention fear, and monkey hits out at things that scare him...

[Zedd]

Honestly for any unmonitored reason..We dont need robots...We'll become really fat and lazy due to having they doing EVERYthing for us what we can normaly do on our own. So in end..If there was it be any synthoid walkign the streets...I'd pop it in the head with an AK-47.

[TheGreyRonin]

Honestly for any unmonitored reason..We dont need robots...We'll become really fat and lazy due to having they doing EVERYthing for us what we can normaly do on our own.

Technology already does this. How many Americans are overweight? How many Americans actually get off their...rumps...and work each day, instead of sitting in an office pushing buttons?

I'm all for technology, but like most things humans have created it has taken over our lives instead of simply enhancing them.

As for sentient robots in humanoid form, I don't think it's a good idea. But consider this: Instead of building an onboard sentient processor, fill a building with one, then link it via your choice of high-speed communications link. (Wi-Fi would work well...)

There's also a large difference between "humanoid" and "human" in function. One could argue that a mannequin hooked into a sentient computer network was "humanoid", but wouldn't be able to even move. I agree that we are a ways off from duplicating human abilities fully with technology, but in general most of a human's range of motion and capabilities are rarely, if ever used in daily life.

The full understanding of organic functions is only necessary if you are trying to build a robot that has the exact functions of a human being. Since this theoretical construct would have no need to breath, sweat, excrete, etc. otherwise, I think it is fair to say that we could indeed have a humanoid robot very soon if enough resources were applied, as long as we were simply trying to mimic, not duplicate full human function.

There's much easier ways to do this. The most popular requires two people and roughly nine months. *grins*

[rt]

Took a quick look, not very impressed initially at either. For realism I was more impressed with the anamatronic monkey head geek gift that went around a few christmases ago.

[Fuyudenki]

I would just like to punch a hole in Valynth's argument by saying that even though a single bit operation is either on or off, by simply adding a second bit, you get the exact same functionality you mentioned for human neurons: 4 possible states.  That's one off, and 3 different ons.(00, 01, 10, 11)

Actually, that's why I'm practicing counting on my fingers in binary, instead of the usual way.  Plus: you can count to 31 on one hand, and 2047 on two hands.  Downside: it takes a little thought at first, and 4 and 20 look like you're flipping people off.

you don't exactly want to say there are *middle finger* this many people at the table, Fry.

[edit]it's 20.  Totally 20, you saw no 12s here![/edit]

[superluser]

Actually, that's why I'm practicing counting on my fingers in binary, instead of the usual way.  Plus: you can count to 31 on one hand, and 1023 on two hands.  Downside: it takes a little thought at first, and 4 and 12 look like you're flipping people off.

Huh?  Why would people think you're flipping them off if you show them your ring finger (or your ring finger and pinky)?  By my reckoning, 2 and 18 are the numbers you need to stay away from.  Then again, my system has a bit of a NUXI problem.

[Fuyudenki]

the thumb is used for counting, as well.  If you don't use your thumb, you can only count to 15 on one hand, and 255 on both.

4=2^2.  Counting from either pinky or thumb, 0, 1, 2.  2 is the middle finger.  Also, I use the pinky as my LSB when counting on one hand, or my right pinky as the LSB, and the left pinky as my MSB, when counting on two hands, so you git middle finger, or middle finger plus thumb, both of which are flipping people off.

[edit]what are the odds that two identical black cats would walk past that door?[edit]

[Tezkat]

Hmm... took a look at the vids...

I'd say that the difference is that the Disney animatronics are on their way down the uncanny valley while these new ones are climbing up the other side. Interestingly, I think the robot worked much better in Japanese. The stiff movements and flat affect mirror the sort of body language one commonly sees in Japan, but they look less natural from a Western cultural perspective.

llearch, I don't think you're grasping the scope of how complex the mere process for transmitting data from one neuron to another is.  For one thing a nerve doesn't simply go on and off like a computer processor (all coding is simply an evolved form of binary that is easier for humans to understand), it has three "degrees" of on (based simply on the three chemicals found when a nerve fires) and we have yet to truely map out exactly what those chemicals tell the nerve or muscle cells to do.  Yes we can get a general idea of what they are telling the cells to do, but we don't understand -precisely- how they do it or the exact movements they entail.  In otherwords unless we can make a processor and learn to code it with three degrees of "on" with the degree "off" it could be awhile till we create a being who can think in analog and therefore be considered truely "sentient" as analog thinking allows us to be creative on our own rather than having to be told everything by an outside source.

Whaaa? "Three degrees of on"? Neurons either fire or don't fire, depending on whether or not the membrane potential crosses a certain threshold voltage. They're binary in that sense. Of course, since we're talking about a voltage, "off" represents not a null value but a distance from the threshold, mediated by ion concentrations on either side of the membrane. (Presumably that's what you meant by "three chemicals"--or were you referring to neurotransmitters and the rest of the chemical messenger system?) It's still something that can be quantified to whatever degree of precision you choose.

The functionality of neurons in isolation is fairly well understood, and we now have pretty good computer simulations. Indeed, we can already model ion channel activity. What we lack is the computational capacity to scale those models up into large, multi-neuron systems--complexity grows exponentially with the size of the network.

We're still decades away from machines with the computational capacity to simulate a human brain. However, unless you imagine some kind of barrier at which computers would cease to progress, getting there is just a matter of time (see Moore's Law). Granted, so much of the human experience depends on the rest of our bodies that modelling that as well might be necessary for such a digital brain to function. Such brute force approaches to artificial humanity would be more the province of people seeking immortality than those interested in thinking machines.

That's not to say it isn't a useful avenue of study. The resolution of nuclear imaging has made amazing progress in the past decade, allowing us to examine ever more detailed areas of the live brain. Combining that with higher level approaches that study human behaviour can lead to an understanding of our brains at the algorithmic level, which is far easier to translate into a digital medium. There's a lot of that sort of work going on in fields like visual processing. The human brain is very good at processing sensory information, so figuring out how we do it has great practical value.

There's no overwhelming reason to do things the human way all the time, however. The state of the art in mathematical cognition, for instance, currently involves studying how... say... humans add single-digit numbers together. We're a ways off from implementing a computer that does math the way people do. But honestly... why the heck would we even want to? Computers are very, very good at doing that sort of thing their own way.


Of course, all that leads into questions that are more the domain of philosophy than engineering. What constitutes "sentience"? When do we consider another being capable of "thinking"? Philosophers have been attacking that problem for millenia without all that much progress. Engineers will make it matter in within decades. Given the recent progress in AI and cognitive science, I doubt that Turing Test capable machines are as far off as some people believe.

For example a camera will only tell you what it can see, if you put it in a box it can no longer see the room and therefore the room does not exist anymore to the camera since it is pure logic (I do not see it therefore it does not exist).  An analog thinker, however, knows that the room still exists even though he cannot see it and has not been told to assume it still exists by its programmer.

That makes no sense. By your argument, anyone unable to recognize object permanence (which, incidentally, includes very young humans and most other mammals) does not think in analog (whatever that means). And how is a camera pure logic?


Your "having to be told everything by an outside source" suggests an impression of AI based on the very formal, rule-based systems that, frankly, went out of style back in the 1970's yet. Modern AI has largely eschewed that approach in favour of systems that learn the "rules" themselves based on examples--not terribly unlike the way humans do. If you examine software designed to learn the rules of English grammar, for example, you'll find the kinds of mistakes they make very similar to the generalization errors made by young children. They've actually taught us something about how we learn language ourselves. Our current AIs are limited by immaturity, not impossibility.

[superluser]

Also, I use the pinky as my LSB when counting on one hand, or my right pinky as the LSB, and the left pinky as my MSB, when counting on two hands, so you git middle finger, or middle finger plus thumb, both of which are flipping people off.

If I'm understanding you correctly, the forbidden numbers should be 4 and 20 (the thumb represents 16, right?).

Anyways, if you're using this to count, make your index finger the LSB.  That one has the most manual dexterity, and you'll have to switch it every time.  If you use the pinky for 8, you'll have to switch it only one eighth the time.

(I have to count one-handed, so I worked this out years ago)

[Fuyudenki]

I use my feet to turn on my computer, log in, and put CDs in the drive.  Manual dexterity isn't a problem for me.

and yes, if you'll notice, I corrected my posts.  It's 4 and 20 to worry about.

So how long have you been counting in binary?

[superluser]

So how long have you been counting in binary?

I started working it out in 6th grade.  I don't think I started counting in actual binary until somewhere around 9th grade.

[Madmann135]

Robots imitating humans is so WRONG.  If you want to have a humanoid robot have a robot that has a humanoid frame but that's it.  Those machines need to be dealt with by using a compactor (or removing their skin). 
When you start making robots look human you are crossing into such wrong territories.  The Will Smith i,Robot's were OK because you could easily tell what was human and what was machine.

[Fuyudenki]

So how long have you been counting in binary?

I started working it out in 6th grade.  I don't think I started counting in actual binary until somewhere around 9th grade.

Pretty cool.  I've been doing it for about 6 months, now.

Madmann, what do you have against humanoid robots?  It's not like they're killing real humans to make them.  Do you have something against Final Fantasy movies, too, because the characters are made too realistic?(personally, I think the Advent Children characters have mostly crossed to the far side of the Uncanny Valley.  Quite a feat, indeed!)