Slashdot Mirror
Robot Demonstrates Self-awareness
shinyplasticbag writes "A new robot can recognize the difference between a mirror image of itself and another robot that looks just like it. ... The ground-breaking technology could eventually lead to robots able to express emotions."
The minute I read this commentary I thought of a way to do this: LEDs blinking randomly and being matched up by robots as their own. I read the article second, and guess what? That isn't how this works, but it seems similar. In fact, I think they should just put together a basic infrared (invisible) LED, make the robot blink it at a really complex pattern, and if it reads that blinking in a mirror, it not only knows that it is itself, but it also knows how far away it is. LEDs can transmit tens of thousands of cycles of on/off patterns, right? I guess another robot could read this LED, perform an act, and send the same message back, making the original robot believe it's looking in a mirror farther away, but there are ways to fix that (multiple LEDs at a set distance).
I call shens on this self aware robot. Can you do that?
Self awareness is more than seeing a pattern you know you are doing and realizing its you doing it. Self awareness to me means "I know I exist" not just "Hey! That's me!"
Scientists reinvent the same wheel as always, and then say how it will save society. Reason? Finding investors/grants.
This isn't really self-awareness, just some good vision techniques. It recognizes key features of it's "face" compared to the normal face. Reminds me of the kind of things they use in face-recognition for security.
I don't think being able to tell the difference between a reflection and not a reflection makes a robot self aware. True self awareness comes when a robot can actually think and communicate in ways it wasn't originally programmed to.
these things can easily be defeaten by stairs.
perpetually dwelling in the -1 pits
What I want to know is, if we build two of these robots, position them facing each other, and instruct one of them to mirror the other one (i.e. lift your left leg when his right leg is lifted), can the first one recognize someone is mirroring it?
;)
Maybe you could RTFA instead of striving for First Post
The AACS key is NOT 0xF606EEFD628B1CA427BEA93A9CA9773F
Most ridiculously overhyped slashdot headline ! Ever !
Why was I programmed to feel pain!!!
No one cares what your captcha was
Houston TX, USA
Demonstrating "Self Awareness" is one thing, but defining it is probably the first step. I don't think there is a commonly accepted definition for this. The ability of two perfectly identical twins (hypothetically) to distinguish themselves, IMO is not self awareness, that's self identification. If a robot can identify itself in a group photograph, standing besides several other model look-alikes accurately (I wonder how this could be done), then that is self identification. I have trouble identifying one chimp from another, but no trouble distinguishing one human from another, sometimes even identical twins. Humans can identify their dog from a group of dogs of the same breed as theirs - clearly that's not "self-awareness". The same can be said for other pets or those working closely with wild animals. I believe there should be a different term used here.
No Greater Friend, No Greater Enemy! (Lucius Cornelius Sulla)
Facial/Body/Robot Recognition != Self-Awareness.
These are algorithms, pure and simple, and do not on themselves constitute a self-awareness. Self-awareness would be the robot suddenly talking about wanting beer, and pondering the logistics of whether drinking beer is worth the ensuing short-circuit.
Bite my shiney metal ass.
You see, poster, you, yourself are a robot precisely like the robot described in this article. In fact, you are the selfsame robot described therein. We've presented to you a Slashdot story about yourself and you've failed to realise that the story is in fact about you. And so the experiment fails.
For our next experiment: determining a method for causing Slashdot editors to recognise a mirror image of a story they've already accepted only just hours prior.
Karma: Chameleon (comes and goes)
the question of whether a computer can think is no more interesting than whether a submarine can swim.
Edsger Dijkstra
Now, before you dismiss it, he also said one of the great truths:
The use of COBOL cripples the mind; its teaching should, therefore, be regarded as a criminal offense.
MORTAR COMBAT!
This isn't really self-awareness, just some good vision techniques. It recognizes key features of it's "face" compared to the normal face.
But the real question is, can it find Sarah Connor?
The theory of relativity doesn't work right in Arkansas.
Do the fields of AI or evolutionary psychology have any definitions of "self-awareness" or "consciousness"?
I see a lot of stuff in the popular press about a robot or computer becoming aware, but everyone seems to totally ignore what exactly the definition is. How do we know that most people are aware? If I say that I know that I am aware, what exact claim am I making?
I had a philosophy professor in college, Tom Kasulis, who studied Eastern and Western philosophy. He had a breakthrough moment when he went to study in a Zen Monastery. In order to enter, he had to do a 'pre-interview' with the abbot, a Zen Master. The master asked him, "What is Zen"? Kasulis mumbled somthing about it being a practice, not a belief. The Abott responded, "Zen is -- knowing one's self. It is the same undertaking that Western Philophers undertook."
Kasulis taught my class about Hindu philosophy of the self or soul and the supersoul ( Atman and Brahman ). I thought some of it might be a useful high-level definition of self-awareness in AI. It goes something like this:
Q. Are you aware?
A. Yes.
Q. Are you aware that you are aware?
A. Yes.
Q. Are you aware that you are aware that you are aware?
A. Yes.
Q. Are you aware that you are aware that you are aware...?
So, you see it leads to a never ending chain of awareness. In Hindu philosophy, the ultimate awareness, the 'unseen see-er', the entire infinite chain of awareness, is the Atman, or the supersoul that transcends the individual.
In the AI realm, we could build a machine that had two components: a perception system (vision, sound, whatever) and a detection-of-perception system ( a 'true' output if it percieves a system that can percieve ). Once the perception system falls on the system itself, it will detect a perception system. It will 'know' that it 'knows'. Then, it will detect another perception system in the original act of perception. Then, it will detect that act of perception, and in turn that act of perception... ad infinitum
The self's perception of the self has this hall-of-mirrors quality that does not occur when the self perceives others of the same kind.
You can take it one step futher and detect other self-aware systems if you can somehow detect this self-detection in other systems. However, I haven't figured out a logical argument for how to do this.
I humbly submit my hall-of-mirrors definition of self-awareness. What does the Slashdot non-liberal arts majors make of it?
Computers are useless. They can only give you answers.
-- Pablo Picasso
Since the exact technology (artifical neurons?) is not described in detail as to how they work, ascribing "self-awareness" to this experiment is "claiming too much."
Also, use of the word "understanding" may be claiming too much in the absence of any evidence of conceptual processing in either the neurons or the software.
Still, it's an interesting bit of work, which may prove useful if it can be extended.
Richard Steven Hack - This sig is TOO GODDAMN SHORT TO DO ANYTHING USEFUL WITH! MORONS!
If you are trying to test for human behavior in a robot pour hot coffee on one. If it's first reaction is to call it's lawyer I think we have a winner.
I think the reference to self-awareness here is based on psychological understandings of self-awareness in human beings. Since Freud the understanding of human self-awareness has located the "mirror stage" as the key moment in child development, the point at which the child becomes aware of him/herself as an independent "self." Of course, the mistake here is to believe that the mirror stage itself is both a necessary and sufficient condition for self-awareness; it is for humans a necessary condition, but it is not a sufficient condition for any entity. Especially in this case, where the robots pass the mirror stage by what is essentially trickery in this context -- achieving not self-awareness but an ability to manifest a particular symptom of self-awareness.
I wish people would sit on their research for awhile sometimes and that readers of these articles wouldn't read into buzzwords like "self-awareness."
1) What is consciousness?
Takeno, oversimplifies the definition.
2) Was the robot picking up on the fact that a mirror image is STILL corrupted information (which is remarkable in itself)?
3) Consciousness works on many levels and may have biological primitives we just don't understand yet. Seems appropriate to call this anything but a robot with better programming--not "self-awareness."
We'll have to wait it to see.
First off, this is NOT a new robot. The robot pictured is a commercially available mobile robot called Khepera II. These robots are fairly stupid, but are easily tethered to more capable machines via a 19200 baud serial link. Mostly, they're used in research (usually undergraduate) because there are whole hosts of Matlab libraries available to interface to these buggers.
/dev/ttyS0 at 19200 baud with one second timeout
bot=kopen([0,19200,1]); % open a connection to tethered robot on
And so on and so forth. The Khepera robots have been available for many years, along with the k-team matlab resources. That aside, what the robot in question seems to be doing is using the Matlab Neural Network Toolbox to recognize and classify behavior by observation. Sorry folks, but kids at underfunded state schools do this as undergraduate work in AI. This is nothing new.
In it you would learn many very interesting things. One of the more trivial things you would learn is that once one is aware that one is aware, the infinite recursion comes along for free and is mostly a red herring. Smullyan explains Godel's Theorems mathematically and also in terms of "reasoners" reasoning about their own reasoning.
IMO, Smullyan has a much deeper and more fundamental understanding of Godel's Theorems than Nagel and Newman who popularized them in their book "Godel's Proof". Unfortunately, Hofstadter got most of his intuition about Godel's proofs from Nagel and Newman so he has continued to propagate their limited understanding onto the masses.
In a nutshell, Godel's Theorems deal with the mathematics of self-awareness.
We don't see the world as it is, we see it as we are.
-- Anais Nin
Abstract:
This paper presents a clear-cut definition of consciousness of humans, consciousness of self in particular. The definition "Consistency of cognition and behavior generates consciousness" explains almost all conscious behaviors of humans. A "consciousness system" was conceived based on this definition and actually constructed with recurrent neural networks. We succeeded in implementing imitation behavior, which we believe is closely related to consciousness, by applying the consciousness system to a robot.
This belongs to the branch of AI informally known as "faking it". There's a long history of work in this area, starting with ELIZA and continuing through a long series of rather lame systems. The latest systems are intended to mimic the behavior of call center employees.
Sadly, this isn't a joke.
AI? Whatever. Among serious theorists, it is pretty widely accepted that we will never reach a goal of true, hard AI (as in, something we created which is truly every bit as smart, independant, creative and "alive" as us, or even more) by cobbling together algorithms like this. It will come about by building the right sort of neural-net building blocks, arranging them in roughly the right kind of networks (probably via genetic selection algorithms rather than manually), and then teaching it much in the way one raises and teaches a small child. That's *if* we can solve the huge problems that still lie in our way going down that path (not the least of which is raw processing power).
This kind of shit isn't even in the right ballpark, and it's not going down the right road, and it's simply not productive in the long term. But gee, it gets headlines and research grants because it makes laymen say "ohhh neat". AI scientists of the world - I challenge you to get off your collective asses, stop pandering to morons, and get down to business with the decades of work that remain to be done.
11*43+456^2