I thought this article was interesting, especially considering our latest topic in class. Apparently Japanese scientists have created a robot that can “taste” food.
Japanese unveil robot wine steward
This reminded me of “The Robot Reply” which is part of the Searle article. The argument is: if we put a computer inside a robot which apparently perceives the world around it (through seeing, doing, tasting, etc.) couldn’t we say that the robot really has the ability to understand? Searle says no, because the robot is still just manipulating symbols and facts.
All right, imagine this…. the total Turing test. Scientist build what they claim is an intelligent robot. It can decently carry on a conversation. It can see using cameras and scanning devices. It can taste with a more evolved version of the technology described in the article above. It can also detect sound, touch, and smell.
So… does this robot have a mind like we do?
My instinct tells me no.
The artificial intelligence tests we’ve discussed the most frequently are chess and using language. I don’t think either one of these is particularly good for comparing a computer to the human mind. Chess is a closed, formal system which the computer can play with probability and calculations. Even language can be sort of calculated, because of the structure that comes with grammar. The computer could pick out key words and arrange appropriate responses according to the Subject- Verb- Object foundation.
I think a better test would be analyzing a painting.
Take the robot I’ve described and tell it to say what it thinks the painting means. The robot sees the picture through the camera lens. It scans the painting from top to bottom, picking up on colors, lines and space. Assuming the picture isn’t abstract, the robot might be able to tell you if the subject is a woman, a building, etc.
The robot also might have certain information about paintings and symbolism programmed in, such as: blue is used in sad paintings, or a + shape might somehow allude to crucifixion; but, I don’t believe a robot could ever give an interpretation on what the painting might symbolize, or for that matter, decide whether or not the painting is beautiful. The thought process is too subjective. These decisions can’t be made with the manipulation of symbols.
I realize I’ve given somewhat of an “X- factor” response, but analyzing art is somewhat different from some of the other X-factors listed by Turing. Some of the examples he gives include: falling in love, caring, enjoying strawberries and cream, and having a sense of humor. I agree with Turing’s critics that the human mind is somewhat needed for these abilities, but they don’t necessarily have anything to do with the kind of intelligence that artificial intelligence tries to replicate. Learning art, however, most definitely requires intelligence…. the kind of intelligence computers simply cannot have.
Thoughts on this?
Erin Murphy | 20-Feb-07 at 8:24 pm | Permalink
By the way, the “tasting” robot scanned the cameraman and identified him as bacon. There’s just something funny about that…
Daniel Estrada | 20-Feb-07 at 10:17 pm | Permalink
I’m planning on showing these short films during Thursday’s screening, but since it is relevant, I’ll link it here:
The film-making robot.
Some additional commentary can be found here.
Along the same lines, here’s a painting bot.
Kerry Phelan | 21-Feb-07 at 2:37 pm | Permalink
I agree with you Erin. I don’t think that a robot can analyze a painting in the ways that you describe. The thought process invovled in that is far too extensive for something that I believe can’t think on it’s own.
Danielle Foster | 21-Feb-07 at 7:45 pm | Permalink
I agree. I think that there are so many abstract ideas that could come from looking at and analyzing a painting. A painting can evoke so many emotions that a robot could never feel. I’m sure that the robot could spit out some generic thoughts about the painting but that they could never really understand the true meaning behind the painting.
Annie Werly | 22-Feb-07 at 5:26 pm | Permalink
Erin- I think your argument is well thought out and I certainly agree. The human mind is too complicated to be mapped using simple things such as input and output responses to complex issues or interpretations. There is no way that a robot can possiblity understand and interpret certain symbols and situations: it is simple not possible, nor is it beneficial.
Ashley Graham | 26-Feb-07 at 3:29 pm | Permalink
I agree that the robot will never be the same as a person. There is no way for a machine to have the same indepth understanding and emotions of a human. They will never love, show compassion, or sadness. There are certain situations which can only be understood by humans, and you captured this well in your response.
Krystal Kniep | 26-Feb-07 at 3:52 pm | Permalink
I agree with you Erin that chess and language are not great ways for comparing a computer to a human mind. These tests can be programmed in a computer through prbability and calculations. More authentic tests would be dealing with emotions and feelings. If you give a robot something open-ended, like your painting example, I hightly doubt the robot will be able to give his own true interpretation.
Erin Murphy | 26-Feb-07 at 4:05 pm | Permalink
I don’t even think you really need to test emotions to see if a computer is intelligent like a human.
Art (good art, anyway) will usually inspire emotional responses, but you don’t necessarily need to have that kind of reaction to make an interpretation.
Like, take this part of the Sistine Chapel ceiling.
http://battellemedia.com/images/sistine%20chapel.jpg
Would a computer be able to tell you that the two people in the picture are God and Adam, and that God is giving life to Adam?
Daniel Estrada | 27-Feb-07 at 12:28 pm | Permalink
If the computer is sufficiently instructed in art theory, then it should be able to make the identification. It doesn’t even need to be very sophisticated- try searching ‘Adam and God’ in Google’s image search, and the Sistine Chapel makes up the first page of responses.
Why do you think brute identification is impossible for the computer?
Erica Yuenger | 28-Feb-07 at 3:01 pm | Permalink
I think a computer would be able to identify subjects of a painting that are “famous” because the iconography of certain people such as God, Zeus, or Aphrodite remains the same throughout many paintings.
I do agree,however, that a computer could not generate a subjective interpretation of a painting, because, as I have mentioned in other comments, the emotion that a painting evokes comes from experience.
Erin Murphy | 28-Feb-07 at 5:15 pm | Permalink
Doesn’t Google just pull up the image because of keywords and tags? It’s not making its own judgment about who the two men are.
Because this painting and its intentions are well known, a computer might be able to respond with a pre-programed identification. But let’s say that this is a brand new, unheard of painting. How could a computer understand its meaning? On the surface, it just looks like two men, one of them naked, touching fingertips. How could the computer interpret who these people are, and that the naked figure in a way represents all of humanity?
Daniel Estrada | 11-Mar-07 at 3:58 pm | Permalink
Erin: Its not that simple, because the keywords and tags aren’t entirely provided by Google’s designers, nor are they entirely provided by the authors of the webpages that Google is indexing. The internal structure of Google is largely based on its own ‘assessments’ or relevant categories and groups.
So yeah, Google doesn’t know who God and Adam are, but it knows a whole lot about how they are associated. Its a subtle but important difference.
Annie Werly | 12-Mar-07 at 7:13 pm | Permalink
It is certainly more than emotion that the robot cannot identify: the fact that it judged people as bacon and prosciutto indicates its complete unfamiliarity with anything that it is not specifically programmed to do. Even the things it IS programmed to do are limited- it can only know a few dozen wines at best. They are certainly making strides, however the technology is still quite limited.