Miscellaneous > The Lounge

HP face recognition doesn't do negroes

<< < (4/5) > >>

Aloone_Jonez:

--- Quote from: piratePenguin on 23 January 2010, 17:16 ---Can your cat scan through billions of web pages and find occurrences of search terms?

Different tools for different jobs. For now, though, HP should've used cats.  8)

There are two identical twins I know in college, well I see them (normally one of them) around every week or so often. Never call them by their names. Even if I recognize the difference in their faces, I get mixed up with names anyhow and can't associate them with the correct name reliably. So usually if we're going out I ask someone "is mick/kev in the white shirt?" or figure it out like that. Not good at learning small differences in things, and very bad with names. But if I was a cat and I didn't need names I'd probably know mic/kev since I wouldn't associate their differences with a name and then get mixed up so I can't recall it confidently.

--- End quote ---

A search engine isn't really intelligent because it doesn't actually understand the information it sifts through. I'm not sure that a cat wouldn't either, apart from pictures of other cats, people and mice maybe. ;D

Just because you can't remember names very well doesn't make you unintelligent, it's just that your memory isn't very good. I'm not good at matching faces to names either: I generally need to vaguely get to know people before I can remember their names.

Intelligent animals can do things and solve problems that they would never normally face in the wild. My cat can differentiate between me and other people, even though it probably doesn't know my name but that wasn't what I saying. My point was that in the wild it only has to recognise feline faces, and is only able to recognise humans because it has intelligence, which a machine does not.

No doubt some face recognition software could be designed to sift through a huge database and recognise people better than any human can but because it's not intelligent because it'll completely fail if you asked it to differentiate between two cats.

Animals totally pawn even the most powerful computers when it comes to problem solving and common-sense reasoning.

For example my cat has figured out to open sliding doors, we didn't train it to, it figured it out for itself. What's funny is, if it knows I'm watching, it doesn't bother, it plays dumb and looks up at me to get me to do it. The crafty cat has figured out how to get me to open the door because it's too lazy to do it. ;D Believe it or not, this takes brains, this behaviour can't have been preprogrammed: cats evolved long before people or sliding doors. A robot wouldn't be able to do this unless it was specifically programmed -even a cat is far brighter than any AI currently is.

EDIT:
Added quotes and last paragraph.

piratePenguin:
Computers do what they're programmed to do. So if you program a robot to sense things and what seems like a good idea in the physical world and what seems like not a good idea (such as resting your hand on a hot stove), and record and learn from these things: even communicate these things to buddy robots via the internet, maybe you can call it intelligent?

We may never see "intelligent" robots in our lifetime, but if they do everything we program them to, and we program them to learn about the world then if they're programmed correctly they'll do that.

worker201:
One of the reasons AI usually seems so dumb is that we don't really understand our own intelligence.  Like with your cat, you may think that it recognizes you, and you could maybe prove that it recognizes your face, but you will probably never know what it is about your face that makes you recognizable to a cat.  So far, facial recognition is based on theories and experiments, but we don't know for sure how we recognize faces, so it's mostly stabs in the dark.

Lead Head:
The biggest problem with most modern AI systems is that they don't really think or learn. They are still mainly very scripted preprogrammed things. With just pre-written responses to expected events. The only difference is now is that with the large amount of space and processing power, you can have very large and very complicated "decision trees"

I saw a video from a couple of years ago, and at the time the most sophisticated "learning" AI in the world was only able to do very basic things. Like this certain robot was given a car to play with, the most it was really able to figure out was that the car could only be slid easily forwards/backwards and not side to side, and even with that basic ability the computing hardware required was immense.

I also wouldn't under estimate the intelligence of animals when comparing AI to them. Especially when you consider that (Koko) the Gorilla knows over 2000 words of spoken english, and over 1000 sign-language signs. Your average domestic dog has far higher reasoning and logic skills then chimps and 3 year old human infants.

Facial recognition programs aren't really AI, they just have algorithms designed to look for shapes, ex: ovals, circles, and certain colors. They aren't like, "Oh hey thats a pair of eyes on a human head", its more like "Two ovals within a larger oval/circle - Focus Here"

Aloone_Jonez:

--- Quote from: worker201 ---One of the reasons AI usually seems so dumb is that we don't really understand our own intelligence.  Like with your cat, you may think that it recognizes you, and you could maybe prove that it recognizes your face, but you will probably never know what it is about your face that makes you recognizable to a cat.  So far, facial recognition is based on theories and experiments, but we don't know for sure how we recognize faces, so it's mostly stabs in the dark.
--- End quote ---
You're right, we don't understand our own intelligence, let alone that of other animals. I'm pretty sure my cat recognises me as it's shy around strangers but I can only assume it uses sight, it could be smell but being a visual predator I doubt it, I don't know about dogs though.


--- Quote from: Lead Head on 25 January 2010, 00:53 ---The biggest problem with most modern AI systems is that they don't really think or learn. They are still mainly very scripted preprogrammed things. With just pre-written responses to expected events. The only difference is now is that with the large amount of space and processing power, you can have very large and very complicated "decision trees"
--- End quote ---
Yes, they lack the basic reasoning, all animals with brains seem to have.



--- Quote ---I also wouldn't under estimate the intelligence of animals when comparing AI to them. Especially when you consider that (Koko) the Gorilla knows over 2000 words of spoken english, and over 1000 sign-language signs. Your average domestic dog has far higher reasoning and logic skills then chimps and 3 year old human infants.
--- End quote ---
I know that many animals are more intelligent than human infants but how is a dog brighter than a chimp?

I don't think that's true, unless you're comparing it to a baby chimp or some other ability such as smell.


--- Quote ---Facial recognition programs aren't really AI, they just have algorithms designed to look for shapes, ex: ovals, circles, and certain colors. They aren't like, "Oh hey thats a pair of eyes on a human head", its more like "Two ovals within a larger oval/circle - Focus Here"

--- End quote ---
If I were clever enough to design an AI image recognition system, I would base my design on a mantis shrimp. They're clearly the most intelligent crustaceans with the most complex eyes in the animal kingdom. They have hyper-spectral vision, are sensitive to polarised light and each eye has neurons which processes the information before it gets to the brain. They're capable of recognising each other and there's anecdotal evidence they can recognise people.

http://en.wikipedia.org/wiki/Mantis_shrimp
http://www.uemis.org/es/magazine/nature_and_science/peacock_mantis_shrimps_pugnacious_predators
http://blogs.thatpetplace.com/thatfishblog/2008/07/03/mantis-shrimp-order-stomatopoda-–-breaking-research-and-care-in-captivity/

I'm more interested in alternate paths to intelligence: we know it's evolved in mammals but it's present in: birds (not as stupid as people once used to believe), fish (yes sharks are actually quite bright despite their reputation) and invertebrates cephalopods (octopus, cuttle fish and squid), stomatopods (mantis shrimp, see above) and possibly also jumping spiders.

"Crows in urban Japan have innovated a technique to crack hard-shelled nuts by dropping them onto crosswalks and letting them be run over and cracked by cars. They then retrieve the cracked nuts when the cars are stopped at the red light."
http://en.wikipedia.org/wiki/Bird_intelligence#Tool_use

Octopus 'turns off' a light which is annoying him.
http://www.youtube.com/watch?v=YBghUIEfDxg&NR=1#

Octopus opening jar to get its dinner.
http://www.youtube.com/watch?v=ocWF6d0nelY&feature=related#

These animals have developed brains down a different evolutionary path, especially invertebrates which have totally different nervous systems to ours. Suppose s scientist figures out exactly how we recognise faces? They might believe that it's the only way of doing it until they investigate how a squid does it which is totally different and might be easier for us to develop an implementation.

Navigation

[0] Message Index

[#] Next page

[*] Previous page

Go to full version