Is Artificial Intelligence Even Possible?

AI already exists, we have computers that can detect objects from images, drive cars, detect diseases from MRIs, defeat the best human chess players, and win at Jeopardy. Computers can now do many creative and intellectual tasks better then humans can.

These things are pretty impressive mental tasks that certainly involved "intelligence" and that used to be done solely by humans.

Beyond that, you will have to define what "real AI" is. One measure of what AI is the Turing Test, where a computer would have to respond in a way that is indistinguishable from a human (ironically this would mean being stupider at things like retrieving knowledge and making quite calculations).

I think that the Turing Test will eventually be passed by a computer. I also think that this is further off then many people are predicting.

Do not forget self awareness that both my cats and dog seems to have but no computer program seems to have reach as of yet.

AIs to this point are very narrowly focus on such things as driving a car or playing chess and to reach true self awareness such narrowed focus need to be done away with.

But true self awareness seem possible to me.

How would you measure self-awareness in a computer? I believe that with animals, they do tests to see if the animals recognize themselves in a mirror. Computers can do that.

If self-awareness doesn't mean being able to recognize yourself or your impact on the environment (which computers have just fine), then please define it in a way we can test with an experiment.

This gentleman, professor Junichi Takeno, built a robot that passes the Mirror Self Recognition test which is used to determine whether an animal possesses self-awareness. This robot demonstrates the same level of self-awareness that dogs and cats have.

http://www.s2is.org/Issues/v1/n4/papers/paper4.pdf

In my opinion, this is pretty cool.

Of the many things I know about and am thus ocncerned with, AI isn't among them. By the time AI gets good enough to actually worry about, we'll have either killed ourselves off anyway or overcome such issues and made any threat from AI moot.

Climate change is a worry now. Not some nebulous thing down the road in the uncertain future.

That doesn't sound like AI so much as pattern recognition. Differentiating itself from other things, or knowing when it's seeing itself like. That's not all that impressive. Basicly just a kind of facial recognition.

I wonder if his robots would recognize themselves in the mirror if he'd put some make up on their faces, or a new hairdo...

I am asking a very specific question HesDeltanCaptian... what would impress you.

You have an AI in front of you that I claim can "differentiate itself from other things, or know when it's seeing itself".

How would you test this? What experiment would you need to do to convince you that this AI differentiated itself from other things. There are tests that scientists have used with human infants and with animals to determine self-awareness.. the robots can now pass this test.

What test would you need this poor robot to pass?

(By the way, facial recognition is very impressive... if you knew what it takes to do this you would agree).

I think the answer is yes... you will have to read the paper for yourself.

I think the answer is no, but I don't have the time to read his paper...

So you formed an opinion without looking at the evidence, and you are sticking with it.

Whatever works for you.

Well, yes, but you did the exact same thing...

No I didn't. I read the article and whatever else I could find on the topic.

Since you read the article, did he try to change their look through larger and larger alterations, testing whether they still recognized themselves say with a different face?

I am asking this because people can recognize themselves even though their face is always a bit different from yesterday: with different hairdo, a pimple on the nose, a bruise here or there, etc. There's more to this skill than just matching two static images.

To qualify as AI it has to do a lot more than recognize physical patterns. Computers could not even do that if we did not first tell them what they look like and then tell them exactly how to go about doing recognition. And of course a programmer would have to instruct them to go about the job of recognition. That is a LONG way from AI.

The AI that we are being warned about is a machine that could become self aware enough to independently decide on a goal that the programmer did not set. Self driving cars, pattern recognition, etc do not qualify.

A programmer could intentionally or accidentally program for doing random things but again, that is the programmers decision, or mistake, not the machine.

I'm guessing True independent decision making is not going to happen because we do not yet know what makes that possible and may never know.

Again you will have to read the article to see what the article says. The experiments he did involved actions.. he was modeling the robot behavior on animal behavior by trying to use the same criteria that is used to judge that dogs, or chimps or ants are self-aware. The specific test you are suggesting is not in the article.. but by inference it is likely that the robot would pass this test because it isn't comparing what it sees in the mirror with a static image.

That being said, the experiment you are suggesting can be done by computers. Facial recognition software can detect faces with well beyond an 80% accuracy rate (using the benchmark "faces in the wild dataset". This test includes identifying people with different makeup or a hat or a new hairdo.

This is a different problem though, but in the advances in this field are pretty impressive in my opinion.

Actually facial recognition as it exists today is basically simple. Starts with things like: what are the ratios of the triangle sides formed by the eyes and nose. Do they match that of the target?

Really simple, not AI.

You clearly don't know what you are talking about. Do any amount of reading on the modern research into facial recognition, the advance cutting edge techniques they are using, and the success in laboratory tests, and you will quickly change your mind.

Google and Facebook are both getting very impressive results in the field. Some very prominent researchers are applying cutting edge techniques to the problem.

Do a little reading before you spout off about something you know very little about. (For the record, I work in computer speech recognition which is a related field, although I personally look up to the facial recognition people... they have a more complex problem).

Holy **** man, you get hostile mighty easy.
I gave only the most basic starting point. Of course the techniques get more and more sophisticated as you add details and yes, it works pretty good. But the point within this thread is that computers are only doing what the programmer directs, no matter how sophisticated the algorithms might be.

The premise of the 'AI warnings' is that our abilities are a result of a very large collection of bio-logic gates and can be replicated by a sufficiently large computer with clever enough programming. I'm saying they have reached a wrong conclusion.