First let me ask this: Is AI possible at all? In my opinion, yes, it is possible. I would like to cite a sentence from the movie matrix first episode (as far as I remember right now): "What is reality? How do you define reality? If you say reality is what you see, touch, hear, then reality is simply electrical signals interpreted by your brain"....
Is this reduction really justified?
Think about a 2 year old...Touching, looking, observing everything he/she can. To learn and discover the world through her senses.....Is she really just introducing data to her brain to be stored in some form of memory/experience (in a flexible form, not like a computer)?
All laws in science are elegant. They are nice, but most of all, simple. It doesn't matter if you need to explain that rule with pages of math formulas...They are the most beautiful, simple and true form of something... So must be the rules of the brain's operating system....Today, as of december 2008, after tens of thousands of years of human civilization, we have discovered many laws of science. We have come up with theories...But as of right now, we have no general theory of how our most intimate organ, through which we see, hear and think, works....We do not have a theory close enough to at least imitate it artificially in some simpler form...That is what we are looking for after all....
If we define true AI as an entity capable of thinking like humans do, then we must look at what humans have to think like that....They have a magnificient, astonishing organ called the brain....We have to learn it. Then we have to replicate it...No computer program, or software or algorithm or a robotics experiment which has a cpu like an ordinary computer with sensors attached will ever be successful. (today's computers and the brain are not even comparable anyway) Because this is not what we use. This is not what you use right now to read and understand and agree or disagree with this blog. You use your brain. You use the brain's structure and rules of operation in order to understand this blog. All we have to do, and the only thing we need to do, is to imitate biology. After that, we may come up with a better version. But only after doing that.
Again, do not even think about computers when you think about AI. Computers and the brain are no more similar to each other than an apple and a tea pot. Think about your car right now. Do you remember every detail, every pixel of its picture? You don't, yet you still recognize your car when you see it. Think about reading and "understanding" this sentence. How did you do it? Think about a person from the opposite gender that you like at first sight. Why did you like him/her and how is the decision made? You exit a building and it's raining outside. How do you decide all of a sudden to change your path to walk under canopies in order not to get wet instead of the open sidewalk? And how do you think about avoiding to get wet first of all (and why), even before a single drop of rain touched your body? You are riding in an underground subway train, and you can not see your surroundings. Yet you know where you are approximately, and on top of that you have a sense of time about when you will get off the train. You are reading this paragraph right now and thinking about all the images these words create in your mind. When I say "mind" for instance, what did you think? And why what first comes to your mind and my mind are similar to each other but never exactly the same? First of all how do you define similarity? How do you define it for a physical object? And how do you define similarity for abstract things? When you see a wallet you have never seen before, how do you know that it is a wallet? If you have a general picture in your mind about a wallet, how close that new wallet needs to be to your mental picture? And close in terms of what? When I say "happiness" what do you think, and how are you able to understand what it is? When I say "you should always try to avoid being sad" how did you combine that image with the rest of the words in that sentence and come up with "something" in your brain that enabled you to "understand" me? And when you reply "that's true but life is too tough sometimes" how were you able to determine the response, and put together those words? And examples go on to no end....What I am trying to say is, computers and human brain have no significant similarity at all. The brain "machine" capable of doing things described here and a trillion more feats, is totally different form a machine which is only able to follow a fixed set of instructions and produces some output, without even "understanding" what it is doing.
Humans were able to fly because they imitated birds in some form. Life has the most necessary rules and laws stored in itself already, and they are working. No need to look for something else, no need to introduce unnecessities, no need to tweak it, change it. Just admire it and replicate it, and then make it better if we can...This is our only way to success, if we want AI to be true.
Considering the ever accelerating rate of technology, who knows, this point might be closer to us than we think. Do not expect a technological advancement in th enext 20 years as it was in the last 20 years. Because it is not linear. It is exponential, parabolical. There might be 10, 50 times as more technological advance in the next 20 years compared to last 20 years. So, who knows, may be we are so close...
After all these discussions about the brain, we might still be missing a fundemental point though...And that is, even if we were able to replicate a brain 100%, we still might not succeed. Because human brain and human body are all connected and working together. Our brain learns and interact with the world through the senses muscles and our whole body. Think about the very simple sentence "the weather is cold". You only understand "cold" because you "felt" it before. You only understand the "weather" because you have seen different forms of it, and "felt" their heat, humidity, saw the sun, clouds, rain, snow and you know how they "feel". May be replicating just a brain is not close enough at all, and it will still not work, and we will need to have a good replication of all ourselves. Quite a task....
Until we come up with a working AI, all we will see is, the "dumm" AI improving, which is what is happening right now, with more and more historical data to base its output and ever more developed software. Such as the current cognitive search engine algorithms or chat programs, or checker play software, or autonomous driving software. This ever improving "dumm AI" will reflect on robots being better at what they are doing, and as long as they are capable of their tasks, we will not even care after some point. It is an interesting question that "how far can the dumm AI go?" May be it will improve to a point that it will be "just a little short" of the real AI. And who knows, by taking advantage of the processing power and being able to access huge databases and predetermined solutions, it might still perform better than the real intelligence, even for complex tasks, that require operation of robots in real environments. Even when that happens, it will change our lives considerably, having the same huge impact the computers had on our civilization in the last 30 years.
