What is artificial intelligence?

There are days when I think my intelligence is artificial . . . or is that wasted? I’m not sure.

Actually, this subject has a lot to do with science fiction because of the idea of androids like Data of Star Trek. He has a ‘positronic’ brain, whatever the heck that means. Actually, Isaac Asimov coined the term. It’s a fictional concept in order to explain how Data’s brain works.

The idea of artificial intelligence (AI) goes back a ways. Essentially, AI is a machine intelligence created or run by computers. Thinking machines go back to the ancient Greeks, but actual implementation of the concept had to wait until computers became powerful enough to even consider making AI work.

The first working example of this idea is Deep Blue, an IBM chess-playing computer that beat Gary Kasparov, a chess champion on May 11, 1997. This established AI as a viable concept.

But how could a real AI brain work? That’s a great question because the human brain is extremely complicated and it’s not well understood how it actually works. There are some basic things that an AI machine must do however: Deduction, Reasoning and Problem Solving. These three functions are absolute requirements for any AI system. This is based on the idea that humans use a step-by-step process in solving a problem. It turns out that this is not necessarily the case. There is a lot of intuition and random choice involved, processes that are hard to make a computer program do.

And, programming is where it’s at when it comes to AI. All computers can accumulate data, information and even visual images, but how does one program a computer to use these things to learn and reason? That’s a process called ‘machine learning.’

Machine learning is something that’s necessary for reasoning. It involves pattern recognition of images acquired from computer vision. For example, a computer sees a round block and it must reason that it has to go in a round hole, not a square one. It does this by pattern recognition and programming that tells it how to use this recognition to make the right decision.

The next hurdle is for AI to learn how to process language. A machine can talk and listen, but can it assemble words into a sentence that makes sense? Most talking robotics have a repertoire of sentences spoken and recorded by a human. A real thinking robot would need to make its own sentences to fit the occasion. An AI system must be able to listen and understand human speech and possess the ability to read. In order to do this one must program semantic indexing. This involves indexing large amounts of abstractions from language to speed up the process of understanding.

As an adjunct to language processing, an AI machine must learn social intelligence. It has to simulate human reactions and moods by recognizing and interpreting these things in a human it might be communicating with. How would one program a computer to have empathy or even emotion? How could we teach a machine to have imagination? These are difficult problems to overcome.

An AI system must have general knowledge and be able to use it in making decisions or even talking with a human. This involves a tremendous amount of storage and the science of cybernetics, which involves the scientific study of control and communication in a machine.

There are many theoretical approaches to this problem. One involves trying to duplicate how the human brain thinks. This is still a theoretical idea that has had very little success, and needless to say it’s very complicated and involves what is known as a ‘neural network.’ A neural network is what our brains use to process information. It involves interconnections between nodes (neuron cells) and algorithms on how to make it work. Duplicating this in computer programming is a nightmare turned real.

The real question is: if we do create an artificial intelligence machine like an android, would this result in it becoming smarter than we are?

That’s a question for another post.

Thanks for reading.