I have been involved with Decision Theory and Artificial Intelligence for the past 50 years and have some comments on the subject.
Both Algorithms and Artificial Intelligence have their place, but there are some key and significant differences. For years we have used both Algorithms and Decision Trees which are both easy to teach or represent on a two dimensional black board or text book. The problem is that our historical methods of teaching or representing knowledge has been very limited. As a result few of us have progressed beyond simple Algorithms and Decision Trees.
Here is another way to look at the situation. I am going to use the crash of the two 737's for my example. The pilots appear to have been trained in "how" all the controls worked and were able to operate them. The normal training procedure after the pilot knows how to operate all the controls is to allow them to get more "experience" with more flying hours. If you think about it, exactly what is "experience" supposed to provide to the pilot? The pilot has been armed with a few "algorithms" that he can use in situations that are likely to occur, but these "algorithms" don't cover 100% of the possibilities that the pilot may encounter. As you can imagine, in real life, there is an almost infinite number of different situations that any pilot can expect to encounter. How can any system prepare a pilot for an infinite number of possibilities? See the problem?
This is where "Artificial Intelligence" comes in. AI is not an algorithm that only covers one or two situations. It is a set of "rules" or "principles" that, given a situation, generates the "best" action to deal with the situation. This is the way our brains work. From what we have learned from personal experience allows us to generate new solutions that fit the situation, but which we have never before used or considered. It is not magic that some people seem to know how to solve all kinds of problems that they have never encountered before. It is that they have applied rules or principle that they had gained in the past to allow them to develop "new" solutions to new problems. Exactly how one's brain does this is not clearly understood, but we know that it must be complex and that it makes use of different kinds of prior knowledge. The challenge is to figure out how to do some of what our brain does with a computer program.
I think that you can see that it would be impossible to ever come up with enough different algorithms to compete with a human brain, but it may be possible to try to do it the way that the brain does. There are all kinds of computer systems that try to duplicate what our brain does that I will not go into here, but is seems clear that algorithms are just a start, but not the final solution. You may even think about this like the "law" where our judges are always making what looks like different decisions based on their interpretation of the laws on the books.
Back to the pilots. They knew "how" to operate all the controls, but they did not know "when" to do "what" or "how much" or "how fast". If you have ever ridden in a car with a young driver, then you must have noticed how they "under" or "over" steer the car. I think that this is exactly what the pilots did. They over steered the plane until it crashed due to "PIO". Pilot induced oscillation. A good AI program could have prevented the crash.
Both Algorithms and Artificial Intelligence have their place, but there are some key and significant differences. For years we have used both Algorithms and Decision Trees which are both easy to teach or represent on a two dimensional black board or text book. The problem is that our historical methods of teaching or representing knowledge has been very limited. As a result few of us have progressed beyond simple Algorithms and Decision Trees.
Here is another way to look at the situation. I am going to use the crash of the two 737's for my example. The pilots appear to have been trained in "how" all the controls worked and were able to operate them. The normal training procedure after the pilot knows how to operate all the controls is to allow them to get more "experience" with more flying hours. If you think about it, exactly what is "experience" supposed to provide to the pilot? The pilot has been armed with a few "algorithms" that he can use in situations that are likely to occur, but these "algorithms" don't cover 100% of the possibilities that the pilot may encounter. As you can imagine, in real life, there is an almost infinite number of different situations that any pilot can expect to encounter. How can any system prepare a pilot for an infinite number of possibilities? See the problem?
This is where "Artificial Intelligence" comes in. AI is not an algorithm that only covers one or two situations. It is a set of "rules" or "principles" that, given a situation, generates the "best" action to deal with the situation. This is the way our brains work. From what we have learned from personal experience allows us to generate new solutions that fit the situation, but which we have never before used or considered. It is not magic that some people seem to know how to solve all kinds of problems that they have never encountered before. It is that they have applied rules or principle that they had gained in the past to allow them to develop "new" solutions to new problems. Exactly how one's brain does this is not clearly understood, but we know that it must be complex and that it makes use of different kinds of prior knowledge. The challenge is to figure out how to do some of what our brain does with a computer program.
I think that you can see that it would be impossible to ever come up with enough different algorithms to compete with a human brain, but it may be possible to try to do it the way that the brain does. There are all kinds of computer systems that try to duplicate what our brain does that I will not go into here, but is seems clear that algorithms are just a start, but not the final solution. You may even think about this like the "law" where our judges are always making what looks like different decisions based on their interpretation of the laws on the books.
Back to the pilots. They knew "how" to operate all the controls, but they did not know "when" to do "what" or "how much" or "how fast". If you have ever ridden in a car with a young driver, then you must have noticed how they "under" or "over" steer the car. I think that this is exactly what the pilots did. They over steered the plane until it crashed due to "PIO". Pilot induced oscillation. A good AI program could have prevented the crash.
