Today I continue my exploration of the state and future of Artificial Intelligence.
As we have seen, there are several types of AI. One is a limited AI also known as an Expert System. This works based on a set of rules. If you feed in inputs of a certain type, it will make decisions based on the rules it knows. While this sounds cool, it is very limited as to what it can do. It has to have it’s inputs conform to certain predetermined values.
One of the best known Expert Systems is Watson. This computer was able to beat humans on Jeopardy a few years ago and it now being touted by IBM for medical field related tasks. For this, it is well suited.
From an article I found on Gizmodo yesterday…
A team of researchers from the University of Illinois Chicago recently set out to discover just how advanced our artificially intelligent computers have become. Like they’d do with any food-eating human, they gave the computer an IQ test. The machine in question, a ConceptNet 4 artificial intelligence system developed by a bunch of eggheads at MIT, took the Weschsler Preschool and Primary Scale of Intelligence Test, a standard IQ test for children, and scored about as high as a four-year-old would have. It turns out that while it did well on questions with cut-and-dry answers, the computer had a lot of trouble with the “why” questions.Oh and there was one other thing. “If a child had scores that varied this much,” said Robert Sloan, lead author on the study, “it might be a symptom that something was wrong.”
Note it has trouble with “why” questions. Of course my Smart Toys concept is designed to interact with children, who’s primary question is “why”. :) Obviously that concept isn’t ready for primetime yet.
Still, it should be possible to use the current limited AI to create tutors for students in a variety of subjects. Even a teacher can’t answer every question thrown at them. Naturally, they can answer a lot more than a computer, but perhaps we will see limited AI as a way to augment and supplement what a teacher can do? I imagine this will be an extra you can buy when you buy your e-textbook. There will also be a variety of different AI tutors available of differing capabilities. These AI tutors will get better over time also because they will learn from the types of questions they get. Finally, AI tutors will greatly improve the efficacy of MOOCs.
I really wonder if limited AIs will be able to replace teachers in a lot of situations since the circumstances are restrictive. Could the AI deliver a virtual lecture? If you do it online, the lecture could be tailored for each students. Could it also answer questions? The answers could relate back to previous questions the student has answered or incorporate interests the student has. All of these things, a good teacher can do, but not to a whole bunch of people at once. Human teaching can’t both scale and deliver a personalized experience. You can do one or the other but not both.
You know who this will replace first? Teaching Assistants. Why have a class taught by a grad student of questionable quality when you can have a AI teacher do it better?
I don’t have a good timeline for this to happen but I would bet sometime in the next 10 years (maybe sooner) we will see a lot of teaching done by AIs, especially online. I think it will happen first in the private sector.
As always, I invite your comments!