I’ve always been interested in the MIT Media Lab and also artificial intelligence since university. I’ve read a few books about projects there and have read some information about the Media Lab on the web. The projects are always interesting and usually end up in consumer products eventually.
Recently I was looking for any developments for software agents and came across a project that contains striving to teach a computer common sense. This project intends to teach computers that which is fairly easy for humans. Internet users can register at Open Mind Commons and become the computer’s teacher. It will probably be years before a computer will understand what we call Common Sense but this kind of project is a genesis for such a venture. If you are interested in the research paper, please visit http://www.openmind.org/papers/commons.pdf.
I just found out about it over the weekend and became addicted to trying to teach the system new concepts. It is incredibly frustrating and yet also invigorating to be the first to tell it something. For example, it seemed to understand the concept of money fairly well but struggled with the concept of a storm. It has the option of asking questions about certain words and I was getting things like “Can a storm be used as a weapon?”.
There are a few things that you realize with such a system. First of all, “Common Sense” really isn’t common. Secondly, it is very hard to convey meanings in text alone to a computer system.
Probably the hardest aspect to detect is nuances. In this category is words that have multiple meanings. Tonight I had an issue with Darwin being both a man and a city. I solved it by calling the person “Charles Darwin”. I think the computer system is still confused.
I recommend trying to teach the computer system what you consider common sense. Some of the questions it asks are quite good and also quite funny. It knows more than I suspected and also is very good at detecting duplicate information.
From a Citrix angle, this kind of research would benefit computers making sense of environments and the best operations from learning concepts that apply. It would certainly fit with the computer systems becoming more autonomous. It would also benefit the users that would be provided with a system that has more common sense.