I was at the NERCOMP Annual Conference last week. There were some really interesting presentations that I attended, but I should say that the first keynote by Gerard Senehi was less than optimal for a conference to open. Danah Boyd, on the other hand, was fantastic, talking about  how even the younger members of our society care about privacy, contrary to the myth that they don’t.

One particular talk that I liked and want to follow up has to do with open educational resources. The powerpoint presentation is available along with the abstract, so please review it. Though some of the panelists are from institutions that are very different from us, we feel that there is something here for us to learn from and educate our community.

Artificial Intelligence has been in the news recently and frankly, trying to define it in clear terms is something I am not capable of. It has morphed over the years thanks to advances in computing. Is it possible for machines to emulate humans in the way we think? This is a loaded question as you can imagine.

Theoretically speaking, an artificial intelligence system must pass the Turing test. This test involves a party game where a man and a woman play with a third person who is trying to guess the genders accurately. The man provides all answers to convince the third person that he is a man while the woman provides tricky answers to convince the third person that she is the man. Turing proposed that if you switched one of them with a machine then the person needs to guess who is a human and who is a machine. If the person failed to guess correctly more than half the time, then the machine will be declared having passed the test (that it has enough intelligence on its own to fool the third person).

There are a lot more underlying details to this of course, because of the availability of massive amounts of data and the computing power, even the “brute force” computing can be confused with intelligence.

I distinctly remember the Artificial Intelligence (AI) course I took when I was a master’s student in CS back in the early ’80s. It was mostly about pattern recognition of simple three dimensional objects. We used a specialized language called LISP, which was favored for AI. It actually came in handy for me later on as a user of an editor called emacs  which uses a variant of Lisp called elisp. BTW, emacs is my favorite editor and I continue to use it after all these years!

Machine learning is a subfield of AI that has become prevalent we can relate to easily. If you think about the way most of the computing takes place, everything we want is explicitly programmed in. If the program ends up doing something that we did not intend for it to do, we don’t call it intelligence, but “bugs“. However, there are many problems that cannot be explicitly programmed. We humans somehow seem “to figure out” things in real time that we are unable to translate into steps that can be programmed.

There is a tight coupling between statistical methodologies and machine learning. If you are interested in this field, you can find many approachable courses in the MOOC world. Supervised and unsupervised learning are two broad categories that is used in machine learning. In the former case, you train the system with a variety of known inputs and outputs (the more variety the better) and in the latter case you leave it to the system to infer the hidden information without necessarily training the system. It is easier to understand these through examples.

You all are aware of many popular AI examples ranging from IBM’s Watson, and Deep Blue to Google’s self driving cars  to personal assistants (such as Siri, Google Now and Cortona). Here is a short list of everyday things where AI has a part to play.

Like all technologies, there is always many sides to any technology. Whereas AI has resulted in some really fantastic applications mentioned here and elsewhere, the recent fiasco involving a chatbot called Tay from Microsoft points to the unintended consequences of technologies one fears.

Unfortunately, just the way you find both good and bad amongst humans, the AI systems will have both good and bad aspects associated with it for a variety of reasons. Before it gets totally out of hand, we need to exercise extreme care, unless the self thinking machines themselves are capable of winning over the bad machines!