In the coming spring, I will be teaching a broadly targeted course at Harvard offered through the Extension School and open to Harvard College undergraduates, graduate students, and continuing education students titled “Building the Brain: A Survey of Artificial Intelligence.” It is an in-depth course, as far as survey courses go, where I try to impart an understanding of natural neural networks, such as those found in our brains, and artificial neural networks (ANNs). Although students will be offered the opportunity to build from scratch a perceptron, considered the first artificial neural network building block, (see a great intro here) and apply deep learning to solve problems, it will not churn out experts in machine learning. It is intended to teach what AI is and more importantly what it is not. Its promise and fundamental limitations. This course came about for a number of reasons, not the least of which was my surprise at magnitude and prevalence of misconceptions about ANNs.
Many who know me know that I am bullish on artificial intelligence and have been since I was a teenager in the 1980’s. I have always had an affinity for neuroscience and computer science. My father is a renowned professor of neurology, and I would sit in his lab and hack Apple II games in my preteen years eventually even programming something useful for his laboratory. I had the privilege of working on neuronal cultures at the Georgetown and later contributed in a small manner to the Neuron Project, an early multi-research center attempt to model neuronal patterns accurately, while still in high school. After co-founding the computer Systems Engineering group and being hired as teaching staff in CS prior to graduating at my alma mater, the University of Wisconsin in Madison, I eventually received my B.Sc. in Molecular Biology with a special interest in neurophysiology. In spite of my belief that I would one day become a research physician like my father and his father before him, I was destined to live and breath computational science and information technology.
A little over a decade ago, I was lucky enough to sneak into Harvard when no one was looking and become Head of Neuroinformatics at the university’s Center for Brain Science among other hats. More recently I took on the role of Assistant Dean for Computing and CIO for the newest school at Harvard, the Harvard John A. Paulson School of Engineering and Applied Sciences, but I plan to return to more hands-on science and teaching next semester.
Computers have beguiled many with the promise of utopian and dystopian futures for the better part of a century. As my friend and notable experimental psychologist, Joshua Greene, recently noted to me, although we have now had many systems that pass the Turing Test, there are none yet that can fool any clinical psychologist worth their salt. If, or perhaps when, this should happen, it will pose a number of extraordinarily challenging questions from animal rights to more fundamental existential ones, but those I hope to discuss with my class at the end of the semester, not here.