My History with Artificial Intelligence

Chad J Woodford
Machines Learning
Published in
4 min readFeb 15, 2018
Officer K from Blade Runner 2049, a replicant

I’m excited to announce the launch of my new AI website, a space for me to publicly explore my renewed interest in artificial intelligence. I’ll be using that space to talk about developments in artificial intelligence and related topics, including the ethics of AI, practical uses, assessments of the state of AI, and implications for humanity and the global economy. I’m fascinated by what’s happening in machine learning, intelligent assistants, and practical applications of advances in machine intelligence, such as smart cities. So I’m excited to have a dedicated place to discuss it. As with this post, I’ll occasionally cross-post here.

My personal connection to AI goes back about 25 years. I studied electrical and computer engineering at Clarkson University, and did graduate work there immediately afterward in artificial intelligence from 1993 to 1995, at the tail end of the second “AI winter.” Back then, I was intrigued in particular by neural networks (one approach to machine learning) and by fuzzy logic, the latter a field of AI you don’t hear much about it anymore. And I was enamored with the work of Douglas Hofstadter, a giant in the field at the time and very deep thinker. Unfortunately, my thesis advisor at the time was biased against these emerging fields and consistently attempted to dissuade me from pursuing my thesis in training a neural network to play Go. In fact, he would often send me academic papers criticizing machine learning and fuzzy logic. I should have transferred schools at that point.

I was a fairly philosophical graduate student (this trend continued in law school with my classes on jurisprudence). At the time, I figured we either tackle the challenge of achieving artificial consciousness or nothing at all. Remember, we were a good twenty years out from any kind of practical applications of artificial intelligence, like medical diagnosis or smart speakers.

At the time, I concluded that if we couldn’t achieve what’s known as artificial consciousness — a self-aware machine — why bother. Maybe I was too ambitious. Or too idealistic. Or just too interested in the nature of the mind and of consciousness. In any event, once I understood what it would take technologically to achieve artificial consciousness, and that it may not even be possible, my interests in the field quickly waned.

So my disappointment in the plodding pace of advancement in AI combined with lack of support from my advisor motivated me to drop out and go into business with two fellow grad students who were equally disenchanted with academia. (Our first idea was to helping cable companies in upstate New York offer high-speed internet access to cable TV customers, an idea well ahead of its time in 1995, especially in rural New York. We eventually pivoted to software consulting and I left to join a startup in Vermont for the snowboarding.)

Given my work in grad school on a program that could learn to play the game of Go, I was excited to see that DeepMind finally created a computer program that beat the world Go champion, about a decade earlier than expected.

AI TERMINOLOGY

Which brings us to AI terminology: The term ‘artificial intelligence’ has become very overloaded, and misunderstood. Most of the applications of AI making headlines today are actually weak AI, instead of strong AI or artificial general intelligence. Think Siri (weak) versus Roy Batty from Blade Runner (very strong) — or Officer K if you must (pretty strong). Impressive recent applications of narrow AI and machine learning like AlphaGo are trained on a very narrow domain with a very large data set. We’re a long way from general-purpose or artificially conscious AIs. So, in subsequent posts, I will be careful with these terms.

Speaking of strong AI, here’s a fun anecdote from my grad student days: Back then, the first two Terminator movies dominated the general population’s understanding of the future of thinking machines. One night at the local pub, my then girlfriend confessed that she had started to worry that, in my research, I was helping to usher in the irreversible and dreaded Rise of the Machines. We ended the night by betting each other a dollar that machines would/not take over the world. She never did pay me that dollar. But we never set an end date either…

Anyway, welcome. I hope you’ll follow along with me as we plunge more deeply into the era of practical, weak AI.

--

--

Chad J Woodford
Machines Learning

Philosopher of technology, yoga & meditation teacher, product manager, lawyer, writer. Tweeting @chd. bio.site/cosmicwit