Monday, June 9, 2014

Learning to Think

Hello, world. I'm a Computer Science PhD student studying Artificial Intelligence, which suddenly seems cooler when I write it down [1]. Today, I'd like to talk about intelligence, brains [2], learning, and solving problems. It's the kind of stuff that keeps me up at night, and amazingly enough, it also manages to pay my rent.

I discovered Computer Science in an Intro to Java course my freshman year of college. Programming hooked me with its instant feedback loop: write some code, and now I'm playing chess against myself. Write some more, and now the computer is playing against me. Write more, and now the computer is winning, despite the fact that I taught it (quite literally) everything it knows! It's a rush that I still don't know how to accurately describe. It's the kind of creation that drives artists, inventors, and scientists to work long hours for low pay and poor recognition, and once I tasted it I couldn't stop.

During college, I channeled this energy into a passion for solving real-world problems with computers, which the AI community calls "Machine Learning" [3]. That led me to enroll in a PhD program [4], where I joined the Autonomous Learning Laboratory at UMass.

I'd like to tell you about the grand projects we undertake in the lab, with shiny anthropomorphic robots following our spoken instructions and monolithic supercomputers that project the future in amazing detail [5].

Instead, my PhD research is mathy and esoteric. It turns out that humans are particularly good at exploring the unknown in small, understandable chunks, while computer algorithms excel at making shallow insights on massive scales. On occasion, however, the pieces come together to produce world-changing results [6].

My point in all this is that there's nothing mystical about intelligence, whether it originates in cells or silicon. We still have much to discover about the universe and our brains, but I believe that by teaching computers to learn, we're also giving computers a chance to teach us.

[1] Usually at this point in a conversation, someone chimes in with: "Artificial Intelligence? Like in that movie where science goes too far and creates something it can't handle, resulting in disastrous consequences?" Yep. That kind.

[2] You can't talk about AI without a tangent about the brain, and all too often someone smugly quotes: "If the human brain were so simple that we could understand it, we would be so simple that we couldn't." It's a nicely constructed sentence, but it doesn't stand up logically. Understanding the brain is not only possible, it's essential. I decided to leave the brain research to the neuroscientists, though, because it turns out there's a shortage of living human brains to experiment on.

[3] In truth, this is what AI has been about all along, but a period of overpromising and bad PR led researchers to re-brand the field.

[4] If you're considering a PhD, don't believe all the things you read about it on the internet. Each student's experience is different, based on the school, department, and advisor you work with. Your fellow graduate students also make a huge difference!

[5] That's actually not so far off, though it doesn't seem as futuristic in real life. I just asked my phone "Will it rain today?", and Google's massive network of datacenters spoke back in a pleasant female voice: "Yes, the forecast is 84 degrees with a chance of storm". My lab's robots look more like trash cans than people, though.

[6] Can you remember what finding information was like before modern search engines? It wasn't pretty.

CJ Carey
Amherst, Massachusetts, USA

No comments:

Post a Comment