Who are you? And what do you think of intelligent machines?

Let’s answer the first question before moving on.

Emotionally, at least, the answer to who you are is that you’re the sum of your experiences. Otherwise known as your memories. Ah, but those memories might not be what you think. Every year we live, we experience more than 31 million seconds, high-def, live, and at regular speed. But the flash drives in our brains don’t just timestamp running footage like a security camera. We only recall bits and pieces, moments.

Or chunks, as Chris Baldassano might say. Baldassano is a postdoctoral research assistant at the Princeton University Neuroscience Institute who studies how the brain constructs its memories and, by extension, our experiences and sense of self. He’s also an optimist about the future of intelligent machines, less in the Matrix camp where computers will use us for batteries and more in the Star Trek camp where they will help us conquer new frontiers.

Baldassano will be a presenter at TEDx Carnegie Lake, a gathering of regional thinkers on a slate of topics ranging from sports to prison reform scheduled for Saturday, October 28, from noon until 5 p.m. at the John Witherspoon Middle School auditorium, 217 Walnut Lane. Cost: $35.

Other presenters include Olympic gold medal rower Amanda Elmore; Gregory van der Vink, founder of Novametrics; Joel Hammon, co-founder of the Learning Cooperatives; Karen Nichols, managing principal of Michael Graves Architecture & Design; Kate Germond, executive director and investigator at Centurion Ministries, a Princeton organization that works to vindicate the wrongly convicted; documentary filmmaker Marvi Lacar; Serena Burla, a professional marathon runner and cancer survivor; Suzanne Buchanan, executive director of Autism New Jersey; and Yusuf Dahl, an entrepreneur and real estate investor. Visit www.tedxcarnegielake.com.

How the memory and AI components come together is that memory is a necessary element of intelligence. Memory, in effect, is recalled experience, which means it’s how we learn. You can read this very sentence because your brain has learned what the letters are, how they sound, how they make words, and what those words mean. Your brain has just called up all that information without you even knowing it was doing it and used it in an intelligent way. Without that ability to take in pieces of information, store them, and use them, every word you’ve so far read would be gibberish.

Baldassano’s research looks into how our blank-slate brains start building and using the information we take in. Memories, he says, “are how we construct ourselves over time,” but the real question in there is, “how do we start building event models?”

A couple of terms to keep in mind: scripts and chunking. Scripts are exactly what they sound like: our brains’ way of building expectations and responding accordingly. Take walking into a restaurant. It’s familiar, even if it’s a new restaurant. You know generally what kind of experience you’re going to have.

“As soon as an event starts, you have a lot of expectations,” Baldassano says. That’s because we’re following the script, the social constructs, the behaviors learned. It’s why when we see the sidewalk stop and a lawn start, we know not to step into someone’s yard. But a bird or stray cat doesn’t have that concept, because they don’t develop scripts that tell them that one thing signals something bigger.

Part of what Baldassano is studying is why and how our brains are able to read these cues and translate them into scripts we can follow easily. Another part is understanding chunking, which is how our brains store our life experiences. Rather than recording like a filmstrip, Baldassano says, our brains pick out the highlights and store those in a way that takes up less space while giving us the most memory bang for our buck.

So when it comes to building artificial intelligence, the big holdup is that though we know how to do so much — though we know how to use our brains — we don’t actually know how the brain is able to allow us to do that. We are, therefore, unable (so far) to program machines to be able to recognize scripts and processes and use them in intelligent ways, Baldassano says. Sure, great strides are being made, but if we’re looking to build personality into a machine, we’ll be a while.

But it would be a mistake to think that Baldassano’s work is all about figuring out how to make machines fall in love with us or learn how to develop a business, even if his blog is titled “Rooting for the Machines.” If we better understood, in the present, less-sci-fi world, how experiences get stored — and how experiences shape our personalities over time — we would be able to develop better treatments for personality disorders, brain injuries, and countless other cognitive issues. It’s just a matter of solving that itsy-bitsy question of how our brains area actually able to do the astoundingly complicated things it can do.

The roots of this question probably go back centuries, when brains were the only technology people were aware of. But in the scientific research sense, the question started in earnest in the 1950s, Baldassano says. Back then, a group of scientists at MIT started what was to be a fun little project for a few months, designed to teach computers how to recognize objects. Sixty-odd years later, they are still trying.

“Some of these things seem really simple to us because we’re really good at them,” Baldassano says. Like recognizing words. But again, we don’t know how we’re able to recognize them, we’re just good at doing it.

One of the uphills in this area of research, Baldassano says, is that the brain can be studied on a near infinite sliding scale. There are researchers looking at whole brains, others looking at regions, others looking at one-millimeter segments, and still others looking at the 10,000 neurons in every millimeter-sized piece of a brain. Until all those scales can be unified, Baldassano says, research will continue to be slow going.

Baldassano, like many in the rapidly growing field of neuroscience, came to his vocation from the outside. He grew up in the Philadelphia area in a decidedly science and education-friendly house. His mother was a teacher and his father a physician. He always had a thing for science, so much so that he says he grew up “conflicted about what I wanted to do.”

In middle school he found computer programming and latched onto it as his career path. He earned his bachelor’s in electrical engineering from Princeton in 2009, with great attention on intelligent systems. He went to Stanford for his Ph.D., where he realized the connection between neuroscience and programming. He earned his doctorate in 2015, in computer science with a concentration on cognitive neuroscience and artificial intelligence.

Baldassano became a post-doctoral research assistant at Princeton in 2015. In 2018 he will take a position as assistant professor in Columbia University’s psychology department. And he is not the only Baldassano who found the path to neuroscience from elsewhere. His brother started as a chemist who became a doctor interested in doing brain interface work; his sister is a veterinarian who has gotten into the neuroscience of veterinary animals.

“People get pulled into neuroscience because there are so many interesting questions,” Baldassano says.

Once we understand how human brains can do what they do, Baldassano believes the future of AI will be a lot less like the Terminator and a lot more like those hopeful science fiction futures where machines can replace a lot of the dangerous and tedious work humans still do. Computers are already showing signs of helping doctors by being able to recall reams of medical information that human minds haven’t yet learned to chunk adequately, he says.

But if you’re still fearful of the rise of the machines, do try to remember, we don’t understand ourselves well enough yet, and what we’ve been able to translate into computers so far is way less impressive that what we’re capable of.

“Right now, artificial intelligence is dumb,” Baldassano says. “It’s not interested in ruling the world.”

Facebook Comments