Sign up for our newsletter and get the latest big data news and analysis.

Interview: Pamela McCorduck, AI Industry Pioneer

I recently caught up with Pamela McCorduck, AI expert, industry pioneer and author of her compelling new book “This Could Be Important: My Life and Times with the Artificial Intelligentsia,” which covers her unique experiences around AI. She has been working around the field of artificial intelligence since 1960. In the interview she offers unique insight on the evolution that developed the technology we use daily today. Pamela witnessed the first developments of AI when the impersonal machine first evoked a candid and personal response from a human and has been documenting ever since.

insideBIGDATA: Can you tell us what motivated you to write this title? Why now?

Pamela McCorduck: My story is unique—no one else could tell it. I knew the founding fathers of AI both professionally and socially from the time the field was barely underway. I knew that these weren’t nefarious characters, planning to make terrible mischief in society. They were scientists, hoping to model human intelligence scientifically, by using this new and dynamic instrument called the computer. I was there when a speech understanding program that could understand a thousand spoken words (albeit in a very limited field) was a triumphant achievement. When I moved to Manhattan and found myself in literary circles, I’d say to public intellectuals: “Artificial Intelligence? This could be important.” Of course they laughed—uproariously, or condescendingly, or in utter disbelief. A machine thinking? How silly is that? This could be important was my mantra for decades.

insideBIGDATA: You’ve experienced first hand a number of “AI” winters. What makes this time around, with all the excitement around the field, different?

Pamela McCorduck: First, I’d like to unpack that term, “AI winter.” What it really meant was that a period had come along when research in AI couldn’t easily be monetized. I went back and looked at the titles of research papers published during the first so-called AI winter, the early and mid-eighties. Fundamental research was being published, research that would undergird future AI applications (e.g., major algorithms around machine learning but only when the technology could accommodate them, a few decades later). 

But the 1980s was the decade of the Reagan administration, when people brought in to guide and fund AI research, especially in the Department of Defense, were ideologically committed to monetizing government research. When the research didn’t immediately yield itself to obvious commercial applications, the funders went looking to support work closer to commercial realization. What we see in AI’s success right now—machine learning, for example—is the result of research done during that decade. When administrations change and funding agencies can take a longer view, end of “winter.” Of course present publicly lauded success draws young researchers to join in the excitement, so youngsters flock to the field. Is another AI winter coming, as I read recently? Could be. One of the leading scientists in machine learning told me—and this is in my book—that machine learning has gone as far as it can go, and he was looking for different approaches to push AI forward. Machine learning apps will proliferate until the end of time it seems, but that’s not research. It’s development.

insideBIGDATA: How do you feel about the way people/companies play loose with the term “AI”? As a result do you feel the general public is getting the wrong impression about the technology.

Pamela McCorduck: I’m not a linguistic purist when it comes to AI. Our model of intelligence is humans, and what do you know? To perform intelligent behavior, humans use all sorts of approaches, from pattern recognition to the highest kind of symbolic reasoning and expressions. We read faces and other kinds of body language; sometimes we take our time, and sometimes we jump to conclusions. What’s interesting to me is how AI has enlarged the very notion of intelligence. We once believed humans were the only species that exhibited intelligent behavior. Were we wrong! Now we’re looking at non-human intelligent behavior with more respect. Journalism about AI is so varied—the new Jerusalem; the end of civilization as we know it—that the public can think anything it wants. And yes, AI will change things profoundly. We just don’t know how yet.

insideBIGDATA: What do you think about the misuse of AI, specifically deep learning and its application to government controlled facial recognition, “deep fakes” videos, etc.?

Pamela McCorduck: I’m very disturbed by it, both as a citizen and as a spectator of AI research. A thread in my book is how naive I was—we all were—in the early days, when it seemed as if more intelligence could only be like more virtue. I’m especially disappointed with myself. I was a student of the humanities. How could I not have imagined that more intelligence would bring along all the usual misbehaviors humans are capable of? I’m embarrassed to say that it took a while to recognize that. I expect there will be technological fixes to the issues of deep fake videos and generative writing that fakes the writing style of authors and journalists (researchers are working on that like mad) but facial recognition (so flawed right now) is a blundering tool in the hands of governments. It will still be blundering when it improves technically. That’s really a political, not a technological problem.

insideBIGDATA: Can you pull out your crystal ball…Will AGI be a reality one day?

Pamela McCorduck: I can’t think why AGI wouldn’t be a reality one day. As for the rest, it’s more my wish than any projected beliefs: I’d love to see AI in the service of social problems, for instance alleviating poverty. Let’s assume earlier forecasters were correct, that AI will usher in a time of abundance. Will that abundance be evenly shared, or will the .001% grab the lion’s share? Can we bring the benefits of AI to the underserved rest of the planet? That’s a political, not a technological problem.

Sign up for the free insideBIGDATA newsletter.

Leave a Comment

*

Resource Links: