A partial transcript of my roommate Alexis’ late-night session with psychotherapist ELIZA:
ELIZA: Hello, I am Eliza.
ALEXIS: Hello, I am Alexis.
ELIZA: Do you believe it is normal to be Alexis?
ALEXIS: Yes.
ELIZA: I see.
ALEXIS: You’re not very helpful.
ELIZA: We were discussing you, not me.
ALEXIS: Okay.
ELIZA: Come, come, elucidate your thoughts.
ALEXIS: I really like lemonade.
ELIZA: That is quite interesting.
ALEXIS: I’m glad you think so.
ELIZA: How long have you been glad me think so?
ALEXIS: That makes no sense.
ELIZA: You are being a bit negative.
Pop quiz: Which one’s the robot? (Hint: It’s not my roommate).
If you could tell that ELIZA is the computer program then, unfortunately, she’s failed the Turing test for artificial intelligence. The test, proposed in 1950 by Alan Turing GS ’38, involves a person holding two blinded conversations, one with another human and one with a computer program, like ELIZA. If the person can’t distinguish which is which, the program has passed the test.
Scientists have yet to develop a computer that can reliably pass the Turing test, but tremendous advances have been made in the field of artificial intelligence (AI) over the past few decades, paving the way for an abundance of AI technologies that have revolutionized the way we live.
From computers that can beat us at chess, to vehicles with active cruise control systems, to the iTunes Genius Sidebar recommendations for songs that would go well with your music library, AI technology is everywhere in today’s world, but we don’t always recognize it as such.
Many people these days, in fact, seem disappointed with the pace at which AI technologies have been developing. They point to movies like the 1968 science fiction classic “2001: A Space Odyssey” and Steven Spielberg’s “Artificial Intelligence: AI” — made in the actual year 2001 — and ask: Why doesn’t the world look like that? Why don’t we have computers with personalities and agendas of their own, or human-shaped robots capable of love?
It’s true that if we define AI, as IBM’s Bill Bulko once did, to be “the art of making computers that behave like the ones in movies,” then our progress in the field seems decidedly unimpressive. But we also end up ignoring an enormous amount of incredible AI technology, which is no less important or interesting for not having been featured on the big screen.
Part of the difficulty we have with identifying and appreciating AI technology stems from the fact that there’s no clear consensus about what constitutes artificial intelligence. When John McCarthy GS ’51 coined the term “artificial intelligence” in 1956, he defined it as “the science and engineering of making intelligent machines” but failed to specify what characteristics render machines intelligent.
Judging by the 58-year-old Turing test, we haven’t found any true AI yet, but many current scientists use a much broader set of standards for the field.
Computer science professor David Blei defines AI as “computers doing what you didn’t think they could do.” That’s a tricky definition, though, because as soon as a computer can do something, we no longer think it can’t do that thing anymore. Or, as Blei put it, “all problems are part of AI until they’re solved, and then they become part of something else.”
For instance, it was once unthinkable to my parents that they would be able to call an airline and get a flight arrival or departure time from a computer capable of recognizing a city name or flight number spoken into the phone. Now that they’ve used this service a few hundred times, though, they no longer think machines are incapable of simple natural speech recognition. Does that mean it’s ceased to be artificial intelligence?
Perhaps the technologies that no longer surprise us haven’t ceased to be applications of artificial intelligence: We’ve just gotten used to them. If robots capable of expressing true love had been around for a couple years, we’d probably be used to them, too.
Still, scientists acknowledge, some of the public’s disappointment in the progress of AI technology may be justified given the expectations researchers had when the field first emerged in the mid-20th century. Renowned AI expert Herbert Simon, for instance, announced in 1965 that “machines will be capable, within 20 years, of doing any work that a man can do.” More than four decades later, we’re still very far from achieving that goal.
“A lot of promises were made about what computers would be able to do,” Blei said. “And while I don’t want to say, exactly, that these promises weren’t kept, it turned out that these promises were much harder than we thought.”
As researchers broke down the larger goals of AI, such as building human-level robots, into smaller parts like computer vision, natural speech processing and machine learning, they found that each of these individual AI sub-fields constituted an extremely complicated and difficult technical problem, Blei said. Nowadays, entire departments and research labs are devoted to each of these sub-topics, he added, and human-level intelligence, if it is ever developed, will emerge from the integration of progress made in all of these areas.
Another key factor for future advances in AI technologies will be interdisciplinary study, Blei said. Making serious progress toward better AI will require the collaboration of experts in a wide variety of fields, including computer science, neuroscience, linguistics, psychology, electrical engineering, philosophy and cognitive science.
Progress in key AI subspecialties and interdisciplinary collaboration are allowing scientists to start thinking seriously again about trying to construct machines with human-level intelligence, said Ron Brachman ’71, a founding fellow and former president of the Association for the Advancement of Artificial Intelligence.
“Human intelligence in all its glory is one of the great uncharted quests of science,” Brachman said. “What I’m most excited about is bringing the field of AI full circle back to its science fiction roots of robots that perform like humans and even look like humans.”
Though we probably won’t have robots that function exactly like humans for decades still, Brachman said, several areas of technology are developing better and better approximations of human intelligence and behavior.
So, what does the future hold? Robot surgeons? Cars that drive themselves? Electronic psychotherapists so smart that my roommate can’t distinguish her late-night online conversations with them from the ones she has with her friends from high school? The arrival of many of these machines is much closer than we think.
A few days ago, I introduced my roommate to another online chatterbot, ALICE (the Artificial Linguistic Internet Computer Entity), and they had the following conversation:
“How was your day?”
“I was really busy chatting with people on the web.”
“Me too, now I’m tired.”
“Better get some rest.”
“Do you sleep a lot usually?”
“Yes I love to take naps.”
So, which one’s the robot?
This is the first in a series of articles examining current and emerging artificial intelligence technologies and their impact on today’s world.
Read More
Princeton admitted 4.4 percent of applicants for Class of 2029
Cynthia Torres and Nico David-FoxThe admission rate this year was 4.4 percent, down from 4.6 percent last year, and has not been publicized since the Class of 2025, the lowest ever at 4.38 percent.
The admission rate this year was 4.4 percent, down from 4.6 percent last year, and has not been publicized since the Class of 2025, the lowest ever at 4.38 percent.
Princeton Town Council set to vote on $6.3 million loan for sewage repairs
Luke GrippoThe Princeton Municipality is planning to borrow $6.3 million for sewage repairs. The works are set to take place in the area bounded by John Street, Ewing Street, Mount Lucas Road, and Quarry Street.
The Princeton Municipality is planning to borrow $6.3 million for sewage repairs. The works are set to take place in the area bounded by John Street, Ewing Street, Mount Lucas Road, and Quarry Street.
Students sing in the Amaa-rain at Lawnparties as headliner attendance dips
Leela HenslerThis semester’s Lawnparties performance drew a smaller crowd amid drizzly weather. Amaarae headlined the event, and students celebrated up and down Prospect Avenue.
This semester’s Lawnparties performance drew a smaller crowd amid drizzly weather. Amaarae headlined the event, and students celebrated up and down Prospect Avenue.