AI Explained as 'Intelligence Augmentation'

UA computer science professor Mihai Surdeanu tells a College of Science lecture series audience that artificial intelligence lacks an in-the-world human dimension.

By:
Robin Tricoles , University Communications
Computer scientist Mihai Surdeanu speaks at the College of Science lecture series in Centennial Hall. (Photo: Bob Demers/UANews)

 

Rest easy, people. The recent revolution in artificial intelligence won't be spawning computers hell-bent on doing in the human race. Instead, artificial intelligence, or AI, will be working with us and not against us, in what Mihai Surdeanu terms "intelligence augmentation," or IA.

Surdeanu, a University of Arizona associate professor of computer science, delivered Monday's second of six talks at Centennial Hall as part of the College of Science lecture series. His talk, titled "The Minds of Machines," touched on the differences between artificial intelligence and the human mind — and where artificial intelligence is headed next.

The series, titled "Humans, Data and Machines," is designed to delve into various aspects of the revolutionary social change now underway with the convergence of the physical, digital and biological worlds. 

Surdeanu said that one of the starkest differences between the human mind and artificial intelligence was noted by 20th-century philosopher Martin Heidegger, whose work focused on what it means to be.

"Essentially what Heidegger is saying is that what we are, and what the world is, are mutually interdependent," Surdeanu said. "There's no objective world apart from our experiences of it, just as our experiences cannot be separated from the world in which they occur."

In other words, "Existence is interpretation, and interpretation is existence," Surdeanu said.

In fact, Heidegger went to the trouble to coin the word "Dasein" to describe humans as being in the world — meaning that humans have the capacity for self-interpretation, unlike computers, thus limiting computers' capacity for humanlike intelligence.

For example, when it comes to learning human language, artificial intelligence has a limited ability to grasp syntax, sarcasm or metaphor. Another example is decision making. Artificial intelligence can make decisions, but it can't yet explain why it made the decision it did, Surdeanu said.

"We shouldn't trust in decisions without an explanation of why they're happening," he said. "In the future, we will have systems that have the ability to explain themselves, which is better than most of us can do."

Of course, artificial intelligence can help humans with monumental tasks, such as quickly mining and sorting data that humans can later use for research that only humans can do.

That's why Surdeanu said he prefers "intelligence augmentation" to "artificial intelligence." After all, artificial intelligence originally was designed to complement us, not to replace us.