Professors discuss fears and benefits of artificial intelligence

    Northwestern professors talked to students Monday night about technology familiar from science fiction movies, iPhones and IBM's Watson – artificial intelligence. Speaking from their various fields of computer science, philosophy, psychology and linguistics, the professors discussed the technology's future on a panel hosted by the Neuro Club.

    What is artificial intelligence?

    EECS Professor Doug Downey categorized two concepts of artificial intelligence. In one, he said, we imagine a computer that thinks and acts like a human does. The Turing test, invented by Alan Turing, is the ideal result of this definition. In the test, if a person can talk to and interrogate a computer without being able to determine if it's a computer or a human, the computer passes.

    Artificial intelligence could also mean a computer that can perform tasks with rational intelligence, but it might not necessarily be as recognizably human to us, Downey explained.

    Downey said humans have an internal experience while processing information that is vastly different from how computers process information. Our technology today works in a way that is non-human and makes non-human mistakes, he said.

    Psychology professor Satoru Suzuki said human intelligence can feel varying, and as a human thinks through a task they may struggle to find the right words, something that might also categorize artifical intelligence. "Can the computer feel the same intellectual struggle?" Suzuki asked.

    What can humans and computers do well?

    Klinton Bicknell, a linguistics professor, said that computers are very good at chess and other arenas where they have defined boundaries and rules to work in, but not as good at world knowledge. He asked the audience to finish the phrase, "The children went outside to..."

    Most people answer play, Bicknell said, but a computer would have an extremely difficult time answering the question. Language technology, like the type that makes Siri possible, is less advanced in this way, he said.

    People also gather information differently than computers and can keep a network of connections in their brains at the same time, according to Suzuki.

    "When things are unconstrained, we still behave in a productive way," he said. "You have a world of experience, a copy of that, in your brain network of 100 billion neurons."

    Transistors can send signals significantly faster than neurons, Downey said, but they can't be connected as densely or powered as easily in computers. Switching to an analog system to create artificial technology instead of using digital would be closer to how neurons work, he said.

    Should we fear artificial intelligence?

    Philosophy Professor Sanford Goldberg said that he stays up at night wondering whether there are levels of intelligence above human understanding already or in the future.

    "Are there kinds of intelligence that are so high-level that humans won't recognize them? Should we be worried?" he said.

    However, the current level of artificial intelligence is far from a self-aware world takeover borrowed from a science fiction novel.

    "Currently every artificial intelligence techonology that we've built is with a goal, like speech recognition," Bicknell said.

    Artificial intelligence technology is a complex field and each system that has been built can usually only handle its prescribed task, Downey said.

    "It's not that generalizable," Downey said. "If you take [the technology] outside its constrained domain it's not that intelligent anymore."

    Celebrities and scientists who call for a fear of artificial intelligence, like Elon Musk and Stephen Hawking, may need to trust in computer scientists, who know the field well, Downey said.

    What are the potential benefits of artificial intelligence?

    Currently a huge mass of data exists in the world without enough people to analyze it, Downey said, a job that computers do well.

    "We need a ton of additional people to help us learn from data," Downey said.

    For example, more medical records are becoming digitalized with "the potential to transform healthcare," he said, but this task is also incredibly challenging. Computing systems could tackle the data.

    Though many people fear computers will create more competition for factory jobs, Downey said the next sector to become embellished by computers is actually high-paying jobs with little motor or sensory skills necessary. Autonomous systems have the potential to help with a variety of tasks.

    "We want to run [artificial intelligence systems] in a way that's good for us, whatever that means," Downey said.


    blog comments powered by Disqus
    Please read our Comment Policy.