(Photo Courtesy Elemental Cognition)
(Photo Courtesy Elemental Cognition)
The head of the development team that created IBM Watson, the computer that bested world “Jeopardy” champions in 2011, said last week that robots will be able to learn the way we do within five years.

Scientist David Ferrucci led the team that built Watson, which was specifically designed for the “Jeopardy” challenge. Watson was able to successfully synthesize huge amounts of data and learn from its mistakes.

Since then, IBM has invested $1 billion to adapt Watson for use in health care, finance, education, and other fields. Starting in November, IBMWatsonHealth will be used by GlaxoSmithKline in interactive online advertising designed to lead flu and cold sufferers to a personalized response (and presumably a suggestion to buy a Theraflu product). Watson stops short of diagnosis, according to Computer World.

Ferrucci went on to found Elemental Cognition, a company developing artificial intelligence that goes beyond finding the closest restaurant or providing simple answers to questions. He said artificial intelligence is evolving from synthesizing huge amounts of data to the next step: refining the information through context provided by language, culture and emotion to provide richer explanations. Real understanding and interactive communication are next. He sees the result as moving from a know-it-all Watson to a machine that can collaborate with humans by asking questions and interacting to pursue knowledge and solve problems. 

Ferrucci gave an example of a simple scenario between children where it is relatively easy for us to understand context such as age and weather, but difficult for artificial intelligence to do so: 

John and Mary are running a race when John falls and hurts his knee. Mary looks back, and even though she wants to win, she runs back to help John up.

Right now, artificial intelligence wouldn’t understand how to apply empathy and other factors. When it does, it will be able to come up with ideas on its own.

“So, a machine could say: I have a new idea for applying stem cells to organ regeneration. Can we talk about that?” he said.

He optimistically predicted that artificial intelligence would reach that capacity within the next five years. 

A PopTech audience member asked if he was concerned about robots taking over.

“No, I’m not worried about machines eating humans for lunch,” he said. 

The people controlling the systems are more of a problem, he said, since hacking the systems that control security — such as in driverless cars — could cause problems.