UPDATED: WILMINGTON–AI experts from across the state gathered at the Cucalorus Connect conference in Wilmington Nov. 8-9. Aimed at exploring the interface between technology and creativity, the conference was divided into several tracks, including HealthTech, FinTech addressing the future of banking, Community Brew showcasing trends in social entrepreneurship, the Startup track featuring futurist Charlie Fink, and GigTech with an emphasis on AI and the sharing economy. Each track included several breakout sessions conducted by industry leaders, with participants coming together for social gatherings sponsored by the NC Tech Association and the NC Biotechnology Association.

Artificial Intelligence was a main focus of the GigTech panels. Curry Guinn, PhD, department chair and professor of computer science at UNCW, and Julian Keith, PhD, department chair and professor of psychology at UNCW, teamed up to discuss progress in the AI field to reproduce human reasoning and learning capabilities.

Dr. Keith presented the contribution deep mind research can make to the development of AI systems. Diagram from deepmind.com

Mike Orr, retired from IBM after a career that spanned developments from hardware the cloud, with two patents to his credit, moderated Guinn and Keith’s panel, titled AI: The Third Generation of Computing. In his introductory remarks, the 38 year IBM veteran offered a “quick trick” to help determine if an AI system is really helpful.

The “Magic test”

“I call it the magic test,” he said. “Replace machine learning, neural network, AI with the word magic… do your sentences still make sense? If so, you probably aren’t getting anywhere.”

Orr, who worked on both IBM Watson and Watson Health, also fielded a question about the Facebook chatbots which reportedly developed their own language, unintelligible to humans, last year. “The researchers freaked out and shut it down,” he responded. “They were in time to keep it from spreading… they hope.”

IBM veteran Mike Orr moderated the Third Generation of Computing panel. Photo by R Wright copyright Capitol Broadcasting A.R.R.

Bill Stuart, a senior data scientist at GE, explored the major challenges facing companies using AI applications in his panel, AI: It’s All About the Data. Data scientists, he says, spend the majority of their time – some 80 percent – gathering and organizing data, and not enough time cleansing the data and addressing biases that may affect ML results.

“It’s critical to determine why our algorithms are making decisions,” he said. “One study looked at how a program was deciding whether a picture was of a wolf or a husky. An analysis found that the algorithm gave a wolf result to any picture with snow in it.”

GE data scientist Bill Stuart spoke on the problems of dirty data. Photo by Renee Wright. copyright Capitol Broadcasting A.R.R.

Stuart said the paradigm hasn’t changed significantly since the 1950s. “We still haven’t figured out how to get a computer to create humor or creativity or intuition,” he said. “That hasn’t been solved.”

Among the hottest topics on the GigTech track was the rapid advance of machine learning and its effects on the economy.

Curry Guinn, who chairs the computer science department at UNCW, told the crowded seminar that changes predicted in the ’60s and ’70s are finally happening now.

Chair of the UNCW computer science department, Curry Guinn, PhD. Photo by Renee Wright.copyright Capitol Broadcasting A.R.R.

“In the next 5 to 6 years we’ll see some pretty radical changes,” he said. Robots in factories, bipedal robots, IBM Watson, Amazon Go and driverless cars are among the major advances he named.

He added, “With the proliferation of these AI technologies, many jobs for humans will disappear. Occupations particularly likely to be taken over by AI include sales people and cashiers, assembly line workers, secretaries, bookkeepers and tax preparers.”

Guinn related a story of a factory with only two employees, one man and one dog. “The man will be there to feed the dog. The dog is there to keep the man from touching the equipment. I actually saw a factory like this, with just one employee,” he said. “A janitor came in every couple of weeks, so that made two.”

Julian Keith, PhD, chair of UNCW’s psychology department, discusses a slide of future AI expectations. Photo by Renee Wright. copyright Capitol Broadcasting A.R.R.

However, there’s still one problem that AI developers haven’t been able to conquer. “We’ve got a problem with speech recognition and conversational language processing.

“Anyone who has tried to have a conversation with Siri or Alexa knows what I mean,” Guinn said. “And there’s no obvious solution in sight, certainly within the next five years. Natural language in one of the big barriers.

“So if you have a job where you have to communicate with other humans, your job is safe. If it doesn’t, you’re in trouble. But if the language barrier is surmounted, well…”