TEDx Peachtree sheds light on theme of ‘Illuminate’

There are more than a dozen presenters at TEDx Peachtree. Here is a sampling:

Keith McGreggor is the director of VentureLab at Georgia Tech. He is also the lead instructor of the National Science Foundation's Innovation Corps program.

Thad Starner is a pioneer in the world of wearable computing, a professor at Georgia Tech's School of Interactive Computing and a technical lead for Google Glass.

Steven Cantor, who performs as HeaveN Beatbox, grew up in Alpharetta and has been beatboxing — vocally replicating, creating and mimicking music — since age 14.

Shannon Zirbel is a doctoral student of mechanical engineering at Brigham Young University. Her doctoral research has focused on applying compliant mechanism theory to space technology with the objective of creating lightweight, low friction mechanical solutions that enable greater advances in space exploration.

Genna Duberstein is a multimedia producer for heliophysics at NASA's Goddard Space Flight Center. Her project, Solarium, is a large-scale video installation that will be featured in Georgia State University's Window Project this November.

Chantelle Rytter is an Atlanta-based artist best known for creating the Atlanta Beltline Lantern Parade, an annual illuminated parade that draws more than 10,000 people.

How do our eyes make sense of the world?

An untutored brain — or a computer — could go nuts with the visual input we process every moment, like trying to sort a swimming pool full of broken stained glass.

Georgia Tech instructor Keith McGreggor, a pioneer in artificial intelligence, says teaching our smart machines how to understand the visual world is a critical way to make them even smarter.

McGreggor will be among the speakers Oct. 17 at the TEDxPeachtree conference at the Buckhead Theatre, a day-long program organized around the theme “Illuminate.”

Among 12 others addressing the conference will be Chantelle Rytter, organizer of the Atlanta Beltline Lantern Parade, and Genna Duberstein, lead multimedia producer for heliophysics at NASA’s Goddard Space Flight Center.

The TED conferences, which began in California in 1984 and became a yearly event in 1990, started as a way for Silicon Valley tech gurus to chat with each other about advances in “Technology, Entertainment and Design.” The conferences spawned smaller TEDx events around the country. The TED approach to presenting information is based on the idea that informed teachers and leaders can compress their most passionate views into an 18-minute talk that will change the world.

Atlanta’s TEDxPeachtree conference began in 2009; now smaller TEDx Salons take place four times a year in Atlanta, culminating in the TEDxPeachtree conference each October. TEDx Peachtree organizer Jacqui Chew said McGreggor, like the best TED speakers, has the knack of “taking a complicated, inaccessible topic and making it understandable for the rest of us.”

McGreggor’s tech credentials are impeccable. An early and enthusiastic programmer, he grabbed Apple’s attention by writing the first 3D drawing program for the Macintosh, then moved to Cupertino, Calif., and ran the graphics group for Apple.

Six years ago McGreggor returned to Atlanta to head Georgia Tech’s VentureLab, a nursery for start-ups aimed at commercializing the bright ideas of Tech students and faculty. (More than 100 new enterprises have grown out of the tech incubator.)

McGreggor also pursued his longtime interest in artificial intelligence, focusing on the visual. Plenty of scientists have worked on visual recognition in computers, creating machines that can identify objects. McGreggor is designing programs that recognize patterns and see novelty — or subtle change.

Humans do this as easily as breathing, instinctively going wide-angle or telephoto as the situation requires, and seeing the patterns at every scale. “You zoom in, you zoom out, you do this effortlessly, and you do it almost unconsciously,” McGreggor said.

Humans also effortlessly choose the level of abstraction with which they view the world. They can see how one leaf looks like another leaf, and how the branching veins of a leaf resemble, at the next scale up, the branches of a tree. McGreggor calls this “fractal encoding.”

“Instead of using (artificial intelligence) to think about the world geometrically, my research focuses on seeing the world fractally. When we do it in that fashion, our ability to notice novelty and to choose the level of abstraction happens automatically.”

While the concept may sound abstruse, the reality of smart devices — self-driving cars, for example — requires that our devices be smart in the same way that we are. That means visual intelligence.

“Our intimate computing needs to react more and more like us,” said McGreggor, “like family, like your best friend, like you’re closest ally.”