THE BLOG
12/16/2013 01:11 pm ET Updated Feb 14, 2014

A Computer Scientist's Unexpected Run-In With... a Shark

Click here to watch the TEDTalk that inspired this post.

"SHARK," yelled John as he dove into the water. Looking out, I saw our 200 feet of tow line with two of the world's leading researchers in marine mammal science hanging from it; the third was somewhere in the water, too. John, the first mate of our boat who was responsible for the safety of our swimmers, had disappeared under the waves.

"I wonder what fraction of the world's knowledge on dolphin communication could be lost in the next few minutes," I thought, uncharitably.

Sharks were yet another complication in our interdisciplinary research. As a computer science professor who specializes in pattern recognition and wearable computing, the CHAT (Cetacean Hearing Augmentation and Telemetry) project is one of the most difficult of my career. I have discovered that I am personally not well-suited to the open ocean, but I still signed on for the 10-day trip in the open Atlantic. I didn't want to miss the deployment of the new hardware, and long ago I resolved that the only way to design a wearable computer for a task is to experience that task firsthand. This year was the third my team had joined Dr. Denise Herzing in her summer research trips, but it was the first of the trips for me as I had been consumed with other work (Google Glass).

John, despite swimming in these waters since he was three years old, had spotted the largest shark he had ever seen. -- Thad Starner

The task seems simple enough at first: make a wearable computer that will allow Denise and her associates to use whistles to call for, and recognize calls for, different objects and people underwater. The hope is that the wild dolphins will observe these whistles and mimic them to "ask" for their favorite toys. Creating such a wearable computer is the first step in a larger goal: automatically record, analyze, and cluster dolphin vocalization patterns into fundamental units and allow a researcher to play back those fundamental units in new orders to the dolphins and observe the dolphins' behavior. The pattern discovery task is most interesting to my group; we have new algorithms that test well on known tasks like "discovering" the individual signs in videos of American Sign Language or the digits in spoken numbers. Perhaps our algorithms could help perform research in a domain where the answers are not known? That is how we found Denise. We had a tool; she had a problem.

First, however, we had to make a computer that could record and produce sounds in a dolphin's hearing range. In addition, the computer must thrive in a saltwater environment; handle baking in the sun on the deck of a boat; be usable in the water; be able to deliver its data to the main computers on the ship; and be maintainable on a tossing deck with the crew seasick. The first year was spent understanding what Denise needed. Using Georgia Tech's GVU Rapid Prototyping Lab, student volunteers iteratively refined the computerized guts for the custom-milled box. Testing that year involved chasing leaks, replacing parts burnt out by salt water, and realizing that it was very hard to buy or make an underwater speaker that was loud enough for our purposes. Another challenge was making a device that researchers could don and doff quickly.

By the second year, the system was ready to be tested in the field. However, the keyboard failed whenever the system submerged for more than a few seconds. I was told of the problem via satellite phone from the boat while at a social gathering of nerds in California. Suddenly, I had the help of a lot of interested friends. Was the problem from the electrical characteristics of being surrounded by salt water? Was it a leak in the cable? Was it a break in the connector? No. After much testing, it turned out to be a bug in the code that communicated between the microcontrollers in the main box and the keyboard. Unfortunately, while diagnosing the problem, we sprang a leak. The equipment was unusable, and we went back to the lab to iterate and be prepared for the 2013 season. We created new speakers; designed new amplifier circuits; added a simple display to our keyboard; milled three new, larger boxes; and made the boxes wifi-ready when out of the water. We thought we were prepared.

When John disappeared into the water to chase the shark, it was during the first test when everything seemed to be working properly. Denise and Adam were using the interface fluidly, and the recognition accuracy between the two CHAT wearable computers was near perfect.

"What type of shark?" Denise yelled, but John was still nowhere to be seen. We scanned the water, worried about continuing the test. I suddenly became concerned about what stray electrical currents we might be producing from the CHAT boxes. Sharks are very sensitive to electrical currents. Combined with swimmers in the water and loud whistles that are similar to those of dolphin calves, a shark might become curious.

John resurfaced and said, "It's OK now." John, despite swimming in these waters since he was three years old, had spotted the largest shark he had ever seen. It was a nurse shark - generally not a threat. However, when John had tried to scare it off, it had circled around and made another pass at our swimmers. John kicked it in the head to dissuade it... something I would never do!

Unfortunately, we never saw one of Denise's spotted dolphins that trip. We also never saw any of the bait fish on which the dolphins feed. Perhaps that is why the nurse shark was so interested in our researchers - there was no food in the area. It was the first time in 28 years that Denise could not find the pod. For the rest of the trip, we drilled Denise on the system and showed John and Captain Pete how to field-strip the equipment. However, it was not until the last trip of the season that the crew found the dolphins. We do have tantalizing data from the final trip, and we are analyzing it now. However, we will have to wait until next year to get enough time in the water with the dolphins to do the work Denise wants. In the meantime, my team is writing grant proposals and creating new computers and sound cards to try to create and capture whistles that are in the higher frequencies the dolphins prefer. Wish us luck; we need it!

Ideas are not set in stone. When exposed to thoughtful people, they morph and adapt into their most potent form. TEDWeekends will highlight some of today's most intriguing ideas and allow them to develop in real time through your voice! Tweet #TEDWeekends to share your perspective or email tedweekends@huffingtonpost.com to learn about future weekend's ideas to contribute as a writer.

Sign up for our email.
Find out how much you really know about the state of the nation.