|
Getting your Trinity Audio player ready...
|
Stark against the white Antarctic terrain, a small, yellow robot moves roughly one centimeter per second. Too fast, and it might spoil its mission: to gather data on a colony of emperor penguins that call the Southern Ocean home.
Artificial intelligence recognition tools help the robot, called ECHO, spot when the penguins start to get spooked. A green box over the penguin signals that it’s calm, but a red box signals stress.
To keep the boxes green, ECHO must approach very slowly, said Daniel Zitterbart, associate scientist with the Woods Hole Oceanographic Institution and the project lead on ECHO.
“There is no AI that can help me go out in the field and catch a penguin. To learn about the natural world, you still need to go into the natural world.”
“The good thing is a robot doesn’t get bored,” he said.
AI tools are quietly transforming how ocean scientists gather and analyze information from the watery deep. Researchers from WHOI, the University of Rhode Island, and the UMass Dartmouth School of Marine Science and Technology (SMAST) have either employed AI tools for decades or are looking to incorporate the technology into future projects, exponentially increasing the volume of data they’re able to collect.
These new applications, often going hand in hand with work seeking to better understand climate change, come as the technology faces backlash from environmental advocates concerned about the amount of water and electricity necessary to sustain it.
In some ways, researchers say the technology is a boon in an era when scientists need to do more with less federal funding. But these researchers are also quick to point out that AI tools will never replace human scientists — and the public money that funds them.
“There is no AI that can help me go out in the field and catch a penguin,” Zitterbart said. “To learn about the natural world, you still need to go into the natural world.”
Cold-approaching penguins
Thousands of miles from Antarctica, most local fights around AI revolve around the data centers being built to support the technology. These centers have multiplied in recent years to accommodate anticipated demand for AI, and with them come concerns that their construction will strain water and electricity reserves.
No data centers have been proposed in Massachusetts thus far, but Massachusetts Department of Energy Resources Commissioner Elizabeth Mahony expects the day to come sooner than later.
“That is on the mind of anybody working in energy right now,” Mahony told The Light in an interview in July.
Amid this environmental scrutiny, AI is also being used to monitor ecosystems most vulnerable to climate change.
As it charms penguins, Zitterbart’s ECHO robot gathers data that serves as a barometer for one of the most consequential regions on the planet.
“The Southern Ocean takes up [a lot] of our CO2,” Zitterbart said. “It’s cold, it’s vast, it slows down the heating up of the world a lot…. And so understanding how well that ecosystem does down there directly affects us.”
In effect, Zitterbart said, if the penguins are doing well, so is the ocean.
Closer to home, Heidi Sosik, a senior scientist in WHOI’s biology department, has used AI technology to monitor plankton in the North Atlantic for over two decades. Her early work employed technology originally used to identify fingerprints, retooling the software to recognize the tiny, often microscopic species that serve as the bedrock of ocean health.
“At the time, I thought I was overwhelmed because I had 300,000 images,” Sosik said. “Now I have 300 million.”
Unlike the large language AI models most people are familiar with, both of these tools were decades in the making, Zitterbart and Sosik said, adopted by scientists gradually and cautiously as technology advanced. Still, Zitterbart doesn’t trust most AI products being rolled out today, he said, and it will be many more years before he incorporates them into his work.
In all applications, Zitterbart and Sosik said they make sure that no machine makes any final calculations or decisions without a human in the loop. In another of Zitterbart’s projects, WhaleSpotter, AI helps boats spot nearby whales in order to reduce vessel strikes, one of the leading causes of marine mammal death. The AI identification helps captains act quickly, but a human scientist goes back to verify information after the fact.
“It’s a tool like any other,” Sosik said. “It’s only as good as its user.”
The technology can recognize multiple species of whales — with help from the company’s researchers — including the North Atlantic right whale, a highly endangered species in New England’s waters. Vessel strikes account for more than a third of right whale deaths, according to WHOI.
Surveying fish stock
In New Bedford’s South End, SMAST fisheries researchers go through thousands of hours of video survey data taken at sea, hoping to catch a glimpse of starfish, scallop, or other species of interest.
Since the 1990s, SMAST’s HabCam survey has filmed the ocean floor using drop cameras to monitor scallop growth in prime areas like Georges Bank, a fertile scalloping bed located 62 miles offshore from New England. The Atlantic cod video trawl survey uses cameras submerged underwater in the Gulf of Maine to keep tabs on a fish population still coming back from the brink.
Typically, the thankless task of combing through these videos falls on grad students, but in recent years, Kevin Stokesbury, the dean at SMAST, said his team is using AI “as much as possible.”
“The sheer volume of having someone sit there and review a video is so huge,” Stokesbury said.
But the technology has its drawbacks, Stokesbury said.
“Is the fish you saw in the last frame the same as the fish you saw in the next frame? And then what species is it?” he said. “It’s quite a tricky AI problem.”
Kris Lewis, an assistant professor at the University of Rhode Island Graduate School of Oceanography, sees the greatest potential for AI in countries she’s worked in such as Indonesia, where manpower and resources are slim.
Following the money
Tech companies like Google and Amazon are also incentivizing scientists to use their AI tools through grant opportunities. As federal grant options dry up, Lewis is taking a look at private grants she wouldn’t have previously considered, she said, including one Google Impact grant and one grant from Amazon.
The two grants would each fetch about $500,000 to $1 million — less than the up to $2.4 million grants Lewis might get under the National Science Foundation, Lewis said.
Although there were no stipulations around how Lewis would have to conduct her project should she get the Google Impact or Amazon grants, she was concerned about a lack of transparency around who decides which project to fund.
“We know at NSF and NOAA that there are experts that they bring in,” Lewis said, but she didn’t know who would ultimately review her application at Google or Amazon.
At WHOI, researchers are more averse to seeking out funding opportunities from private companies, although the institution has long relied on foundational or philanthropic grants. The institution follows a code of ethics that applies to all of their funders to ensure a level of academic freedom. Zitterbart, for his part, does not work with private companies because they often take ownership of any resulting intellectual property, he said
The current funding landscape — and the pressures it creates — is especially heart-breaking for Lewis, since she thought working in Rhode Island would be different, she said. Lewis gave up a position at the University of Central Florida to move to the University of Rhode Island in 2023, fleeing what she viewed as government overreach over the topics she was allowed to teach.
Losing federal support for her work felt like another betrayal, she said.
“I left Florida to escape that, to come to New England to try to get a safer place to do the work that I want to do,” Lewis said. “Never in a million years did I think that I couldn’t get out.”
Email Brooke Kushwaha at bkushwaha@newbedfordlight.org.
Nathan Collins is an investigative journalist and a Report for America corps member, covering the fishing industry, oceans and the blue economy. Report for America places talented journalists in local newsrooms to report on under-covered issues and regions. Email Nathan Collins at ncollins@newbedfordlight.org.
