Friday, June 21, 2024
Friday, June 21, 2024
The Navy has learned a lot in the eight months since it began intercepting missiles in the Red Sea, but the service is only able to leverage a fraction of the sensor data its warships are collecting due to limitations in shipboard data storage.
According to Lt. Artem Sherbinin, the chief technology officer for the Navy’s Task Force Hopper -- which was established in 2021 to integrate artificial intelligence and machine learning across the surface fleet -- each Navy warship produces about 150 terabytes of raw sensor data each day.
Across a fleet of 258 ships and 11 aircraft carriers, “that is more data than any vessel or any physical piece of hardware in the Department of Defense,” Sherbinin said today at the Nexus 2024 symposium in Washington.
“We are capturing miniscule percentages of that data, which means we're not able to apply that to train any sort of model,” he said.
Though the Navy is investing in expanding shipboard storage, it has a long way to go, Sherbinin continued, saying it is difficult to convince the Pentagon to spend large sums of money on storage and computing for ships.
However, the Navy has learned to extract this data faster since its ships began engaging anti-ship cruise missiles launched by Houthi rebels in October.
At the time, it took weeks or even months to extract sensor data from vessels, Sherbinin said. Now, the service has significantly reduced this timespan and is pushing to collect the data in as little as 24 hours, he continued.
Sherbinin also said software is the Navy’s best bet for improving fleet capabilities before 2027 -- the year by which the Defense Department believes it must be prepared to defend Taiwan from a Chinese invasion.
If this countdown is accurate, “that means the cycle time we have to deploy software technology is much shorter and we also know that new hardware isn’t coming,” he said. “Ships take years to make. Sensors for those ships take years to make. Software is all we have.”