This spring, when the team submitted the results to IARPA, the evaluator team evaluated their performance. In June, the team learned who would move to the second phase of SMART. This phase runs for 18 months. Intelligent Automation, which is part of AFS, BlackSky, Kitware, Systems & Technology Research, and now defense company BlueHalo.
This time, the team needs to be able to apply the algorithm to different use cases. After all, Cooper points out that “designing a new AI solution from scratch is too slow and expensive for every activity you want to search for.” Can algorithms built to find construction find crop growth? He says it’s a major turning point, as it replaces slow-moving, man-made changes with natural, periodic environmental changes. And in the third phase, which begins in early 2024, the rest of the competitors have what Cooper calls “robust functionality,” that is, what they call both natural and anthropogenic changes. Try to get the job done.
None of these phrases are strict “exclusion” rounds, and there is not always one winner. Similar to the DARPA program, IARPA’s goal is to move promising technologies to intelligence agencies that can be used in the real world. “IARPA makes phase decisions based on performance on metrics, variety of approaches, available funding, and independent testing and evaluation analysis,” says Cooper. “At the end of Phase 3, there may be no teams or multiple teams left. The best solution is to even combine parts from multiple teams, or teams going to Phase 3. May not exist. “
IARPA’s investment often leaks beyond the program itself, and as science goes to the destination of money, it can also pave the way for science and technology. “The issues that IARPA chooses to do will get a lot of attention from the research community,” says Hoogs. The SMART team is allowed to use the algorithm for civilian and civilian purposes, and the datasets IARPA creates for its programs (such as satellite imagery labeled swarms) can be used by other researchers. Often published as.
Satellite technology is often referred to as “dual use” because of its military and civilian applications. In Hoogs’ view, the lessons learned from the software that Kitware developed for SMART can be applied to environmental science. His company is already working on environmental science for organizations like the US National Oceanic and Atmospheric Administration. His team has helped marine fishing services detect seals and sea lions on satellite images, among other projects. He envisions applying Kitware’s SMART software to flag deforestation, which is already the primary use of Landsat images. “How much of Brazil’s rainforest has been converted into man-made or arable land?” Hoogs asks.
According to Bosch Lewis, the automatic interpretation of landscape change has a clear impact on climate change research. For example, you can see where ice melts, corals die, vegetation changes, and land desertifies. Finding new constructions can show where humans are colliding with areas of the natural landscape, where forests are turning into farmland, or where farmland is replacing homes.
These environmental applications, and their spin-out to the world of science, are one of the reasons SMART was sought by the US Geological Survey as a testing and evaluation partner. However, the IARPA cohort is also interested in the findings for themselves. “Some environmental issues are very important to intelligence agencies, especially when it comes to climate change,” Cooper says. This is one area where the second application of dual-use technology is almost the same as the first application.