Getting beyond maps and imagery has always been a goal of GEOINT, said Gilman Louie, a partner at Alsop Louie Partners in San Francisco, however it now has a different level of importance.
The world is moving away from single sensor exquisite collection to one of real sense-making, by taking advantage of not only the networks but the ability to process information and make decisions in real time, rather than analysis time.
Louie is one of 15 Commissioners appointed to the National Security Commission on Artificial Intelligence (NSCAI), which found that if the United States continues at the current rate of innovation, the country will be unprepared for the next era of artificial intelligence (AI) and machine learning (ML) compared to competitor nations.
Intelligence will be the biggest benefactor of AI, more than any other branch of government capability. The NSCAI recommended that the country build purpose-built IC platforms to take advantage of all the information with temporal geo tags, in order to amend how the country approaches intelligence.
“We need to rethink how we think about geospatial information, to be much more machine centric, much more computable ready and much more about flows of information over periods of time to get unique understanding,” Louie said.
Today, we are limited by Mark 1 sensors, or our eyes. “If we think about all the things we do to process information, it’s all to present that information in front of our eyes so our brain can process it,” Louie explained. What you see is what you get.
Our computing and collection capabilities have been to the left of the Mark 1, which will reduce a spectrum of information into pieces that we can understand. AI promises to change this. Deep mind breakout processing will take raw information and process it using the latest in reinforcement learning.
The applications of next generation GIS, which have evolved alongside humankind, will not have human controllers. Instead, the human will only act as a supervisor as the sensors process information. Currently, most information is preprocessed and stored in a way that humans understand, so this must change. In order to do so, you must have machine-centric data. “If my sensor sees thermal images then my data, to identify what I am seeing, it needs to be in a format that’s as close to what the sensor is actually sensing,” Louie said. Machine ready data set will also speed up processing times, as no computations need to be completed. Latency can be fatal, so real time processing is essential.
According to Louie’s expertise, GEOINT must start building models of the plethora of sensors available to collect. There also needs to be more computable infrastructure that’s necessary to process this information. It’s pertinent that this process begin now, as other countries are already integrating this kind of technology.
To build a common AI framework that’s interchangeable across domains for the intelligence field, we must invest in the necessary research infrastructure. The systems that are created must be tested for potential bias or blind spots. They must also “be resilient to when we don’t have comms,” Louie explained. Being too dependent on comms will leave the country vulnerable to those whose first target will be our comms systems.