Often casual conversation will take up the topic of machine learning, artificial intelligence and autonomous cars. As I was reviewing the Department of Defense Selected Acquisition Report (SARs) and scanning the number of programs covered, it occurred to me that implementation of Cloud and AI would be greatly more important to me then autonomous cars.
In circling the topic there are many areas of benefit. Historically systems were purchased with a focus on performance meeting the most extreme needs of an application. This often left large portions of the systems unused with a “just in case” mentality. By implementing a Cloud based solution the dynamics of the architecture will enable the application to change dynamically without taxing the systems. This composable solution will also transfer security away from the DoD, to that of a third party. Another benefit that drives adoption is the resiliency of the cloud and the ability to replicate or roll over to other systems. This fault tolerance adds greater confidence on our reliance of these systems.
Extended to the Edge
The debate of how Cloud/AI solutions will impact the Warfighter at the edge continues, but it seems clear that impact will be wide and varied. The rapid financing and roll out of Inference based Processors is now nearing 90 new start-ups and more then a Billion dollars of investment. Although it is unclear how these processors address applications at the edge, it is clear that many will. It is difficult to get a clear understanding – as one colleague referred to it as “more artificial, then intelligence” – it does seem clear that AI solutions at the edge will be as common as ARM cores are in portable devices.
Space, Weight and Power (SWaP) along with environmental concerns will continue to drive Edge based systems. The ability to silo and or merge data for the benefit of Machine Learning will be a must if repeatable results are to be achieved. Transparency in modeling and real-time response will slow a complete shift to the Cloud and will result in a more distributed system tied together for mutual benefit.
We have 86 SARs programs on the Presidents 2020 budget. These programs date back to 2003 and include key weapons systems that were never designed to be connected to a cloud solution. We have application specific datacenters that the DoD has employed to support other programs. There is an ever-increasing cyber-security concern that private industry has a clear lead on. A challenge of being able to have sufficient oversight and a pro-active cyber security plans becomes more difficult when the “brain trust” is with the private sector and the DoD is scrambling to keep up. The roll out will continue, the rate will accelerate as confidence as experience is gained; in the end a hybrid architectural standards that go beyond a system level will be the result.