US military must urgently rethink machine learning strategy

- Tell the sovereign that the British do not clean their rifles with brick: let them not clean them, but God bless the war, they are not suitable for shooting - Lefty distinctly pronounced, crossed himself and died.

The public report , prepared in July 2017 at Harvard, shows how unprepared the US military looks like in such an area of ​​artificial intelligence as in-depth training.





Recently it became known that Google (corporation of good) is engaged in the analysis of video images from military drones. This project is called Project Maven and was proposed in April 2017. Interestingly, Eric Schmidt, the former chairman of Alphabet, and the current chairman of the Defense Innovation Council, DIB, organized cooperation with Google on this project.



As part of the project, by the end of 2017, it was planned to implement advanced algorithms on state computer platforms designed to highlight the objects of interest of 38 classes in photo and video images. Colonel Zukor, who submitted the project, described the algorithm as 75 lines of Python code.

')

Greg Allen, author of the Harvard report, said his insider opinion that prior to the Maven project, no one in the Department had any idea how to properly acquire, implement and test artificial intelligence. Let me emphasize this. A year ago, no one in the US Department of Defense was aware of the above.



For the sake of justice, we note that there are several people in DARPA who have an idea of ​​deep learning. In 2017, DARPA procured a number of modern projects: “Context Adaptation” , “Explainability” , “Biologically Inspired Architecture” .



There is clear evidence that deep learning is not taken seriously in the US defense industry. Here is a description of a vacancy published on March 15, 2018. The US Army is seeking a PhD in deep learning with a $ 52,000 salary salary. Maybe they forgot zero?



The F-35 aircraft costs at least 300 million dollars. For comparison, Google acquired DeepMind for 500 million. But from the point of view of modern progress in deep training, the F-35 looks the same antique military system as cavalry on the eve of the Second World War. The F-35 is so expensive, apparently because it is very complex. However, a swarm of 10,000 drones at a price of 1/300 of the cost of one F-35 will probably be just as effective on the battlefield.



This is the problem of preparing for wars of the future based on the technologies of the previous era. We cannot continue to spend 300 million on an unlimitedly complex system that risks becoming outdated on the very first day of putting it into service against systems inspired by artificial intelligence.