DARPA intends to revolutionize machine learning

Almost every news from DARPA related to robots and artificial intelligence is inevitably accompanied by snapped comments about Skynet. But this time they will be surprisingly appropriate. The Agency's new research program is devoted to probabilistic programming for solving advanced machine learning tasks (Probabilistic Programming for Advanced Machine Learning or PPAML). According to program manager Kathleen Fisher, DARPA intends to do as much as "Do for machine learning what the advent of high-level languages ​​50 years ago did for programming in general."

Machine learning algorithms are already widely used in consumer technologies - anti-spam, speech recognition, car robots and for analyzing gigantic amounts of data in medicine or finance. Naturally, the prospects of machine learning are interesting for the military as well. At the same time, there are no universally accepted universal tools for creating intelligent systems. Because of this, you have to constantly invent bicycles, over and over again implement algorithms similar to two drops of water, build an architecture from scratch.

The set of approaches and paradigms used in machine learning is called probabilistic programming.". Tools, libraries and programming languages ​​for him have not yet left the walls of universities, and their list is quite short. DARPA intends to change this situation.

Among the goals of the program are to radically reduce the complexity of creating machine learning systems, lower the threshold for entering intelligent programming into programs, and improve basic machine learning algorithms, the maximum use of modern hardware technologies - multi-core processors and GPUs, cloud computing, the creation and standardization of AP I for linking machine learning infrastructure elements into a single system

PPAML program is designed for 46 months. Detailed conditions and requirements will be announced on April 10 at a presentation in Arlington, Virginia. For now, you can download the PDF with a brief description of the program.