Protective scientific Management of perspective research programs of agency prepares for the beginning almost 4-year project on development of artificial intelligence which will be capable itself to study and improve.
In the new project speech doesn't go about modeling of a human brain and model start on the computer. This way in the short term is considered many scientists deadlock in view of exclusive complexity of a human brain and low-study of the principles of work of our reason. To DARPA plan to go on other way: to construct cars which will study by means of algorithms of probabilistic programming, that is to sift huge databases and to choose the best options of a solution. During this work the artificial intelligence will study and after a while will be able easily to solve simple problems. Revolt of cars and death of mankind begins with such projects in fantastic novels.
Experts of DARPA suggest commercial campaigns to develop in 46 months new technologies of probabilistic programming for self-training cars, in abbreviated form PPAML. The PPAML technology will help military to solve more effectively a set of analytical tasks which demand today huge human resources. In particular, self-training cars will be extremely useful when conducting investigation, supervision, recognition of speech, driving of the car, information sifting in search of valuable data, etc.
The hardware of artificial intelligence can be various: supercomputers on the basis of multinuclear processors, a network of ordinary personal computers and cloudy networks.
It is hard to say, some years creation of such difficult system of artificial intelligence are how real in the next. Now there is a set of automatic analytical tools which facilitate work of people but while any of them isn't capable to real self-training. While the highest achievement of technology of work with a database in fact there is a checking of archival pictures with the image arriving from a video camera. Therefore in DARPA emphasize that algorithms of PPAML will much more difficult, in particular, they can unite information of different type, independently formulate inquiry and on the basis of the obtained data to formulate the decision.