The development of artificial intelligence is onefrom the priority directions of evolution of modern science and technology. New technologies are entrusted with solving problems that are beyond the power of man. However, at the present stage, the AI is at the initial stage of development and implementation, and therefore sometimes the technology faces unsolvable problems.
DeepMind (part of Google) is activelyThe AI developer conducted an interesting test that was no different from standard graduation tests for schoolchildren in English schools. The tests were subjected to specially trained in the school program in mathematical disciplines AI.
To the surprise of the experimenters, the AI failedthe task and scored 40 of the necessary points total 14. As a result of the analysis of unsuccessful testing, it was concluded that the AI requires a special formulation of the task, stated in a concise, specific form, while standard tests for people use both digital and textual tasks.
The solution of abstract problems without specifiedthe task required the analysis of the machine to highlight the necessary information. Often at this stage mistakes were made to ensure the wrong answer. Experts still have to spend a lot of effort in order to force AI to think and solve tasks on a par with the human mind.