What are the criteria of intelligence of artificial systems? Turing Test (TT) was the starting point of testing intelligence. Now it is understood, that TT should be divided into separate tests measuring different aspects of intelligence. We propose two more tests 'Universal Player'' and 'Universal scientist'' aimed to show up capacity of artificial systems for making metasystem transitions. We consider some candidates for engine of metasystem transitions. One of them is 'self-organization'': automatic adjusting of parameters to critical values.