What is Universal Artificial Intelligence

Machine Learning: A Universal Formula for Intelligence

Paths to Universal Intelligence
  1. Computers are not only better at arithmetic and searching through databases than humans. They also outperform us when it comes to board and quizzing games and even drive their own cars.

  2. However, they provide these services in completely different ways than a human and therefore cannot provide any information about the human way of thinking.

  3. A new concept combines intelligence with the ability to compress data. This should make human and machine intelligence easier to measure and compare.

What is intelligence As psychologists have long known, this central term in their science is difficult to pin down, and current intelligence tests ask about a mix of skills. The American psychologist Howard Gardner went so far in his 1983 book "Frames of Mind" (German: "Farewell to the IQ") to claim that there is not a single intelligence, but many different forms. The idea flatters our ego, especially when we cannot boast of a high IQ: the more special intelligence there are, the greater our chance of shining in at least one of them. At the same time, Gardner's concept met with fierce opposition from experts.

The discussion about the essence of intelligence is being nourished, as recently machines have been performing what the whole world would have without hesitation described as intelligent up to now. When the "Deep Blue" computer dethroned the then world chess champion Garry Kasparov in 1997, it was generally regarded as an epoch-making event. At the time, some commentators pointed out - as a consolation, so to speak - that the programs in the Go board game only showed astonishingly mediocre performance. But now the machines can take on the top players here too. In October 2015, a system developed by the company "DeepMind" succeeded in beating the reigning Go European champion, and shortly afterwards, in March 2016, even the currently world's best player Lee Sedol.

This article is contained in Brain & Mind 7/2016

Artificial intelligence (AI) is now unbeatable in checkers. Since 1994 no one has succeeded in defeating the Canadian program "Chinook"; it pursues an optimal strategy that cannot be further improved. According to game theory, there must be such a winning strategy for all games in this class. Calculating them for chess, however, seems to be impossible for decades to come ...