A Short Story of
The desire of mankind to create rational beings similar to themselves goes back far into antiquity. The pursuit of such beings by the arts and sciences has always been accompanied by fears of disastrous monsters and hopes for godlike supermen. Therefore, the advances of scientists are always examined to determine the extent to which their results can be of benefit to mankind.
The first programmable calculating machines quickly achieved, and exceeded, human intelligence – in certain, clearly defined task areas. When, in 1997, IBM’s Computer, Deep Blue, defeated the world chess champion Garri Kasparov, the machine’s main achievement was to calculate millions of possible moves in seconds. The IBM computer was not able to learn from its opponent or “think” about its position on the chessboard, but …
In the years that followed, computer performance increased enabling enormous amounts of data to be processed. This data was used to develop neural networks and enable them to actually learn. In a game of Go against the world champion, the self-learning system AlphaGo was put to the test; Go is far more complex than chess and defeating a human professional is impossible using Deep Blue’s methods alone. During the game, AlphaGo drew conclusions from his opponent’s past moves, predicted his future moves, and won!
How should a system be built to enable it to apply its abilities to a previously unknown task? Libratus, an AI-supported computer program, conquered the rules of the card game poker by competing against itself and then against a team of professional human players. By learning overnight from its own weaknesses from the previous day, it succeeded in winning the tournament. The real success: this strategy is not limited to poker.
Today, AI-based systems are able to solve some of the problems of our real world and find their application in areas such as media, medicine and mobility. The fears and hopes of people around these applications are as great as the potential of AI.