When angry birds sit on slingshots and seconds later are fired at green, swinish opponents, is "angrybirds"-time. Since 2009, the computer game has been conquering the screens of smartphones, tablets and pcs in its different variations.
While the fan community in germany is anticipating the release of the movie "angry birds 2 on 19. The bamberg student team "bambirds" has been looking forward to september has already successfully shot down its birds and pigs respectively. At the 8. AI-birds world championship, which was held in august as part of the most prestigious conference on artificial intelligence, the international joint conferences on artificial intelligence, took place in macao/china.
However, the bambirds did not execute the moves themselves, but rather an AI program of the same name that they had developed. Because that’s what it was all about: the search was on for the most powerful program that could independently master previously unknown levels of angry birds. The delegation from the university of bamberg had to complete three 30-minute rounds. In the final, it finally beat the strong competition from the technical university of dresden by 228,050 points to 195,350 points and once again brought the title to the cathedral city. The bambirds had already won the world championship in 2016.
The groundwork for victory was laid in the weeks leading up to the championship in bamberg, germany. The bambirds, a team of 17 students, developed an intelligent agent together with their professor diedrich wolter as part of an applied computer science project. This computer program is capable of acting and learning on its own, and can draw conclusions from the current game to adjust its strategy.
What looks like an entertaining pastime on the surface, touches on numerous still unresolved research questions of artificial intelligence. Mastering "angry birds requires, for example, the ability to grasp and exploit physical relationships. "In the future, the technologies researched here could, for example, help robots in the household to learn new tasks", according to wolter.
However, the comparison with humans also shows that there is still a long way to go before robots can solve everyday tasks in a similar way to humans: "in the separate direct comparison ‘human versus computer’, only one program managed to solve one of four levels, almost all humans were better", according to wolter.