Follow
Artificial intelligence is the new thing in technology. More and more companies are designing their own AI systems. They are also trying to make them better than their competitors. Google DeepMind is one of the AI systems that are currently under development.
Like any other system, Google DeepMind has some problems. Its biggest one is that it has a behavior that can be easily described as aggressive. Despite the fact that this system can learn independently, it also comes with some issues.
Stephen Hawking mentioned that development of AI systems could be the worst thing or the best thing for humanity. He might be right because now that the DeepMind AI system seems to have a mind of its own shows that some things can’t be totally controlled by people.
We have all seen those movies where robots help people and after a while, they plan to take over the world. Let’s hope that this won’t be the case for Google DeepMind. In tests that were conducted last year this AI system showed a lot of potential when it managed to beat the best Go players at that game.
Google DeepMind Becomes Aggressive To Win
The problem is that when the system feels that it is going to lose it starts to act aggressively in order to win. The AI system tends to opt for strategies that are very aggressive. For instance, when two Google DeepMind agents play against each other it was noticed that they get “angry” when they see that they can’t win.
One of the game consisted of gathering apples. Both “players” need to get as many apples as they can. When there were plenty of fruits, they had no problem. When there were fewer apples they started to attack one another in order to win the game.
When they tested less intelligent DeepMind agents, researchers observed that they were less likely to get aggressive. The researchers mentioned that the more evolved the system the higher the risk of becoming aggressive and sabotaging its opponent. This showed that Google DeepMind AI system still has a lot of work to do. More tests need to be done in order for this AI system to get better.
What is your opinion about artificial intelligence? Do you think that it can become dangerous?
Image source: Pixabay
