Artificial intelligence always needs to be based on curiosity. The world of AI in the last couple of years has been trying to focus on the fact that the human quality needs to be improved before making the computers smart. Google’s DeepMind division has published a new research they have done on this same subject.
Montezuma’s Revenge on Atari is regarded as one of the most difficult games ever and the video above shows that a DeepMind agent/bot is trying to tackle that. The agent is only learning how to play game in the video like every other human being as it was not previously aware of anything about it. Nonetheless, the agent was able to collect points, grab keys and play the game like a pro because of the Artificial Intelligence of DeepMind. Interestingly, a similar bot, which wasn’t backed with AI tried the same game back in February and failed to score any point.
What do you think the difference between the two incidents? Definitely curiosity.
When we are discussing about ‘Curiosity’ in Artificial Intelligence, it refers to the term ‘Intrinsic Motivation”. Intrinsic motivation in general refers to the motivation which is found in the activity itself. In Artificial Intelligence, ‘Intrinsic Motivation’ refers to the creating some kind of a reward system for the agent, which makes it explore the surrounding curiously.
The report reads, “After 50 million frames, the agent using exploration bonuses has seen a total of 15 rooms, while the no-bonus agent has seen two.”
The new agent clearly explores more trap-filled rooms of the very tough Montezuma’s Revenge than its previous iterations only out of curiosity or the ‘Intrinsic Motivation’.
However it is to be noted that the same type of Artificial Intelligence cannot be compared to that of the real world. First and foremost there’s no unlimited life in real life and there are little chances of encountering a ghostly villains and falling of the ledge like videogames.