- Dec 30, 2012
- 4,809
Here’s a riddle:
100 prisoners stand in line, one in front of the other. Each wears either a red hat or a blue hat. Every prisoner can see the hats of the people in front – but not their own hat, or the hats worn by anyone behind. Starting at the back of the line, a prison guard asks each prisoner the colour of their hat. If they answer correctly, they will be pardoned. Before lining up, the prisoners confer on a strategy to help them. What should they do?
Stumped? Google’s artificial intelligence wasn’t. A deep neural network based on one built by Google’s DeepMindteam – software that has previously taught itself to play several dozen classic arcade games – can now solve riddles like this one. The conundrum is reportedly among those posed to prospective Google employees at interview.
The ability could one day help groups of robots work together to solve puzzles by sharing pieces of information. “It’s basically a first step toward having AIs that can communicate and collaborate,” says Jakob Foerster at the University of Oxford, who collaborated with Google’s DeepMind team on the research. “In the long run it will give them a lot more scalability and allow them to solve tasks that previously weren’t possible.”
To solve the hat riddle, each person was modelled as a separate artificially intelligent agent. These considered the colour of the hats they could see, decided what to tell the others, and then used the information collectively to work out the answer.
As when learning to play video games, the AIs largely worked out how to tackle the problem themselves. “They’ve come up with protocols that are different from how humans solve these problems,” says Foerster. “We don’t yet fully understand what the solutions are, but we know that they work.”
Source
100 prisoners stand in line, one in front of the other. Each wears either a red hat or a blue hat. Every prisoner can see the hats of the people in front – but not their own hat, or the hats worn by anyone behind. Starting at the back of the line, a prison guard asks each prisoner the colour of their hat. If they answer correctly, they will be pardoned. Before lining up, the prisoners confer on a strategy to help them. What should they do?
Stumped? Google’s artificial intelligence wasn’t. A deep neural network based on one built by Google’s DeepMindteam – software that has previously taught itself to play several dozen classic arcade games – can now solve riddles like this one. The conundrum is reportedly among those posed to prospective Google employees at interview.
The ability could one day help groups of robots work together to solve puzzles by sharing pieces of information. “It’s basically a first step toward having AIs that can communicate and collaborate,” says Jakob Foerster at the University of Oxford, who collaborated with Google’s DeepMind team on the research. “In the long run it will give them a lot more scalability and allow them to solve tasks that previously weren’t possible.”
To solve the hat riddle, each person was modelled as a separate artificially intelligent agent. These considered the colour of the hats they could see, decided what to tell the others, and then used the information collectively to work out the answer.
As when learning to play video games, the AIs largely worked out how to tackle the problem themselves. “They’ve come up with protocols that are different from how humans solve these problems,” says Foerster. “We don’t yet fully understand what the solutions are, but we know that they work.”
Source
