As our interactions with machines increase, it’s important that a constructive dialogue can be maintained. A recent project led by the University of Michigan explored whether the ’20 questions’ game could help to ensure that happens.
If you’re not familiar with 20 questions, it involves asking a series of binary questions with a yes or no answer that can hopefully help the player to estimate an unknown value. The researchers believe that this approach could be used by AI to allow machines to ask questions, both of other machines and of humans. The team believe it could help to underpin new methods of interaction between man and machine.
Man and machine
“It is well known that artificial intelligence systems, such as those found nowadays on every smartphone, can answer at least some questions,” they explain. “They can even win a game like Jeopardy, focusing on only one question at the time. A real, purposeful conversation, especially in complicated military environments, is different. It requires the AI system to understand a whole sequence of questions and answers, and to handle every question or answer with consideration of what has been asked or answered before. Such computer algorithms do not yet exist, and the scientific theory for building such algorithms is not yet developed.”
The authors believe it’s currently extremely difficult for machines to efficiently communicate with humans and tap into our expertise. One thing humans are very good at however is answering yes/no questions. The challenge however was to formulate questions that would both minimize the number of queries whilst also maximizing the value of each one.
The actual 20 questions game is usually pretty straightforward, but the researchers appreciated that in their scenario, it was possible that a question may answered incorrectly. It’s something they refer to as the noisy 20 questions game.
The chances of producing such an error were reduced by formulating a sequence of questions that would correctly answer the question “what is the value of x?”. The team next hope to apply the method to improve communication between soldiers and robots.
It’s an interesting project, and you can learn more about it via the video below.