(28th-April-2020)
The frame problem is one of the important difficulties in artificial intelligence and it shows that it can not deal with all real problems that can occur in a robot with limited information processing capability.
It was first mentioned in the paper by John McCarthy and Patrick Hayes in 1969, and there are now numerous formulations.
As a situation, there is a battery to move the robot in the cave, and a time bomb is set on it. In this state the bomb explodes and the battery is destroyed and the robot can not replace the battery, so you have to take out the battery from inside the cave. The robot was ordered to "take out the battery from the cave".
The artificial intelligence robot Unit 1 R1 was well programmed, so it entered the cave and was able to safely take out the battery. However, although It was aware of the fact that the bomb was on the battery at Unit 1, It did not realize that carrying the battery carries the bomb together, so the bomb exploded after leaving the cave It was. This is because the first car knew about the purpose of taking out the battery, but it did not understand the matters that arise secondarily (that is, bringing out the bomb at the same time as taking out the battery) It is a cause.
Therefore, we developed an artificial intelligent robot Unit 2 R1 - D1 which also takes into consideration secondary issues in carrying out the purpose. However, this robot ceased to operate in the cave and came in front of the battery, the time bomb worked as it was, the robot blew away. In Unit 2, in front of the battery, "Whether the bombs on the top do not explode when this battery is moved" "Do I have to move the bomb before moving the battery?" "To move the bomb, the ceiling It started thinking about everything that could happen side by side, such as "Does not it fall down?" "Does the wall color change when approaching a bomb?", And continued thinking indefinitely. This is because there are infinitely matters that can occur side-by-side, and infinite calculation time is required to consider all of them. However, even though it is secondary matter, it does not usually need to take into consideration that "the color of the wall changes or not" or the like.
Therefore, we developed an artificial intelligent robot Unit 3 R2 - D1 which improved to not consider matters unrelated to the purpose. However, this robot ceased to work before entering the cave. Prior to entering the cave, Unit 3 continued to think indefinitely, trying to wash away all matters unrelated to purpose. This is because there are infinite matters irrelevant to the purpose, so it takes infinite calculation time to consider all of them. It is difficult to make a robot R2 - D2 that can be judged like a human being like a thing to the left.
Comments