[darren]
[tarrant]
it's my name
The Frame Problem
Turing was aware of Godel's theorem which states that there are certain things that humans can manifestly do that computing machines cannot do. However, his response was that it had never been proven that humans were not subject to the same limitations. The subtlety of this point is discussed in the later section on Godel's theorem.
Turing's position has been attacked in three main ways (Boden 1988). Firstly, anti-behaviourists have argued that it is not sufficient to behave like an intelligent entity to be intelligent. Secondly, it has been argued that computers could not possibly be intelligent; Searle's Chinese Room argument attacks the Turing position in this way. Thirdly, it is sometimes argued that although in principle a computer could be intelligent, in practice it is impossible to achieve because of the level of complexity that would be required.
One central feature of an intelligent being is that it can "look before it leaps"; indeed, it can think before it leaps. In order to plan an action, it must weigh up all of the relevant information; but how can it even decide which information is relevant and which is irrelevant. In other words, what is the frame of the problem?
The frame problem has been illustrated by Dennett (1984) with a story about a robot named R1 by its designers. The only task for the robot was to fend for itself. One day the designers arranged for it to learn that its spare battery was locked in a room with a time bomb waiting to go off soon. R1 located the room and formulated a plan to rescue the battery. There was a trolley in the room and the battery was on the trolley. R1 hypothesised that a certain action, which it called pullout(trolley, room) would result in the battery being removed from the room. It acted quickly and did indeed remove the battery before the bomb exploded; sadly, the bomb was also on the trolley. R1 had failed to realise that pulling out the trolley would also bring the bomb with it. R1 had missed this obvious implication of its planned act. The designers realised that the next robot must be made to recognise, not only the intended implications of the actions, but also the implications about their side-effects. They called their next robot, R1D1, the robot-deducer. When R1D1 was faced with the same predicament as before, it also decided that it should pull out the trolley to retrieve the battery. It had just finished deducing that pulling the trolley out of the room would not change the colour of the room's walls, and was embarking on a proof of the further implication that pulling the trolley out would cause its wheels to turn more revolutions than there were wheels on the trolley, when the bomb exploded. The designers realised that they must teach the next robot the difference between relevant implications and irrelevant implications. Thus, R2D1, robot-relevant-deducer was faced with the same predicament. The designers were surprised to see R2D1 sitting outside the room pondering its actions when the bomb exploded. When asked to explain itself, R2D1 responded that it was busily ignoring thousands of implications that it had deduced to be irrelevant. All of these robots suffer from the frame problem; if the designers were to build a robot with the adroitness of the fabled R2D2 then they would have to solve the frame problem
Submit a comment about this page
I am available for freelance web design work in the Cambridge, UK area, outside normal working hours. If you are interested, please