Page 45 - DCAP506_ARTIFICIAL_INTELLIGENCE

Unit 3: Common AI Problems

alter planes. He is certain that the information will be obtainable in a form he can appreciate at Notes
the time he will require it.

Notes If the strategy is personified in a program that branches on an environmental state
or reads a numerical parameter from the environment, we can observe it as acquiring
knowledge, but this is perceptibly an easier case than those we have conversed.

4. A problem is more hard if it includes concurrent events and actions. To me this appears to
be the most hard unsolved epistemological problem for AI—how to articulate rules that
provide the effects of actions and events when they take place concurrently. We may
compare this with the sequential case regarded in (McCarthy and Hayes 1969). In the
sequential case we can write S’ = result (e, s) (1) where s’ is the circumstance that results
when event e appears in situation s. The effects of e can be illustrated by sentences relating
s’, e and s. One can effort a similar formalism providing a partial situation that results from
an event in another incomplete situation, but it is hard to view how to apply this to cases
in which other events may influence with the incidence.
When events are synchronized, it is generally essential to consider time as continuous. We
have events such as raining until the reservoir overflows and questions like Where was his train
when we wanted to call him?
Computer science has lately begun to formalize parallel procedures so that it is at times
possible to confirm that a system of parallel processes will fulfill its specifications. Though,
the knowledge obtainable to a robot of the other processes going on in the world will not
often take the form of a Petri net or any of the other formalisms accessed in engineering or
computer science.

Actually, anyone who desires to prove correct an airline reservation system or an air
traffic control system must access information regarding the behavior of the external
world that is less particular than a program. Nonetheless, the formalisms for expressing
details regarding parallel and indeterminate programs offer a start for axiomatizing
concurrent action.
5. A robot must be able to state knowledge regarding space, and the locations, shapes and
layouts of objects in space. Present programs considers only very special cases. Typically
locations are discrete—block A may be on block B but the formalisms do not permit
anything to be said regarding where\ on block B it is, and what shape space is left on block
B for positioning other blocks or whether block A could be shifted to project out a bit in
order to place another block. Some are more sophisticated, but the objects must have
simple geometric shapes. A formalism competent of representing the geometric
information people get from seeing and managing objects has not, to my knowledge,
been approached.
The complexity in expressing such facts is symbolized by the limitations of English in
articulating human visual knowledge. We can portray usual geometric shapes exactly in
English (fortified by mathematics), but the information we access for identifying another
person’s face cannot normally be transmitted in words. We can respond to more questions
in the occurrence of a scene than we can from memory.

6. The relation among three dimensional objects and their two dimensional retinal or camera
images is typically untreated. Dissimilar to some philosophical positions, the three
dimensional object is regarded by our minds as different from its appearances. People

LOVELY PROFESSIONAL UNIVERSITY 39