00100	PHILOSOPHICAL PRINCIPLES OF ARTIFICIAL INTELLIGENCE
00200	
00300	
00400	1. One of the main difficulties philosophers get into in trying
00500	to make concepts precise is the hard case, often one that no-one
00600	has previously thought of.  The distinctions the case suggests
00700	seem pettifogging, and they are often unsuccessful, because someone
00800	comes up with an even harder case.  The solution is to avoid hard
00900	cases by trying to do only the easy cases.  This is best done
01000	by trying to say how an artificial intelligence could be provided
01100	with knowledge about (say) causality sufficient to behave
01200	intelligently in some class of cases.
01300	We then avoid trying to make a definition that fits all possible
01400	cases.
01500	
01600	2. This is an example of the "third person principle".  Instead of
01700	asking, "how do I know" or "how do we know" we ask "how can we
01800	make it know" and in the latter task we do not doubt any of our
01900	own knowledge.
02000	
02100	3. The "principle of philosophical relativity" suggests that I
02200	am nothing special.  I will account for my own experience only
02300	in ways that apply to others too.  In fact, humans with their
02400	particular collections of senses are nothing special.  What
02500	happens to be directly perceivable by humans is not especially
02600	likely to be philosophically or physically fundamental.  The
02700	physicists understand this except when they try to be
02800	philosophers.
02900	
03000	4. Counterfactual sentences are relative to theories.  This is
03100	best exemplified by the automaton model.
03200	
03300	5. Explanations are offered for
03400		a. I can but I won't.
03500		b. Free will
03600	in deterministic systems.
03700		c. causality.
03800		d. counterfactual sentences.
03900		e. metaphysically and epistemologically adequate theories.
04000		f. seeing the dog, imagining the dog, hallucinating the dog,
04100	seeing a dog sense-datum.
04200