p.1 Artificial Intelligence (AI) may be defined as the branch of computer science that is concerned
with the automation of intelligent behavior.
p.21 Much of what we commonly call intelligence seems to reside in the heuristics used by humans
to solve problems.
p.27 While humans plan effortlessly, creating a computer program that can do the same is
a difficult challenge.
p.123-124 heuristics are formalized rules for choosing those branches in a state space that are
most likely to lead to an acceptable problem solution. AI problem solvers employ heuristics in two basic situations:
1. A problem may not have an exact solution because of inherent ambiguities in the problem statement or available data...
2. A problem may have an exact solution, but the computational cost of finding it may be prohibitive... Heuristics
attack this complexity by guiding the search along the most "promising" path through the space... Unfortunately,
like all rules of discovery and invention, heuristics are fallible. A heuristic is only an informed
guess of the next step to be taken in solving a problem. It is often based on experience or intuition. Because
heuristics use limited information, such as knowledge of the present situation or descriptions of states currently
on the open list, they are seldom able to predict the exact behavior of the state space farther along in the search.
A heuristic can lead a search algorithm to a suboptimal solution or fail to find any solution at all. This is an inherent
limitation of heuristic search... Heuristics and the design of algorithms to implement heuristic search have long
been a core concern of artificial intelligence. Game playing and theorem proving are two of the oldest applications
in artificial intelligence; both of these require heuristics to prune spaces of possible solutions. It is not feasible
to examine every inference that can be made in a mathematics domain or every possible move that can be made
on a chessboard, Heuristic search is often the only practical answer.
p.152 In applying minimax to more complicated games, it is seldom possible to expand the state space graph
out to the leaf nodes [checkmate, for example]. Instead, the state space is searched to a predefined number of levels,
as determined by available resources of time and memory. This strategy is called an n-ply look-ahead, where n is
the number of levels explored. As the leaves of this subgraph are not final states of the game, it is not possible
to give them values that reflect a win or a loss. Instead, each node is given a value according to some heuristic evaluation
function. The value that is propagated back to the root node is not an indication of whether or
not a win can be achieved (as in the previous example) but is simply the heuristic value of the best state that can
be reached in n moves from the root [position]. Look-ahead increases the power of a heuristic by allowing it to be
applied over a greater area of the state space.
p.159 In the attempt to bring down the branching of a search or otherwise constrain the search space, we
presented the notion of more informed heuristics. The more informed the search, the less space must
be searched to get the minimal path solution. As we pointed out in Section 4.4, the computational costs of
the additional information needed to further cut down the search space may not always be acceptable... As the information
included in the heuristic increases, the cpu cost of the heuristic increases. Similarly, as the heuristic gets more
informed, the cpu cost of evaluating states gets smaller, because fewer states are considered. The
critical cost, however, is the total cost of computing the heuristic PLUS evaluating states, and it is usually desirable that
this cost be minimized.
p.277 The first principle of knowledge engineering is that the problem-solving power exhibited by
an intelligent agent's performance is primarily the consequence of its knowledge base, and only secondarily a consequence
of the inference method employed. Expert systems must be knowledge-rich even if they are methods-poor. This
is an important result and one that has only recently become well understood in AI. For a long time AI has focused its attention
almost exclusively on the development of clever inference methods; almost any inference method will do. The power
resides in the knowledge. Edward Feigenbaum, Stanford University
p.298 Human expertise is an extremely complex amalgamation of theoretical knowledge, experience-based
problem-solving heuristics, examples of past problems and their solutions, perceptual and interpretive skills and
other abilities that are so poorly understood that we can only describe them as intuitive.
