Case-based reasoning (CBR) is the enterprise of solving new problems by analogy with old ones. Problems and their solutions are stored as cases, and to solve a new problem, a CBR system retrieves a case of a similar problem, and adapts its solution to solve the new problem.
Given the shortcomings of AI systems that reason only from first principles, I believe that CBR will be key to making computers more intelligent than man next century. For complex problems, CBR promises greater efficiency than solving problems from scratch. Cases can suggest solutions even to ill-defined problems in weak-theory domains for which there may be no adequate set of rules or hard-wired algorithms. CBR systems can learn from experience (by acquiring new cases), and so can improve their own efficiency and competence over time. Analogy is key to creativity, since new knowledge derives not from nowhere, but from a combination of old knowledge.
Although many of today's CBR systems use retrieval without automated adaptation, I firmly believe that the power of true CBR comes from the combination of retrieval and adaptation. Only by using both together will computers surpass all human capabilities. Adaptation will be the most major challenge for CBR over the coming decades.