New breakthroughs or dead-ends?

摘要

Abstract Artificial intelligence (AI), at its inception, offered new concepts for formulating psychological theories and a new methodology for testing them. It also promised an ‘existence proof’ that intelligence could be implemented in a physical system. These promises are still controversial, both in AI and in philosophy. Some researchers favour connectionism, a form of AI that has blossomed relatively recently. Others believe ‘classical’ AI insights are needed to model many types of human thinking. Some eschew classical AI (and the associated frame problem) in favour of robots ‘embedded’ in the real world. Similarly, some reject functionalist interpretations of AI, arguing that intentionality cannot be grounded in syntactic and/or simulated and/or non-evolved systems. Consciousness is highly problematic: many doubt that any computational (or even scientific) account could explain it. The papers presented at this Royal Society/British Academy meeting explore these issues. Even without dead-ends, the routes taken in AI accounts of the mind may lead in unexpected directions.