Abstract: | In recent years artificial intelligence (AI) has become more mathematical. The 'empirical programming' methodology is no longer the norm. AI researchers have been extracting 'neat' techniques from their 'scruffy' programs and formalising them. Mathematics is often found to provide an appropriate language for this formalisation. New uses are being found for mathematics in every area of AI. Where existing mathematics is not up to the task, new kinds of mathematics are being invented.
These developments are generally to be welcomed as a sign of increasing maturity in the field and as a way of improving the reliability of AI products. However, it creates some problems which we must try to solve by (1) improved human/computer interfaces to make the new mathematical techniques more accessible, (2) improved teaching to produce AI researchers who are comfortable with these new techniques and (3) more testing of new techniques on 'real world' problems
|