LISP 1.5 Programmer's Manual
| Hi-index | 0.00 |
"Automatic programming" is presently a popular part of research in artificial intelligence. It has been defined ambitiously as "reducing the effort required to get a task running on a computer". That definition has some interest for symbolic computing, because the success of the best-known systems has led to their wide distribution in places where there are no experts to advise casual users, who may therefore have much trouble in understanding the systems well enough to get their computations working efficiently and without excessive numbers of debugging runs. Designers generally exploit their own knowledge of favourable peculiarities of their systems to code users' problems which are slightly out of the ordinary. If the number of types of peculiarity is small, it should be possible to mechanize this part of a designer's intuition and make life correspondingly easier for the inexperienced user. The question is worth more study: evidence provided by inspection of some designers' coding of exercises and by my own automatic production of parts of programs for two problems (one in fluid mechanics and one in thermodynamics) suggests that systematic programming in this direction is feasible. I hope to report some detailed results in a later note.