Since I know that a lot of you (like us here at the MIT AI Lab) are planning to submit funding requests for AI research, I thought that you might want to append some helpful excerpts from the following article to your proposal. >From an article in Defense Computing by Gary Martins (former RAND manager): "The AI field has been a prolific source of hokey new terminology: KNOWLEDGE-ENGINEERING for amateur programming; KNOWLEDGE-BASE for a small, slow unfinished database; EXPERT SYSTEM for a development-stage program that almost solves a toy version of a non-problem. ... The entire expert systems concept is based upon a technique known as rule-based programming --- the heart and soul of ART, KEE, KNOWLEDGE-CRAFT and other infamous AI junkware. ... The rule-based approach was invented, evaluated and then discarded by the business D.P. community in the mid-1960's. It is true that most big-ticket expert systems toolkits throw in other goodies (such as Lisp). However these other goodies can invariably be obtained in superior versions -- and at far more realistic prices -- from real software vendors (Perhaps this point is moot, by now. To judge from the bankruptcies, layoffs and restructurings in the AI industry, it has apparently been a very long time since anyone has purchased one of the big-ticket AI packages.) AI is about the same age as the rest of computing. While all other areas of technology have enjoyed heroic advanages since [1955], AI advocates continue to pick over the same stale chestnuts that seemed so fascinating way back then. ... Are the current fads of social science good models for computing? What are the Japanese doing in AI? Quietly marching away from it, holding their noses. ... Civilian AI projects in the US today are confined exclusively to the technological and managerial backwaters of American industry --- Wall Street and Detroit, of which Citicorp and GM are typically dismal examples. Can these guys make AI work? The question is left as an exercise for the reader. Here is a quick look at how some of the [DOD] loot will be squandered: ... Pilot's Associate. The associate uses AI to provide life-saving enemy-bashing advice to fighter pilots.. That is, it would do so if the Air Force would modify the F-16 to accommodate one or more refrigerator-sized Lisp machines in the cockpit. Currently ... it emits such helpful, war-winning gems as: "Warning -- canopy not closed." If next year's planned upgrades are successful, the associate may achieve functionality comparable to that of a pocket-sized checklist card. There are two key factors that contribute to the survival of AI. The first is the laziness and scientific illiteracy of most newspaper and TV journalists, investment analysts and politicians ... the flamboyant nonsense of AI is designed for PR impact. If DOD spending on AI drops far enough, universities like Stanford, MIT and CMU may even find the integrity to rid themselves of scientifically embarassing, but formerly profitable, AI programs. The quality of CS faculties and budgets at universities across the country will continue to be diluted by the presence of large numbers of AI meatballs.