Yonke, M.D.: A LISP machineshould only be thought of as an expensiveterminal, Proc. of the lhh Annual MicroprogrammingWorkshop,Nov. 1978. [5.1. 5.3.1] such as Mars, and then "rove" over the surface must have rudimentary "intelligence."5 These sample gathering and analyzing devices will be many millions of kilometers from earth. Radio transmissions will take a number of minutes to reach their destination (even from Mars). If the rover suddenly need to "know" what to do, such as "do I go down that 60 ° grade?," the device will not have time to communicate with earth. It will have to decide on the spot. Thus, the probe will need to be able to "intelligently" select from certain programmed "behaviors" and groups of "behaviors" (perhaps programs like CIA 2'3 will be used). As space probes become even more complex and go to still remoter planets, this will be an area to apply AI software. These are a few ideas. No doubt, the reader will think of others, and their ideas and comments are most welcome. Please send them as well as any other inquiries to the above address. PRACTICAL USES FOR AI RESEARCH ALAN E. FILSINGER 804 E. Monroe Circle Pittsburgh, Pa., 15229 AI research is sometimes believed to be an area of limited practicality. A discussion of a few applications of A1 to practical uses may help some to have a less limited view of the field. One of the areas which may be of practical value is in the use of language processing algorithms. An application of this is ticket scheduling algorithm. In the near future, a person who wishes to buy a ticket on an airline (or other vehicle of transportation) may telephone a computer program. This software will ask the caller certain basic questions which require limited types of answers. The program will then interpret the words in the answers of the caller and schedule that person for a flight. The "intelligence" is in the interpreting of the callers words and to a lesser extent the selection of questions. Already the precursors to this can be noted. Certain elementary functions formerly performed by telephone operators are being computerized. While quite limited, on certain calls a computer program poses questions (using a persons voice). The program may ask for money and give out elementary instructions and information. The future growth from this will no doubt involve more "intelligent" programs capable of doing more complex tasks. There is another possible application of A1 research. This is in the area of software design and. engineering. Rich, et al. I, have proposed software to translate programs from one language to another, do debugging, compilation, and even documentation. Another area in compute software is adapting AI programs to work with operating systems. Problem solving software could be used as a nominal operating system adapter to increase through-put and overall system efficiency. As system needs vary, the AI adaptor could "learn" which variables to manipulate. Software by the author titled "Conditioning Implementation Algorithm,''2,3 might be adapted for such a use. CIA simulates the psychological mechanism of classical and instrumental conditioning to "learn" certain basic "behaviors" and groups and series of "behaviors." When the decision on operating system (or associated algorithms4) makes are considered, it is possible to conceive of (certain of the more complex) monitors as, in a sense, being "intelligent." This is more likely to be considered the case where a large and variable demand is placed on the operating system such as at a university. This leads to the following rule of thumb. Generally, when many complex decisions need to be made, such as when human decision making is substituted by computer decision making, then this is a potential practical use for A1. Finally, mention of NASA space probes will be made. Proposed devices which may someday be landed on the planets,

/lp/association-for-computing-machinery/practical-uses-for-ai-research-ECoOuJZsMy