Artificial Intelligence Applications to Fire Management

Artificial intelligence could be used in Forest Service fire management and land-use planning to a larger degree than is now done. Robots, for example, could be programmed to monitor for fire and insect activity, to keep track of wildlife, and to do elementary thinking about the environment. Catching up with the fast-changing technology is imperative. Just what is artificial intelligence (AI), and what could it possibly do for (or to) fire science and management? Before getting into uses that, as we shall see, will be manifold and pervasive, let us examine some of the characteristics and definitions of artificial intelligence. Many potential or actual users of AI will insist that the term is an oxymoron, that is, anything artificial cannot be intelligent, and vice versa. There is a simple operational definition that gets us around this difficulty: the Turing test, named after its inventor, Allan Turing. If you are communicating with someone or something, and you can't tell whether the someone or something is human or machine, then it is intelligent. This test is simple, yet Presented at the Symposium on Wildland Fire 2000, April 27-30, 1987, South Lake Tahoe, California. Research Meteorologist, Intermountain Forest and Range Experiment Station, Forest Service, U.S. Department of Agriculture, Missoula, Montana. effective. We can expand this test just a little to include more behavior than communication, and we have a pretty good working definition of artificial intelligence along these lines: if it walks like a duck, quacks like a duck, and looks like a duck, then it's probably a duck. I don't think there is any question that man can replicate his thinking processes in machinery and use machinery to expand those processes. We've already done this. We will build machinery that thinks to match the job to be done. Imagine coming to work one morning and turning on your desktop thinking machine, only to have it announce that it was calling in sick. It would probably really be fishing. So, we'd probably want our desktop machines to be rather subservient. On the other hand, suppose we send a machine to explore the surface of Jupiter. We would want this machine to have a great deal of intelligence, including the capability of refusing to carry out a self-destructive action (see Asimov's laws of robotics). Where are we now in the realm of AI? Currently, the field is broken into several overlapping categories. The most common divisions are: environmental sensing (vision, hearing), including pattern matching and recognition of things; computer learning and analysis; "natural" language and linguistics; and reasoning, or knowledge-based systems ("expert systems"). Does this sound familiar? It should, because it is just what people do: input data from through their senses (or from internal sources such as a dream), filter and think about the data, apply known processes and remembered data to the result, and act on the result (fig. 1, 2). Some of our actions are