Perhaps this is an appropriate point to bring my speculations to a close and to summarize briefly the course of my argument. Up to the present time, operations research and the management sciences have been largely limited, by the nature of this tools, to dealing with well-structured problems that possess algorithmic means of solution. With recent developments in our understanding of heuristic processes and their simulation by digital computers, the way is open to deal scientifically with ill-structured problems – to make the computer coextensive with the human mind.
That was Herbert A. Simon and Allen Newell’s introduction to the The Journal of the Operations Research Society of America. Volume 6, 1958.
They created the very first artificial intelligence (AI) program just a couple years before writing that statement.
Here is a great reaction from Richard Bellman, Rand Corporation, Santa Monica, California (received March 20, 1958) – “On “Heuristic Problem Solving,” by Simon and Newell.
On page 7 of the article by Simon and Newell, four predictions are made concerning the usage of computers within ten years, predictions concerning the discovery of important mathematical theorems, the writing of worthwhile music, the future dependence of the major part of the field of psychology upon computers, and the dethroning of the current world chess champion by a computer…
For those who are interested in becoming prophets with honor in their own time and in their own country, there is a fundamental principle which we may call the Principle of Optimism: Never make negative predictions.
I believe that 1 of the 4 things Simon and Newell predicted about computers came true at all, and that was in the 1990s when IBM deep blue beat Kasparov in chess. I’ve heard stories that computers are writing music, and can put together some decent looking literature. But no one reads or listens to it, it would seem.
The point is that Herbert Simon, one of the great thinkers, thought that operations research and management could be handled by AI. This means that the AI would be general AI as opposed to the brute force used to beat Kasparov. It would be able to formulate new problems, decide on the relevant parameters, considers which tests are applicable, and so forth.
That is the thing about making positive predictions. It generates enthusiasm and unbounded forecasts of performance. In fact, much of that optimism bias can be good thing, motivating people to tackle big problems. (For example, see Hirschman’s Hiding Hand principle). But it could be dangerous to lock in future plans based on that optimism, or reorganize social institutions around scientific schemes.
Today, it doesn’t seem that anybody knows how far off general AI is. It could be tomorrow, or a thousand years. Neural networks are interesting because they are designed much like the natural processes of the brain. But still, for military and management problems — which are real world problems — it would need constant interaction with the environment. We humans are physical and generate concepts of objects as they exist in the real world. That helps us understand cause and effect in that world. It might then appear that general AI requires some co-evolution with the environment itself, which implies time.
Anyway, who knows where our tacit knowledge comes from with which we conjecture. I’m not trying to say that general AI is impossible, and won’t some day be able to solve ill-defined problems in military and acquisition matters. I’m just saying, the days when operations research and management are performed by general AI is likely still so far away that no researcher or manager reading this in the first days of 2019 will likely see their career superseded by it. We first need to take hold of machine learning, which will transform our workplace and cities over the coming decades.
Leave a Reply