|Peter Boettke|
Listening to NPR one morning, this story about soccer playing robots came on and I was intrigued by the discussion.
When I entered graduate school and was assigned to Don Lavoie, my expectation would be that I would be working on questions related to the problems of economic calculation under socialism. Of course, I wasn't disappointed as in the first 2 years of graduate study Lavoie was finishing up Rivalry and Central Planning, and National Economic Planning: What is Left? But most of my work for Don at that stage of those projects was reference checking, his argument was already completed and written down. So besides learning by reading his books, I learned how to work the open stacks at the Library of Congress and browse the University of Maryland library shelves to double check and track down accurate references that Lavoie was citing. But in the process, Lavoie was also exploring the full implications of Hayek's argument about "knowledge of time and place" and the Hayek-Polanyi argument about "tacit knowledge". This led him, and those close to him, into an examination of the nature of learning, the differences between technical knowledge and practical knowledge, and ultimately into an examination of both cognitive sciences and the philosophy of mind.
So in addition to learning Gould and Ferguson (Micro), Henderson and Quandt (Micro), Varian (Micro), Branson (Macro), Barro (Macro), Chiang (math econ), Silberberg (math econ), Theil (econometrics) and Kmenta (econometrics) during my first year, I was spending weekends reading not only Michael Polanyi but Herbert Dryfus, John Searle, etc. One of the reasons this was so relevant to my core interests in the problems of economic calculation is that many economists at the time thought that super computers would one day be able to solve the problem. (e.g., Lange 1967).
The relevant point for economics turns on the differences between information and knowledge in understanding the economic forces at work under alternative institutional arrangements. It is not the costliness of information that is the critical issue, and thus the incentives that actors face in acquiring and utilizing the bits and pieces of information that are scattered throughout the economy, it is instead the contextual nature of the knowledge that exists only within certain contexts and outside of that contexts does not exist at all -- and therefore cannot be discovered and utilized by actors in the economic system not matter how hard they search.
The enthusiastic hard AI would-be economic planner is ultimately conflating syntax with semantics in 'the grammar of the economic calculus'*. Formalism can provide clarity with respect to syntactic knowledge but remains silient with respect to semantic knowledge. In the affairs of men, however, it is semantics that we must uncover for understanding how the world works. As Searle argued long ago now, machine langue is not the same as human language no matter how much we try to mimic it.
Well now go back to think about our soccer playing robots. The effort is to mimic human playing soccer players, with the ultimate goal of competing against elite soccer players. But in this effort, I would argue that the analogy to chess is misplaced. It is not just the physical dexterity required to play a sport such as soccer at an elite level, but the decision making capacity of the players with so many degrees of freedom. Chess is no doubt an extremely taxing game of mental capacity (perhaps the hardest such game), but it is also played within an extremely structured environment with only so many permutations and combinations. In a sport such as soccess, or basketball, or tennis, the permutations and combinations are in some practical sense far more open-ended, and thus leaving greater scope for, and necessity of, "entrepreneurial alertness" to hitherto unrecognized opportunities for winning plays, counter-strategies, and new techniques and tactics.
It is not just physical dexterity that presents a challenge to what computers can do, but mental nuance. Nuance and semantic knowledge go hand in hand, without access to semantic knowledge nuance is not possible. So the quest for soccer playing robots will continue to be frustrating if judged against the standard of elite human players. This does not mean that (a) robots will not play soccer (or any other sport) against each other in the future, they just will not successfully mimic elite human play, and (b) some sort of singularity momement where man and machine merge will produce a hybrid human player that will be stronger, faster, and more capable than the current human elite player. All I am saying is that the computer will not substitute for the cognitively nuanced aspects of the human player -- computers will not take over those parts of the playing of the game that require nuance; that require the skillful utilization of the semantic knowledge of context; that are grounded in the unique knowledge of time and place.
Bob Subrick and I explore the implications of this sort of philosophy of mind argument in our paper "From the Philosophy of Mind to the Philosophy of the Market."
*This is the proposed title of an unpublished book of Hayek's that he was working on in the decades after WWII.