Why don’t we treat geniuses like professional athletes?


19


ryan_b

I mean in the work-environment sense, rather than the celebrity-and-endorsement-deals sense.

A professional sports team gets a lot of benefits that are designed to keep physically talented people performing at their peak. They have support for things like recovery, sleep, nutrition, physical fitness, and of course the specific skills they use to play the game. This support takes the form of specialized personnel who are also employed by the team, whose job it is to work with the athletes for those purposes. The whole environment is geared towards performance maintenance.

The basics of sleep, diet and exercise are required for optimal performance across all domains; being in an environment that optimizes for them is an advantage.

Sports teams balance practice and games. Practice still makes sense for knowledge work; games analogize to projects/development/etc.

Decomposing practices is where the interesting bits might be. In athletics, there is almost always a ready-made set of drills for any particular skillset, which can be prioritized by a coach working with an athlete. There is already a notion of managing the workload: athletes have overtraining and thinkers have burnout.

In The Power of the Context Alan Kay describes funding people over projects and orienting them with a vision in lieu of specific goals. This sounds suspiciously like persuading geniuses to act as though they were on a team. Alan makes the comparison himself:

“Our game is more like art and sports than accounting, in that high percentages of failure are quite OK as long as enough larger processes succeed."

I am tempted to go further and say that in the context of things like science or math research failures are still a positive contribution insofar as they establish that something doesn’t work, which makes future attempts more likely. This isn’t much of a thing in sports, which are entirely built around repeatable object-level activities; striking out does not make the next batter more likely to hit.

It seems like the object vs meta level distinction also highlights where the analogy breaks down. In a game like football, each player has a specialized role which contributes to moving the ball down the field. They can spend time mastering a set of moves which they adapt on the fly and can be relied on consistently. There is not a clear place for that in service of the PARC vision of “interactive computing as a complementary intellectual partner for people pervasively networked world-wide”.

We could torture the analogy ruthlessly and say that visions are meta-goals and that a meta-athlete could practice their meta-skills of finding how to complement the team’s work in advancing the vision. That even sounds sort of plausible, until you come up against the question of how to define those meta-skills. I feel like a checklist that goes like "Can you solve a simplified version of the problem? Can you generalize a solution from a similar problem?" doesn't seem to cut it. Though such a checklist would hardly be a bad idea. Based on The Rocket Alignment Problem I envision motivational posters hanging around that go "The beacons are lit! Gondor calls for A.I.D: Articulate the confusion; Isolate the confusion; Dissolve the confusion." Also, I am reminded of a talk given by Gian-Carlo Rota called Ten Lessons I Wish I Had Been Taught, in which he makes note of the fact that Erdos and Hilbert both employed a few tricks consistently in most of their work. Perhaps the role of the head meta-coach is to choose people such that their tricks complement each other well.

So we have the problem of identifying what skills exactly should be the focus of training and practice. We also have the problem of a shortage of personnel to serve as trainers and coaches, though I wonder if this is as intractable as it seems at first blush. The lion's share of training is not so much mastery of the skill yourself as having an outside-and-informed perspective on someone else's execution. The question is, how much is enough? A second question is, could this be automated instead?