Friday 26 August 2011

Why are robots crap at Table Tennis? (or "How to do away with Time")

I've just spent a very relaxing week in Devon with family. It's one of the few opportunities I get for playing Table Tennis (but I'm not very good). But I'm better than this...

And that raises an interesting question. Human reactions in table tennis are very fast. Our anticipation of what is likely to happen next feels almost unconscious. It feels as if our movements naturally lead us to hit the ball when it's returned to us (if things go well). Sometimes we might be caught out by poor anticipation, or simply that our movements in one moment make it physically impossible to be in the place that our brains tell us to be in the next.

What is going on in our anticipatory system?

Coming back from the cybernetics conference, I have been wondering about anticipatory systems and our conception of time. In fact, time presents a lot of problems for cybernetic thinking generally, since it is a somewhat unexplored concept (and it is a concept!). Might it be that without time there is no difference, and without difference there is (essentially) no cybernetics as it is currently understood?

One approach to anticipatory systems is the idea that observing systems are 'nested' running at different (faster) clock speeds, where the differences picked up by a low-level observer are detected by a faster-level observer and extrapolated for likely future developments. This seems to me to be a bit like the statistical approach to evolution. And the problem there is that there are too many possible states to be explored for the anticipatory mechanisms to be at all effective. Yet we know they are effective. So what else might be going on?

One way of approaching an alternative response is to do away with the concept of time. In place of the concept of time, let's have an understanding of concrete individual human experience (i.e. a phenomenology). That might mean replacing one abstract concept with another and seeing where it takes us... so I'm thinking of replacing 'time' with 'symmetry'.

This is helping. Because there is something symmetrical in the playing of table tennis: a pattern of circular motions of the body and the ball which somehow come together. And this leads to what is a new insight for me: that what is modelled by a player is not what the other player might do (and what they might be thinking), but rather the emerging form of the relationship between them and what the individual might do to tune into that new form of relationship: it is a continual process of staying tuned-in.

Thinking symmetrically rather than in time means that we start to think about the proportions of adjustments to regulators in response to the changing environment around an organism. And understanding proportional adjustments means that the processes of anticipation become more foreseeable because of the recursive nature of the proportionality.

Once again, I'm thinking of Pauline Oliveros's listening exercises. But also I'm thinking about Faisal Kadri's paper on multiplier feedback and hysteresis. Multiplier feedback is about harmony and proportion in a way that negative feedback is about stimulus and corrective response. The negative feedback model might need time in a way that it doesn't acknowledge (although this is not to throw the baby out with the bathwater).

But it seems clear that with the science of organisation in man and machine, organisation in man is distinct from organisation in machine, and yet the two are closely related. But it is hard to escape the phenomenological roots of it all... but to do that, maybe some of the metaphors of machine do not help us as much as we thought - particularly in the field of human anticipation.

No comments: