12 January 2007

The indefatigable Dr C, a hymn to Mrs T, & the sins of Mr K.

In two posts now, The puzzles of Keros and A small agenda, Dr C has drawn attention to the threats posed by pattern recognition technologies (as visited in my Recognising the future article) to the future employment prospects for radiographers.

He's right, of course, though it hadn't occurred to me until he pointed it out. Radiographers are one of many groups whose skills will inevitably be replaced by AI constructs linking pattern recognition to combinations of heuristic and expert knowledge systems. It's a whole new phase of a replacement process which began with Hargreaves' "spinning jenny" machine in the 18th century – then it was mechanical skills which were sidelined, now it's workers with the mind who will feel the pinch. I have no doubt, personally, that doctors (like Dr C) and research mathematicians (like myself) will also topple in time.

I was still vaguely turning this over in my mind when I visited my dentist (hereafter "Mrs T") for the twice yearly checkup.

Now, I could give you a long list of reasons why I am lucky in having Mrs T my dentist but the three big ones are:

  1. She's an exceptionally good dentist (obviously important).
  2. I trust her implicitly. (This matters a great deal if, like me, you have a lifelong pathological fear of dentists. I have coped better with the terrors of nightmare than I cope with walking through the door of a dental surgery. I have endured a many more serious pains with more fortitude than I display during a simple filling. I am a coward in all things, but never more so than when required to lie back and let someone put a mirror in my mouth.)
  3. She is one of the most interesting people I have ever met. (Not that I meet her very often. Fifteen minutes twice a year, on average, and my contributions to conversation are restricted to terrified grunts past whatever is going on inside my mouth.)

It's that last item that is relevant here. Mrs T knows my fear, and uses my curiosity to distract me from it. That, of course, is one of the things which make her a good dentist. As a bonus, I very rarely leave her chair without having learnt new information or without having had planted the seeds of several new ideas or trains of thought.

On this occasion, the subject was the complexity of the tongue. Have you ever thought deeply about your tongue? I haven't. I know all the basic facts from A-level biology, first aid training, and field medicine (four main muscles providing mobility, gustatory calyculi enabling taste, the lingual artery supplying the whole structure with blood). I've even realised that it's pretty damn clever, and looked at artificial digital replications of its taste function. But I've never really thought about it. I've given a lot of time and thought and study to the eye, and to a lesser extent the ear, but I've pretty much taken the tongue for granted. By the time my inspection, scale and polish were over, though, I was some way into thinking about it very deeply indeed.

Had I ever thought, asked Mrs T, about the information processing capacity required just to simultaneously handle both the continual exploratory movement of the tongue and the associated tactile feedback, not to mention monitoring the informational state of those ten thousand taste buds? Well, no – I hadn't (though I couldn't say so at the time, with my own tongue pressed down to make room for a mouthful of skilled fingers and chrome implements). Now, though, I did think about it. And, she pointed out, all of this is just one small subroutine within the total body monitoring run by the central nervous system - which is, in itself, only one function of that system amongst many others.

Somewhere in there she made a comment which could have been deliberately aimed at both my professional futurologising and my conversation with Dr C, though I'm sure that wasn't intended. "It will be a very long time" she said "before the human system is completely computerised."

Despite my predictions of a artificially intelligent machine future, I agree with her. In fact, I doubt that reproduction of the whole human being will ever happen. Not because it will never be possible (though the possibility, even if it exists, is a long way off), but because it will never make much sense.

Human beings, as I said in an earlier post, are a package evolved by evolution (or, if you are from the other shop, designed) to meet a very different world from the one within which robotics are developing. No robot which is designed (or evolves) now will make the same journey from arboreal to urban via plains and ice ages that we have made. So why would we (or it) want to reproduce for a machine entity the exact package which human beings developed in the course of (or, if you prefer, for) that journey?

[I shan't take up any more space with those parenthesised references to creationist or intelligent design viewpoints from now on. I'm an evolutionist, and I'll simplify things by talking like one, but even if you are not it makes no difference to what I am saying here: that everything relates to the environment in which human beings had to survive before they pulled themselves clear of it through use of tools and intellect.]

There are already digital chemoreceptor devices which mimic the taste and smell functions of tongue and nose. For particular purposes (wine tasting as quality control, for instance, or explosives detection) they may even do a better job than their natural prototypes. There is not, however, a digital device with ten thousand receptors on an incredibly flexible, mobile powerful and conformable motor platform which also carries dense tactile sensitivity and moreover can be used as a manipulator, not to mention containing its own reflex arcs.

And, offhand, I find it hard to think of a reason for building such a device. A flexible platform for "tasting" shaped surfaces, yes. A complex manipulator with built in tactility, yes. But both in one? That it has evolved in humans and other animals is an accident of history and environment which do not apply to new robotic entities.

Even if an exact artificial analogue of the tongue were found to be useful, placing it inside the mouth of a perfect human copy would not. There are, for example, more efficient ways to power a robot than the complex chemical alimentary system which we use – of which the placement and arrangement of the tongue is a part.

Going back to my original point, from which the exchange with Dr C developed, a robot built any time from the 1990s onwards will exist not only within a different physical environment but within a different informational one as well. On earth or within a few light seconds (not just the moon but as far as, say, the Lagrange points L1 an L2), such a robot will inhabit a world where the machine equivalent of "race consciousness" preexists it. The internet makes both data stores and algorithms available to anyone or anything connected to it; to duplicate them onboard is unnecessary, unless security or the transmission delay is unacceptable for a particular purpose. Human (or other animal) bodies and minds are shaped by a universe where each individual is essentially alone; why would a robot (with global connectivity only a radio pulse away), or its designers, want to duplicate the anatomy or psychology shaped by such isolation?

It has often been debated how far a social insect such as an ant or bee is an entity distinct from its anthill or hive. A similar blurring of boundaries would be inevitable between a machine intelligence and the machine society of which it was part. Although I have used books and libraries as analogies for the internet, that holds true only for human beings; a robot's relationship to the internet cyberspace will be far more complex and intimate.

It's not impossible that dentistry, like radiography, will one day be carried out by machines; but not by a humanoid robot. An expert knowledge system would hold the requisite technical expertise, but it could be located anywhere within reason. Probably the reflexes needed to for rapid response to shrieks of pain will need to be within the same room ... but that could mean behind the wall of the surgery, and need only be a copy of a centrally held (and centrally updated) template. The eyes of the robot dentist would be better placed on the tools themselves, rather than on a mechanical head peering into a mirror from outside the mouth. Another expert knowledge system might one day even mimic Mrs T's ability to respond to patient anxiety with psychological diversion strategies such as discussing the complexity of the human tongue – but that, too, could be centralised and available to every surgery.

I don't, let me emphasise, enjoy the thought of such a robot surgery. I'm old fashioned enough, and sufficiently set in my ways, to infinitely prefer having Mrs T there in person. I'm just looking ahead to technological possibilities and their design implications.

But then again ... thinking about how lucky I am in Mrs T reminds me of Mr K. When I was eleven years old, Mr K pulled a molar when I was sufficiently deep in general anaesthesia to prevent motor response but not deep enough to be unaware of everything that was happening. It was also Mr K who told me that something wouldn't hurt; then it did hurt, and when I said so he just laughed and said "nonsense – I didn't feel a thing." The robot surgery seems a suddenly attractive alternative to Mr K.

All robots within fast communication reach of the internet (or whatever its successor may be) will, then, locate many of their functions within it. As their distance from it increases, they will need to be progressively autonomous – but nowhere near as self contained, physically or mentally or psychologically, as a human being or other higher animal.

A robot sent out across interplanetary or even interstellar space would need to be completely self sufficient, of course; the communication delays would rapidly become too great. But even then, its design would be dictated by a very different environment from that which shaped ours. (Perhaps it would not even be a single physical unit but a "swarm" of component entities; this was suggested to me by an astronomer some years ago – see note 6 below)

All of this ran through my mind as I walked away from Mrs T's surgery on Tuesday. It's now Friday – I have taken those three days to find the time for writing it up. In the meantime, the indefatigable Dr C has made another sally (Some thoughts on information) - once again it's good stuff and well worth reading, so trot on over there and take it in.

Post script(s)

Between September 1997 and March 2000, Ray Girvan and I co‑wrote a speculative scientific computing column, Difference of Opinion, which several times touched on issues relevant to this post. If you have the patience after reading this far, you may be interested in reading some of them:

  1. Autonomous robots
  2. Robots as dispersed "swarms", including interstellar travel
  3. Software organisms
  4. Software mutation
  5. Robot implementation of human expertise
  6. Life on the internet
  7. Automated surgery

A complete Difference of Opinion archive can be found here or here.

1 comment:

girl in a gale said...

The wonder of the Internet is how fresh this Growlery post is even after 3 and a half years. Mrs T for Mrs Teeth? :-)