Tuesday, August 6, 2013

Consciousness raiser: Trade-offs

Some recent discussions I was involved in appeared to suffer from two circumstances: First, that many participants were talking past each other because they were not quite agreed on what the controversy was actually about, and, two, a lack of appreciation of certain aspects of reality on one side of the discussion. Aspects of reality, admittedly, that we humans are very likely to get wrong. So this post is once more mostly exploring the issue for my own benefit and perhaps for later reference if ever needed.

Today, let us consider the concept of trade-offs. When discussing the potential for future technological innovations with the kind of people one might classify as singularitarians, transhumanists, cornucopians or simply technology-optimists, one cannot help but become puzzled at their often very simplistic view of technological feasibility and progress. Many of them seem to believe that the history of engineering can be summarized as undertakings that had previously been impossible gradually becoming possible once the awesome power of the human mind is properly applied to them, and that this trend will merrily continue into the future.

That is, of course, a very one-sided perception. Yes, there have been many cases where doubters were proved spectacularly wrong, for example when it was stated that trains could not go faster than a couple tens of kilometers per hour without killing the passengers. However, Jules Verne's cannon shooting astronauts into orbit really has to remain fiction for just that reason: there is no way of accelerating a projectile (as opposed to a rocket) fast enough to reach escape velocity without killing the astronauts. There simply isn't, and that is how it will remain regardless of how much science and engineering progress.

The same is true of perpetua mobiles, for example. Some things are ruled out by the physical realities of our universe, and the same is very likely to be true for many other pipe dreams futurists are coming up with. It may well be that there is no way of traveling to another star system and surviving the journey, of running a stable fusion reaction at a scale smaller than a star, or of uploading a mind into a computer. (Then again, I am only certain about the last of these.)

But beyond plain "this violates the laws of physics" level of impossibility, one of the really under-appreciated reasons that something might turn out to be impossible are trade-offs. What does that mean? It is simply the observation that to achieve some benefit A you often have to accept some correlated downside B.

In our everyday lives, we are very conscious of our inability to have everything we would like to in one package. There is a trade-off between price and quality: the less you are willing or able to pay, the less quality you get. In car driving, there is a trade-off between speed and safety. And so on.

When dealing with engineering ideas, on the other hand, many of us happily discount the reality of trade-offs. "When we can upload a mind into a computer", "when we have nanomachines built from metal cogwheels that do work at the micrometer scale", "when we have cheap solar cells with high efficiency", these are all fairly optimistic visions of what might be physically possible. They all include the implicit assumption that two or more different desirable qualities can go together instead of having to be traded off against each other.

We are quick with making this assumption because our everyday intuitions (e.g. you can live healthy or eat unlimited amounts of whatever you want but not both) are so easily swapped for boundless techno-optimism when discussing engineering solutions. After all, we decide what we want to build and how, so it appears as if the sky is the limit. And that is where it is really helpful to be a biologist, especially with some training in physiology, ecology or evolutionary biology.

Because as a biologist, you are running into trade-offs all the time. Evolution had literally billions of years to play around with possible solutions to the survival problem. In some cases, there appears to be only one, but in others we can find very different solutions that trade one advantage against another. A few examples:
  • There are organisms that have large numbers of offspring that they invest very little in, and there are other organisms (like us) that have very little offspring but invest a lot of resources into each of them. Some are in the middle, but you can't have both extremes at the same time.
  • There are plants that are fast-growing but short-lived and very uncompetitive, and there are plants that are slow-growing but long-lived and ultimately destined to out-compete the first group. The way the first group of "pioneer species" survives is by quickly colonizing disturbed spaces and producing offspring before being overwhelmed by the second group of "climax species". Some are in the middle, but you can't have both extremes at the same time.
  • There are herbivores that are very good at running away from predators (gazelle), and there are others that are good at fighting predators off if necessary (rhinoceros). Some are in the middle, but you can't have both extremes at the same time.
We could easily consider more examples but the idea should be clear. Over time, biological "design" under the harsh supervision of natural selection tends to approximate optimal solutions. Yes, there is no guarantee that evolution has arrived at the global optimum because it had to work with what was there (which is why we sometimes get insanely bad design like the laryngeal nerve). But you can pretty much take as a given that life has been optimized to implement a near-optimal solution under the constraints of being as cheap and self-healing as possible, both qualities that are desirable but not generally in evidence in the metal and plastic based machines resulting from human engineering efforts.

And this is why many of the futurist ideas sound so naive to me as a biologist. When somebody says, "soon we'll have nanomachines fighting our diseases", I hear "soon we'll have magic pixie dust, and everybody gets a pony". The problem is that robots composed of metal cogwheels and suchlike will simply not work at the nano/micrometer scale as they do at the centimeter scale. The physical environment is completely different, with friction, charges and spontaneous molecular rearrangements playing much stronger roles.

The thing is, evolution has long ago invented nanomachines. They are called enzymes. Yes, they are squishy and short-lived, but they are also cheap to mass produce, easily assembled, and they work very well at the relevant scale. Maybe their squishiness and short lifespan are downsides that plainly have to be accepted if you want those qualities.

The same for people who believe in mind or brain uploading and suchlike. Maybe the squishiness, proneness to disease and short lifespan of animal brain cells are an unavoidable side-effect of being so damn good at running complex minds. Surely it would be hard to envision an electronic component that could heal itself, form new synapses and disconnect old ones, and all for the metabolic price of a few kilograms of organic matter. (And no, a simulation of a brain forming synapses is not a brain forming synapses. I am talking about functionally replicating the brain here. If you think that brain simulation will produce a mind I invite you to try taking a train from SimCity to work and see how far you get.)

In summary, don't underestimate the problem of trade-offs, and don't think that anything is feasible just because you can form a mental picture of it. Surely some things may become possible in the future that we cannot easily imagine at the moment. But it is just as well possible that the answers to "how do we build nanobots", "how do we achieve immortality" and "how do we get cheap solar cells" are "design novel enzymes", "have children" and "you have them already, they are called plants, and for the sake of all that is holy reduce your energy consumption". Not as sparkly a set of answers as some may have hoped but the solutions they contain have the advantage of working, demonstrably.


  1. This comment has been removed by the author.

  2. Just a pedantic linguist's comment: perpetua mobilia (if you insist on a correct Latin neuter plural) or perpetuum mobiles (if a straightforward English plural will suffice).