The Intelligent Mouse Project

As far as we know, we are the only really smart species on earth. (Yes, OK - dolphins have learned to be at one with nature, yada yada. But seriously.) We had a brief period of overlap with the Neanderthals, who were probably pretty smart, but now we're all on our own.

So, I was wondering how long it would take us to make another intelligent species.

The logical starting point would be chimps or gorillas (or dolphins), but there's no fun in that. For one thing, they're all big and they breed slowly - perhaps more slowly than is essential for a clever animal - and these factors will be bothersome. Also, if we start with chimps, it's cheating because they're already almost smart, and we'll probably just wind up with something which is smart in the same way we are.

So, we should start with mice, which are small, fast- breeding, amenable and stupid. How long would it take us to breed a mouse (or mouse descendent) with intelligence comparable to our own?

Well, we can take natural evolution as a baseline. Evolution went from small, mouseish mammals to humans in (very roughly) 100 million years. Can we do better?

First, we need to know what resources natural evolution used - specifically, the population size. This is important, because a larger population will (other things being equal) throw up useful mutants faster than a smaller population. For most of our history, we humans were small mammals; I'm therefore assuming that our ancestor's population could have been as high as a billion for much of the time (becoming much smaller, of course, in the later stages of evolution and during bottlenecks). So, we can assume that natural evolution can go from mouse to man given a population of 1 billion animals (or less) for 100 million years.

So, we'll start with 1 billion mice. Assuming that a mouse cage is a 20x20x20cm cube (crudely), this means that we need about 10^13 cubic centimetres of mousery, or a cube about 200 metres on a side.

Of course, we will also need all of the mouse-tending apparatus, including a wide range of intelligence tests, which will become more challenging as we go along. Assuming that everything can be automated, with the cages in huge tiers served by robotic arms, I'm guessing that a building 200 metres high, 400m wide and 400m long ought to do it. This would be equivalent to 10 of NASAs Vehicle Assembly Building, so it's clearly not prohibitive.

The building might need to be enlarged in the later stages of evolution, but we can worry about that later.

OK, so we have a billion mice. Do we still have to wait 100 million years to get a smart one? By Jove, no!

First of all, we can increase the rate of mutation in the mice, using either radiation or mutagenising chemicals in the food. The natural mutation rate (to the extent of producing significant phenotypic effects) is probably pretty low - perhaps a percent or two (in other words, perhaps a percent of all live-born mice have a genetic as opposed to developmental anomaly). We can afford to make this at least ten-fold higher, since mice are prolific and a 10% mortality isn't a problem as long as they don't have to compete with wild mice. (Of course, almost all of these 10% of anomalies will be fatal or deleterious - it's the very rare beneficial ones that we select for).

So, with a 10-fold higher mutation rate, all other things being equal, we will get 10-fold faster evolution. (Pauses to allow outraged comments - but this is a reasonable crude approximation.) So, we're down to 10 million years to make a smart mouse.

Impatient? Of course we are! So:

We have the advantage of artificial, rather than natural selection. In the wild, a mouse which is 10% smarter than its siblings may have only a 1% advantage in terms of survival. In all likelihood, it will be eaten or will die of disease long before it faces a situation where its brains are of any help. Indeed, since a brain consumes a lot of resources, a greater intelligence may even be a handicap (after all, mice do pretty well, and haven't evolved intelligence in the wild). In other words, the selective power for intelligence in the wild is very weak.

But, in our system, we are constantly testing the mice to detect even modest increases in intelligence. A mouse which is measurably smarter can be given a 100% chance of survival, and can also be bred from intensively, with all of its progeny likewise nurtured and propagated. Therefore, I think it's reasonable to assume something like at least 100-fold better selection than nature provides (10- to 100- fold more likelihood of a useful individual being selected for; 10- to 30-fold faster propagation of that individual's genes through the population). So now we're down to 100,000 years.

Can we do any better? Well, possibly. For one thing, nature is short-sighted, and doesn't plan ahead. Perhaps an individual is born with a skull that fuses later (which is good, in terms of allowing for more brain growth in early life), but that mutation is selected against in the wild unless that individual has already evolved the bigger, longer-growing brain. With artificial selection, we can pamper and preserve such pre-adaptations, knowing that they'll be useful later. In other words, we don't need all of our mutations to happen in the right order, in the way that nature does. A little maths suggests that we might get an advantage of maybe 10-fold (or potentially much more) by taking advantage of this trick. So we're down to 10,000 years.

Another advantage is that individuals with useful mutations can be bred together. In the wild, Useful Mutation A will not breed with Useful Mutation B until both mutations have spread to a respectable fraction of the population. In our experiment, where we will cull most individuals and breed the few useful ones at each generation, there will be many more opportunities for synergistic mutations to meet eachother. Sexual reproduction is very popular in the wild, precisely because it allows this meeting of mutations - our experiment will magnify this advantage hugely. It's probably conservative to say that the selected mating (on top of the selection of individuals) will give us another 10- fold advantage, taking us down to 1,000 years. Still a lot of mouse-poop to deal with, but getting better.

Finally, there is one more time-saver. In nature, a bigger brain has to be very well supported by a host of other developments. For instance, the body has to be able to physically and metabolically support the bigger brain while still being able to flee from predators. Hands have to be good enough to make use of the bigger brain to justify this cost, and so on. But our mice are pampered, and are being selected only for intelligence (plus a minimal ability to survive and breed). So, they can get away with somewhat fewer supporting adaptations than our ancestors needed. We might squeeze a two-fold advantage out of all this at the very least, getting us down to 500 years.

Of course, as the mice get smarter and have bigger brains, it's likely that their generation time will increase. But, in starting from our baseline (100Myr) of natural evolution, this was a factor too, so in a sense it is already allowed for. However, to be on the safe side, we would perhaps want to select (on a secondary level) for short generation times. Of course a bigger brain (and somewhat bigger body) takes more time to grow, but it needn't be as long as a human generation time. In some of the junglier parts of Norfolk, generation times can be as short as 13 years for humans. If we remove the need to be physically capable of rearing offspring, I'm pretty sure a species could evolve with a human-sized brain and a generation time of two or three years. All that's needed (in absolute terms) is that the individual can grow to several times its own birthweight. (Of course, once we get a smart mouse, we might then want to breed it for longevity, so that it has time enough to learn useful things. But that's a problem for later.)

Now, to stand back for a moment and see if this makes sense. Well, does it? First, we can look at selective breeding of things like dogs. (Dogs seem to have more propensity to variation than most species, but of course our mice are being mutagenized, so the comparison is not far-fetched.) We know we can go from generic dog to pekinese or collie in something like a century of selective breeding; we can even breed for quite big changes in intelligence in dogs on this timescale. Our mice will have a faster generation time, a far, far, far bigger population from which to select, and will be selected only for intelligence, so we might expect some pretty impressive results.

We can also try to do some numbers. From what we know of genomics, it's likely that something like 10,000 to 100,000 mutations would have to happen to go from mouse to man (there are many more base-level differences between mice and men, but almost all of these are of no relevance). From a billion mice, with their increased mutation rate (10%) we get 100,000,000 mutations per generation (remember, these are all mutations with phenotypes, since that's how we decide our level of mutagenesis). Of these, perhaps one in a million takes us in the right direction, giving 100 useful mutations per generation. Given a generation time of, say, two months (faster to begin with, slower as the mice get smarter and hence have bigger heads to grow), this gives us perhaps 300,000 useful mutations arising over 500 years. Allowing for some wastage (failure to spot useful mutations, for instance), we're still in the right ballpark.

So, there you have it. For the cost of 10 Vehicle Assembly Buildings and a lot of mouse food, we could have another intelligent species to chat with in a few centuries. As a byproduct, the facility will also produce something like 10,000 metric tons of useful, crop-fertilising mouse poop per _day_, which has to be a good thing. It will also yield 1 billion mouse bodies per year (at least), which will be a boon to the pet snake industry. In fact, with a little coordination and planning, we could run a parallel program to breed intelligent snakes at almost no additional cost.