FT272


When Japanese researchers claimed to have discovered intelli­gence in slime mould, it sounded like a science-fiction nightmare from the fringes of science. But their study was eminently respectable, and further work to understand the anomaly is turning our understanding of intelligence upside down.

Slime mould is a long way down the evolutionary ladder. It’s a fungus-like organism that spends part of its life-cycle as a large number of free-roaming single-celled organisms like amœba. Under some conditions, the individual cells join together into a unified organism that roams the damp forest floor seeking food, in the form of a slime which has been described as looking like dog vomit. A slime mould has no nervous system or any physiology for intelligence.

Toshiyuki Nakagaki of the Institute of Physical and Chemical Research in Nagoya, Japan, decided to see whether the slime moved at random or whether it showed some sort of decision-making capability, by putting it in a maze. Mazes have a long history in psychological research as intelligence tests, as they give a convenient objective measure of ability. They were most famously employed with rats (part of the basis for the term ‘rat race’), but everything from worms to insects to human have been exposed to maze testing at some point. Mazes are still used to assess humans. The Porteus Maze Test is a nonverbal intelligence test, mainly for children but which can also be administered to adults. The mazes are on paper, and the subject traces a path between the entrance and exit, trying to avoid blind alleys. There is no time limit.

Nakagaki carried out the test by putting sections of a slime mould in a maze with food at the entrance and exit. The slime mould joined itself together, which was predictable. What was surprising was that it ended up in a single path, invariably along the shortest route between entrance and exit. Nakagaki concluded that the slime mould’s behaviour demonstrated a “primitive intelligence” which apparently evolved to make foraging as efficient as possible.

Bartosz Grzbowski at Northwestern University in Illinois decided to take the experiment a step further. Grzbowski is a chemist, and rather than using a living thing, his trial involved a drop of oil. The maze was filled with water so the oil drop could float around freely. The exit of the maze was made acidic, and this set up a slight acidity gradient within the maze. The side of the droplet closing the exit would be slightly more acidic than the oppos­ite side. This affected its surface tension, causing it to gravitate towards the exit. While it might be tempting to credit the slime mould with some sort of primitive intelligence, surely this would not apply to an inanim­ate drop of oil?

Both the slime mould and the oil might be described as analogue computers, which solve problems by using physical forces rather than brainpower. Slide rules, mechanical calculators and astrolabes for predicting the movements of the stars and planets are also analogue computers. A mechanical calculator doesn’t work in the same way as a human brain, but it still gets to the same answer (generally much more rapidly).

The fact that you can arrive at an ‘intelligent’ answer without apparent intelligence led to a split in the world of artificial intelligence research. The aim of AI is to recreate actual intelligence. If you have a system that behaves in an intelli­gent fashion and gives all the right answers, then one school – ‘Hard AI’ – says that it really is intelligent. The other school, ‘Soft AI’, says that unless it’s working something like a brain it’s just a simulation of intelligence and not ‘really’ intelligent.

This distinction holds well enough if you take humans as your paradigm of what it means to be intelligent; but nature clearly has other ideas. For example, a colony of ants can find the optimum route to food sources, changing paths when the environment changes.  This seemingly intelligent process is based on the ants laying scent trails. The more efficient a route is, the more ants will use it and the stronger the scent becomes, until only the best route is being followed. It’s not our type of intelligence, but it gets the right answer every time. However, the intelligence only emerges in the interaction of the slime mould with its environment, as both are part of the problem-solving system. The same might be said of humans, whose ‘extended mind’ includes various external items from shopp­ing lists to computers.

The ants’ technique for solving problems is so efficient it is used by robot designers. And even the slime mould itself is now being studied so that its methods can be copied for network design.6 It has genuine problem-solving ability for real-world problems, so it can certainly be described as intelli­gent according to the principles of soft AI. An aircraft doesn’t fly like a bird, but we can accept that the aircraft is really flying rather than just simulating flight. Equally the slime mould is not just simulating intelligent behaviour.

While the scientists may be happy with this resolution, it does leave the rest of us in a slightly awkward situation. If mazes are a valid test of intelligence, and a slime mould or an oil drop can solve them with ease, does it mean that they might be more intelligent than us?