Looking for true intelligence, study shows that slime molds can learn

Slime molds make the news sometimes. It's usually because a scientist figured out that a slime mold could solve a maze in a seemingly intentional way, or that it can grow through a space in an efficient enough pattern to make mathematicians sit up and notice. The implication is usually the same: Slime molds have intelligence. But do they? After all, just because something is complex and seemingly highly ordered doesn't mean it couldn't come about thanks to entirely non-intelligent processes -- just look at the evolution by national selection that gave rise to the slime mold itself.

This week, researchers from France's Toulouse University published(Opens in a new window) the newest piece of evidence in pursuit of a slime mold's true IQ, approaching the question with a clever experimental design. They found that the species is capable of un-learning a natural reaction to stimuli and storing that new understanding for application down the line. Just how it did this is beyond the current understanding of science, but it does confirm what prior research had already suggested: Whether it's the result of high intelligence or pure molecular cause and effect, slime molds are exquisitely well adapted to dealing with all new situations.

The experiment was designed like this: a slime mold was presented with two bridges over a physical gap separating the mold from some food. Some molds had to cross a "bitter" bridge laced with either the substance quinine or caffeine, while others were free to cross bridges clear of any contaminant. Molds presented with the bitter bridge at first recoiled, avoiding the chemical-laced area. This is presumably an evolved response to this sort of sensation, since it's associated with toxic substances -- and indeed, both quinine and caffeine can be toxic themselves, in high doses.

But as small interactions with the bitter edge of the chemical became more numerous, the mold somehow noticed something: bitter or not, the bridge wasn't doing it any harm.

Soon enough, the molds habituated to the bitterness and crossed to eat the food. This began with a tentative finger of slime, a thin projection out into the danger zone that minimized contact with the surface. As even this produced no negative results, however, the slimes lost this prudent measure and simply rushed across. By the sixth day of testing, the organism basically ignored the bitter substance, and treated toxic bridges like clean ones.

This could simply be due to desensitization, so the team tried exposing quinine-tolerant mold to caffeine bridges and vice-versa. They found that the molds all showed a specific distrust of the new taste in spite of their prior conditioning to trust the other chemical. This implies that there could be a real learning process going on -- though just what that process could be, few will speculate.

Does the behavior of slime molds constitute intelligence? Slime molds are a testament to what you can achieve when you have many simple, independent actors all working together -- which is why they have been associated with some of the very same pattern-finding problems assigned to artificial neural networks, the other big example of a complex network of simple actors.

Really, this question of intelligent mold hangs on the same problem as determining a true artificial intelligence, on the extreme other end of the complexity scale: What level of sophistication in conditioned responses is sufficient to constitute real intelligence? A computer can solve a maze, and map efficient transit routes between cities -- but it can do these things because we programmed it to be able to. The only difference with the slime mold is that it is biological, and thus programmed by evolution, just like humanity itself.