Member-only story
Ants and the Problems with Neural Networks
How Cognitive Science might Transform Neuroscience
Ants are pretty dumb. They live for a week and don’t do much besides walking around, looking for food and carrying twigs to their anthill (now that I think about it, we humans also don’t do much else).
But they are also dumb apart from living uninspired ant lives. They are dumb in a technical sense: an individual ant has just at the order of 250.000 (2,5*10⁵) neurons. As a comparison: an average homo sapiens has on average 80 billion neurons (8*10¹⁰), so if we assume that intelligence scales in at least some ways with the size of the brain (ignoring the fact that some animals have larger brains than we do), then we are approximately 320.000 times as smart as ants.
But despite the tininess of their brains, ants have their moments. They are constantly doing things that are so sophisticated that they could be taken straight out of a university-level math exam. One of these things is called dead reckoning.
Dead reckoning
When ants like the Cataglyphis bicolor go out foraging for food, they leave their home and move into the direction of an expected food source. As they don’t know precisely where the food source is to be found, they walk around in wobbly lines and circles (see this…