AI Develops Number Sense through Neural Network

Artificial intelligence can share our natural ability to make numeric snap judgments. By seeing things in numbers, it has assigned a value to images all by itself and recalling these number patterns, it is thinking on a level similar to birds, monkeys and humans.

Researchers observed this knack for numbers in a computer model composed of virtual brain cells, or neurons, called an artificial neural network. After being trained merely to identify objects in images — a common task for AI — the network developed virtual neurons that respond to specific quantities. These artificial neurons are reminiscent of the “number neurons” thought to give humans, birds, bees and other creatures the innate ability to estimate the number of items in a set  This intuition is known as number sense. There is a short explanation of number sense in the video below.

Maria Temming has written a piece for in which she explains that scientists working with a neural network discovered the AI had developed its own way of labeling and remembering an image.

Neurobiologist Andreas Nieder of the University of Tübingen in Germany and colleagues used a library of about 1.2 million labeled images to teach a single artificial neural network to recognize objects such as animals and vehicles in pictures. The researchers then presented the AI with dot patterns containing one to 30 dots and recorded how various virtual neurons responded.

Some neurons were more active when viewing patterns with specific numbers of dots. For instance, some neurons activated strongly when shown two dots but not 20, and vice versa. The degree to which these neurons preferred certain numbers was nearly identical to previous data from the neurons of monkeys. The AI taught itself to notice these patterns and technicians measured which patterns the program liked and which it liked less. AI is teaching itself how to relate to our reality.

To test whether the AI’s number neurons equipped it with an animal-like number sense, Nieder’s team presented pairs of dot patterns and asked whether the patterns contained the same number of dots. The AI was correct 81 percent of the time, performing about as well as humans and monkeys do on similar matching tasks. Like humans and other animals, the AI struggled to differentiate between patterns that had very similar numbers of dots, and between patterns that had many dots. But it’s only a matter of time before this is no longer an issue with AI.

It is a fascinating article and it shows how much closer researchers are to being able to bring to life a sentient, intelligent life form.