The ability of rats to drive these cars demonstrates the “neuroplasticity” of their brains, says Lambert. This refers to their ability to respond flexibly to novel challenges. “I do believe that rats are smarter than most people perceive them to be, and that most animals are smarter in unique ways than we think,” she says.
Makes me wonder how the neural networks behind some of the autonomous cars compare to rat and human brains in terms of number of nodes and number of connections.
To the Bat computer!
This tutorial for using DeepLearning to drive a simulated car involves 2.7 \cdot 10^7 connections.
This publication deals with models of comparable size.
In both cases, we are dealing with simpler problems than operating an actual car under actual traffic conditions, but then so are the rats.
This convenient Wikipedia page indicates that the brown rat has 2 \cdot 10^8neurons and 4.5 \cdot 10^{11} connections.