Aussie researchers have been trying to get around one of the mostfundamentalproblems of robotic artificial intelligence - namely, how to get robots to solve complex problems without a complicated, large, and organic human brain. The solution, it turns out, is decidedly simple - use bee brains instead.
Researchers from RMIT, along with team members from the Université de Toulouse in France, conducted a study into whether smaller, less complicated brains in insects were capable of conceptual rule learning, a process that allows humans to synthesise 'rules' derived from multiple visual phenomena, and use them to decide a course of action.
Dr Adrian Dyer, the study author and Associate Professor at RMIT, said this kind of thinking is integral to the human ability to churn through complex variables to arrive to a quick decision.
“For example, if a driver wants to turn right at an intersection then they need to simultaneously observe the traffic light colour, the flow of oncoming cars and pedestrians to make a decision,” he said in a press release.
“With experience, our brains can conduct these complex decision-making processes, but this is a type of cognitive task beyond current machine vision.
"Our research collaboration between labs in Australia and France wanted to understand if such simultaneous decision making required a large primate brain, or whether a honeybee might also demonstrate rule learning.”
By setting up a maze with different elements involving above/below and left/right relationships, the researchers established an environment that required the bees to have juggle multiple abstract concepts (that of space, and that of visual difference) at the same time to pass through to their goal, the end of the maze, and pass through they did, rapidly learning what visual cues signalled which outcomes.
"This study reveals a surprising facility of brains to extract abstract concepts from a set of complex pictures and to combine them in a rule for subsequent choices," reads the study abstract.
"This finding thus provides excellent opportunities for understanding how cognitive processing is achieved by relatively simple neural architectures."
If the findings of the study are borne out, it suggests that complex learning schema can be created by relatively simple neural systems, which essentially means you could create learning in robots without a complicated 'brain'.
The study is published in the Proceedings of the National Academy of Science of the United States of America.
[via Science Alert]
EDITOR'S PICKS
