Research – Machine vision as good as human vision

An international research breakthrough with bees offers the possibility that machines might soon be able to see almost as well as humans. French and Australian researchers have shown that honeybees use multiple rules to solve complex visual problems. Dr Adrian Dyer, from RMIT University in Melbourne, said the findings held important implications for our understanding of how cognitive capacities for viewing complex images evolved in brains. Dyer said that rule learning was a fundamental cognitive task that allowed humans to operate in complex environments. “For example, if drivers want 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. With experience, our brains can conduct these complex decision-making processes, but this is a type of cognitive task beyond current machine vision.”

The researchers 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. Research team lead author Dr Aurore Avargues-Weber, of the Université de Toulouse, trained individual honeybees to fly into a Y-shaped maze; this presented different elements in specific relationships such as above-below or left-right.

With extended training the bees were able to learn that the elements had to have two sets of rules, including being in a specific relationship such as above-below, while also possessing elements differing from each other.

Writing in the Proceedings of the National Academy of Sciences of the United States of America, the scientists said that sorting objects and events into categories and concepts is a fundamental cognitive capacity that reduces the cost of learning every particular situation encountered in our daily lives. Relational concepts such as same; different; better than or larger than – among others, – are essential in human cognition because they allow highly efficient classifying of events irrespective of physical similarity, according to the four researchers.  Mastering a relational concept involves encoding a relationship by the brain independently of the physical objects linked by the relation and is, therefore, consistent with abstraction capacities.  Processing several concepts at a time presupposes an even higher level of cognitive sophistication that is not expected in an invertebrate.

The scientists found that the miniature brains of honeybees rapidly learn to master two abstract concepts simultaneously, one based on spatial relationships (above-below and right-left) and another based on the perception of difference. Bees that learned to classify visual targets by using this dual concept transferred their choices to unknown stimuli that offered a best match in terms of dual-concept availability: their components presented the appropriate spatial relationship and differed from one another.

The 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. This finding thus provides excellent opportunities for understanding how cognitive processing is achieved by relatively simple neural architectures. Dr Dyer said the results showed that possessing a large, complex brain was not necessary to master multiple simultaneous conceptual rule learning. This offered the possibility of deciphering the neural basis of high-level cognitive tasks because of the simplicity and accessibility of the bee brain.