Updated
Anyone who has seen a honeycomb knows bees can
understand geometry, but now there is proof their tiny brains are also
capable of learning basic arithmetic — and it could have implications
for AI developers.
Key points:
- RMIT researchers have taught honey bees how to do basic arithmetic using colour-coded mazes
- The findings indicate that big brains aren't needed to do maths
- The achievement of complex thinking with very few neurons could have implications for AI development
Researchers at RMIT have succeeded in teaching honeybees to do simple addition and subtraction in their heads, mid flight.
The experiment used special colour-coded mazes that presented the subjects with a maths problem and two possible answers.
The bees learned, over hundreds of trials, to either add or subtract one element from a number between one and five to select the correct path in the Y-shaped maze.
They earned a reward of sugar water when they made a correct choice, but received a bitter quinine solution if they chose wrong.
The findings, published today in Science Advances, imply that bigger is not always better, as far as brains go.
So just how big is a bee's brain?
"The simple answer is 'very small'", according to Associate Professor Adrian Dyer, the top bee expert from RMIT University in Melbourne."A bee's brain contains around 960,000 neurons — about the same number as in the retina of one human eye," he said."A pin head would be a very good analogy."
There was already evidence that honeybees could count, but PhD student Scarlett Howard wanted to see if they could process numbers according to mathematical rules, Dr Dyer told AM.
She saw the research as a contribution to the debate about whether a big brain is necessary to do mathematics.
She designed the experiment using colour as a code for addition and subtraction, rather than the symbols we use.
"The yellow meant that they should subtract and the blue meant they should add, so they knew that rule long-term," Dr Dyer said.
"Every time they came across a problem that involved blue, they would add information, and if they came across something yellow, they would subtract."
He said the bees learnt to very accurately choose the correct option.
"They probably had the ability to count and we taught them the rule of how mathematics can apply," he said.The mazes allowed the researchers to exclude "confounding factors" like wind and scent from the experiment to ensure robust data.
Working memory and long-term rules
Dr Dyer said numerical operations like addition and subtraction were complex because they required two levels of processing."You've got to call from one part of your brain: 'What is the rule?" he said.
"And then you've got to recognise local elements — two lollies, or two apples — and then you've got to put together the local information with the longer term rule to come up with a novel solution," Dr Dyer said.
He said the findings are exciting because it was unknown whether a "miniature brain" in a bee could do such sophisticated tasks.
"Fifteen to 20 years ago, the well-accepted position was that doing additions, subtraction, basic arithmetic is something we can do as humans, something very special about us," Dr Dyer said.
But now it looks like advanced numerical cognition may be found much more widely in nature among non-human animals than previously suspected.
"Many species can understand the difference between quantities and use this to forage, make decisions and solve problems," Dr Dyer said.
"But numerical cognition, such as exact number and arithmetic operations, requires a more sophisticated level of processing."
Ms Howard, the lead researcher, said mathematical ability was vital for human societies to flourish.
"Our findings show that the complex understanding of maths symbols as a language is something that many brains can probably achieve," she said.
Why and how?
It is unclear exactly why bees have evolved the ability to count, Dr Dyer said, but it probably has something to do with flowers — or more specifically, resources."We think counting flowers, keeping track of resources is possibly also very important for navigation," he told AM.
Although they are unlikely to need to add or subtract in the wild, bees' ability to acquire and manipulate learned information to make decisions could help in foraging, the researchers explained in the published report.
For example, remembering which flower traits, such as size, colour and shape, are associated with quality essential resources, and which flower traits are not.
He said bees' advanced capacity in relation to brain size showed they had to be very clever in how they used their resources for maximum efficiency.
And this new contribution to the big brain-small brain debate could have further reaching implications.
"If learning maths doesn't require a massive brain, there might also be new ways to incorporate interactions of long-term rules and working memory into designs to improve rapid AI learning," Dr Dyer said.
No comments:
Post a Comment