If you ask a team of British computer scientists, cell wisdom It is more than a poetic turn of phrase. These researchers have actually studied the brains of honeybees to develop a new kind of machine intelligence for decision-making.
While a bee’s life seems simple on the surface—flying through the air, sampling nectar and pollinating plants wherever it lands—the decision-making process behind this seemingly random walk is more complex than meets the eye. The basic method of operation within the apian neural network seems to be one that locates ideal floral targets within seconds with a high level of accuracy. At this point, the bee flies off to its new feast. And all of this takes place inside a brain the size of a sesame seed with fewer than a million neurons.
“It’s surprising that bees’ decisions are not only highly adaptive and precise but also faster when it comes to (making) choices.”
—HaDi MaBouDi, Opteran Technologies
HaDi MaBouDi is a senior researcher at Opteran Technologies – a brain biomimetic start-up at the University of Sheffield – and one of the co-authors of the new study. Mabode says he and his colleagues were drawn to studying the process behind bees’ decision-making because of the lessons that artificial systems might learn.
“Algorithmic decision-making is often deficient in terms of adaptability, speed, accuracy, and risk aversion, especially when resources are limited—qualities that bees possess,” he says. “By incorporating principles derived from biological models, AI systems have the potential to enhance their efficiency, robustness, and risk aversion.”
To better understand what really turns the gears in the bees’ brains, the research team first observed the behavior of twenty bees as they explored colorful flowers in an artificial garden. The flowers had a mixture of sugar syrup, tonic water, and distilled water. After observing the bees’ choices in a number of experiments, Mabodi and his colleagues saw that the bees made decisions in a way that went against common sense.
“Surprisingly, bees’ decisions are not only highly adaptive and precise, but also faster when it comes to making the right choices than when it comes to making the wrong choices,” he says. “This contrasts with the typical speed-accuracy trade-off observed in animals and synthetic systems, where accurate decisions tend to take longer than imprecise ones…bee decision-making exhibits a level of complexity that parallels certain aspects of decision-making in higher animal species.”
And all of this takes place inside a brain the size of a sesame seed with fewer than a million neurons.
This means that bees’ split-second decisions about a flower are more accurate than decisions they take longer to consider. Confident, “low-level” decision-making can benefit synthetic systems by reducing the amount of training and rule organizing they need, says research co-author James Marshall.
Marshall is Professor of Theoretical and Computational Biology at the University of Sheffield and co-founder of Opteran Technologies.
To mimic this behavior, the researchers designed a model with two parallel decision paths—one for acceptance and one for rejection. Like the biological neural networks of bees, these pathways are designed to be adaptive and to weigh the quality of stimuli to aid their decision-making. The model also retained memory information for previous stimuli to help it remember stimuli that were not worth exploring a second time.
“Simulating the bees’ risk-averse strategy was of particular interest, as they will only accept a flower if they are confident it will provide a reward, and reject it otherwise,” says MaBouDi. “This strategy allowed the bees to focus their efforts on the flowers with the highest potential for nectar.”
To see how well their model stacked the bee, the team gave it 25 randomized trials of high-reward and low-reward stimuli. They found that the model’s response rates were equal to those of real bees, and even resembled the physical layout of bees’ brains. However, that doesn’t necessarily mean that their model mimics bee brains quite faithfully, says Mabodi.
“Given such a large discrepancy in scale between biological brains and computational models, direct implementation of such a model on devices comparable in size to a bee’s brain is highly unlikely,” he says. “[But]focusing on key principles and mechanisms can lead to the creation of intelligent and efficient systems that demonstrate similar decision-making capabilities within the limits of available hardware.”
Mabodi, Marshall, and their colleagues are excited to explore what this could mean for the future of neural computing as well as for miniature devices. Robotic decision making inspired by biological models such as bees could play an important role in helping autonomous robots, such as those used in mining or search and rescue, navigate unfamiliar terrain and make adaptive safety decisions.
Marshall says the bees are just the beginning. Opteran is actively investigating other types of insect brains for other solutions to AI, robotics, and computer science problems that may be hiding in nature.
“Our task now is to understand more and more of the insect brain and how it generates the behavior it does, especially for some of the more interesting insect species such as social insects, for example ants,” he says. “How individual insects interact may help us come up with better traffic-management behaviors for fleets of autonomous warehouse robots, for example.”
The group published its research last month in the journal eLife.
of your site articles
Related articles around the web