Artificial intelligence learns from one of the most important insects on the planet, honeybees

New research shows how honeybees are able to make quick and accurate decisions about which flower to land on, which could help design more efficient artificial intelligence (AI) in the future.

When it comes to making quick and accurate decisions in their arduous and efficient daily work, honeybees are number one in the animal kingdom. They need to know which flowers are most likely to offer them nectar or pollen. But how do they do it? To understand this, scientists at the University of Sheffield examined the choice process these animals go through.

Whenever a bee starts collecting nectar, it must use small variations in color or smell to decide which flower to land on and explore. An erroneous decision in this respect is wasteful, since it wastes unnecessary energy and exposes the bee to risks and dangers. The extraordinary thing is that, for this process, bees only have at their disposal a brain the size of a sesame seed, with less than a million neurons.

For research purposes, the scientists released 20 bees into a field containing artificial flowers of five different colors. The bees were trained to link each color with a certain probability of receiving a sugary liquid as a reward or bitter quinine as a punishment; they were then tested on this knowledge. During 18 trials, each color offered the bees a different probability of reward and punishment.

The five colors offered the reward in 100 %, 66 %, 50 %, 33 % and 0 % of the training trials and were punished otherwise. The color rewarded on 100 % of the training trials was never punished, whereas the color rewarded on 0 % of the training trials was always punished.

From the response times and accuracy rates they revealed a complex pattern of decision making. “How quickly the insects made a decision and the types of errors they made (such as deciding to explore an unrewarded flower or ignore a rewarded one) depended on both the quality of the evidence and the certainty of the reward. Such sophistication and subtlety in decision-making is comparable to that of primates,” the researchers stress in the study.

Before the bees accepted or rejected the stimuli, the researchers found that the bees fluttered near and in front of the stimulus. “We assume that the bees were sampling information about the stimulus.” The results revealed that, if they were sure that a flower had food, they decided to land on it within 0.6 seconds. If they were sure there was no food, they decided not to land on it.

According to the authors, the research reveals the sophistication and subtlety of honeybee decision-making while emphasizing the importance of considering both acceptance and rejection responses in animal behavior research. “By shedding light on the neural mechanisms underlying decision making, our findings not only provide valuable information on honeybee behavior, but also propose a potential framework for the development of robust autonomous decision-making systems with applications in the field of robotics.”

Indeed, for Professor James Marshall of the Department of Computer Science at the University of Sheffield, “research has shown how bees are able to make complex autonomous decisions with minimal neural circuitry. Millions of years of evolution have led bees to have incredibly efficient brains with very low energy requirements. This biology may inspire the future of AI.”

Published by Emirates Herald, news and information agency.

Leave a Reply

Your email address will not be published. Required fields are marked *