All the buzz: As technology advances, scientists and engineers wage a daily war against limitations in size, scalability, and power consumption. Rising AI and data demands continuously drive the need for large-scale power and cooling in data center hardware, which performs everything from simple analysis tasks to complex computation. Despite these growth trends, a group of EU researchers is looking to tackle these challenges from a very different angle – and they're doing it by studying some of nature's smallest and best navigators.
GPS is more than just a way for us to get from point A to point B. Our smartphones, vehicles, drones, cameras, wearable devices, and countless other technologies all rely on it, and the complex computation supporting it, to enable the modern conveniences we enjoy everyday.
While some client devices may be small, the amount of compute power supporting them to process, transmit, and present data GPS data is immense. What if some of that compute and decision making power could be offloaded and delivered by something no larger than a seed using mere microwatts of power? That's exactly what the team at InsectNeuroNano is trying to do today.
To put that into perspective, a bee is navigating a 3D space and making flight corrections based on real-time data while communicating with other bees and locating food – all on roughly 10 microwatts of power.
A typical honeybee measures approximately 1/2 inch (12.7 mm), can weigh anywhere from 0.0035 to 0.007 ounces (~100 to ~200 mg), and may fly anywhere from 1 to 6 miles (1.6 to 9.7 km) in search of food. Bees come equipped with an "onboard GPS," which allows them to determine their position by analyzing patterns in the sky and evaluating those patterns against their own speed.
This ability enables them to make multiple trips to and from the hive each day without getting lost, while processing the equivalent of an estimated 10 trillion operations per second and using anywhere from 10 microwatts (0.00001 watts) to one centiwatt (0.01 watts) of power.
To put that into perspective, a bee is navigating a 3D space and making flight corrections based on real-time data while communicating with other bees and locating food – all on roughly 10 microwatts of power. An Intel Core Ultra 285K, which can be more than 300 – 500 times the weight of the average bee, would require anywhere from 10 to 15 watts to power both the NPU and SoC while processing a similar 10 TOPS workload. Even at a conservative 10 watts, the bee uses 1,000,000 times less power to complete its workload.
Lund University's Anders Mikkelsen, along with researchers from several universities across Europe, is on a mission to build a low-power, insect-inspired chip that can mimic the bee's ability to determine its position with no external processing or overhead.
The team's approach focuses less on traditional AI learning and more on concepts rooted in natural intelligence, using nanophotonic circuits. Unlike traditional circuits, nanophotonic circuits guide light through extremely small chip structures, measuring billionths of a meter across. The end result could yield highly specialized chips capable of performing one very specific but highly efficient task.
If successful, these tiny specialized chips could support uses ranging from environmental sensors to small autonomous robots tasked with cleaning up pollution, building structures, or enhancing pollination in targeted areas. But we're not likely to see clouds of robotic insects in the skies anytime soon. According to Mikkelsen, we may still be as many as ten years away from today's prototypes becoming tomorrow's bug swarms.
Image credit: Honey Bees by Terski