Against the Grain

The intelligent part of artificial intelligence (AI) is mimicking our thought processes – making decisions based on logic and knowledge, and perhaps learning new rules along the way to hone these choices. But as deep-learning models grow more complex, the logic driving them can be lost in the code. Pictured under a scanning electron microscope, spotting different varieties of pollen grains, is a fresh challenge for artificial intelligence. A team of computer scientists and bioengineers sort grains by their different chemical and shape (morphological) properties using a form of explainable AI (xAI) – in which the artificial ‘thought processes’ are transparent and easy to follow. Apart making an excellent testing ground for xAI methods, the team are able to compare different models for classifying pollen, which may ultimately improve how we estimate pollen counts, make forecasts and promote advice to millions of people suffering allergies worldwide.

Written by John Ankers

John Ankers

Writer

A former researcher in systems biology, John is now an online biology tutor and professional coach supporting the wellbeing of scientists and NHS professionals. He's writing, too, of course.

Website

• Image from work by Sanja Brdar and colleagues
• BioSense Institute - Research Institute for Information Technologies in Biosystems, University of Novi Sad, Novi Sad, Serbia
• Image originally published with a Creative Commons Attribution 4.0 International (CC BY 4.0)
• Published in Scientific Reports, February 2023