Microscopy competition winner sparks debate over artificial intelligence in science

A prestigious videography competition featuring microscopic life is facing a debate after scientists questioned the biological accuracy of its winning entry. The Small World in Motion competition, run by Nikon Instruments, named optical engineer Ning Xu as its winner on September 15 for a video showing hair-like cilia from lung tissue. Researchers on social media soon raised concerns that the video contained features that do not occur in actual biology and might have been generated by an artificial-intelligence tool.

Contest Details And AI Use

Xu took his tissue sample from a child with primary ciliary dyskinesia, a rare genetic disorder affecting the lungs, sinuses, and ears. His winning video shows cilia alongside red, purple, and blue structures that were not explained in the original contest blog post. University of Florida bioengineering researcher Edward Phelps noted online that the purple shapes resemble mitochondria and the blue shapes look like nuclei, but neither behaves or appears like real biological structures of that type. Nikon Instruments updated its blog post on September 22 to state that Xu used an unsupervised artificial-intelligence model to help with post-processing.

Responses And Context

Xu wrote on LinkedIn that his team used post-processing to improve the visual look of areas below the cilia without making claims about what those features represent. He stated that artificial intelligence was not used to create the original experimental movie or the cilia motion, but was applied afterward to grayscale data to color structures with similar shapes for a more engaging presentation. Developmental biologist Melanie White from the University of Queensland cautioned that scientific images serve as data rather than mere illustrations, meaning researchers must be able to trust that visuals accurately reflect underlying measurements.

Context

Artificial-intelligence tools are increasingly used across scientific research and data visualization, raising new questions about how to verify the accuracy of processed images. Competitions that blend scientific subject matter with artistic presentation often navigate a fine line between visual appeal and strict data integrity. When imaging software alters or enhances underlying data, researchers face challenges in ensuring that published figures do not introduce misleading artifacts into scientific observation.