Japanese optics company Nikon has stated that it is “carefully reviewing” information submitted by optical engineer Ning Xu of the National University of Singapore after commentators accused him of using AI in his winning entry in the 16th annual Small World in Motion competition. Small World in Motion bills itself as a celebration of the beauty of optical microscopy. According to Nature, the video presented by Xu showed the movement of cilia in lung tissue samples from a child with a chronic disease called primary ciliary dyskinesia. Unidentified red, blue and purple structures can be seen beneath the cilia. Nikon Instruments, the division of Nikon that sponsors the contest, announced Xu’s winning entry in mid-September and reposted it on LinkedIn last week. It quickly faced a firestorm of criticism from users who pointed out major discrepancies. For example, Edward Phelps, a bioengineering researcher at the University of Florida in Gainesville, posted on LinkedIn: “The purple structures resemble mitochondria, but a subepithelial structure composed of extracellular mitochondria the same size as cell nuclei does not exist in biology.” Phelps added that the blue nodules looked like nuclei but did not act like nuclei, and that he has no idea what structures are shown in red. Another LinkedIn user, UT Southwestern Medical Center doctoral student Ian Donovan, noted that the video had a SynthID watermark, according to CNN. Those watermarks were pioneered by Google’s DeepMind research lab and have since been adopted by a host of AI companies. Most AI detectors that attempt to detect AI by other means perform poorly, and without SynthID or equivalent watermarks, it can be difficult or impossible to determine the provenance of an image with complete certainty. Fifth place winner Patrick Hickey told the BBC that microscopists do not have many opportunities to show their work, which often has “very abstract and artistic qualities”. A widely cited set of scientific guidelines allows images of stained samples or colored images, as long as the changes are revealed and described. That article was from 2010, before generative AI came along and started making pixels directly. “There’s a specific set of rules in the contest, and one of them was pretty clear: you can’t use generative AI to produce things, and the other obviously needs to be taken under a microscope,” Hickey added. In updates on LinkedIn last weekend, Nikon said Xu was cooperating. “Xu respectfully complies and has provided detailed technical documentation describing the microscopy equipment, imaging methods, and processing techniques used to create the original video,” Nikon wrote. I had previously updated a blog post about the winning entry to clarify that “An unsupervised neural network method was subsequently used as part of the visualization and post-processing workflow to distinguish and visualize features in the grayscale data, creating a more vivid and visually appealing video.” Xu has said the same thing in other LinkedIn posts, Nature noted, saying his team had used post-processing on the structures beneath the cilia “without making anatomical claims about what those depicted features represent.” Xu also stated that the original video of the cilia and their movement is real, and that AI was only used on the “reconstructed grayscale data to distinguish and color structures with similar morphology.” As of Friday evening, a spokesperson for the contest organizers did not have any updates on the Gizmodo situation. Developmental biologist Melanie White of the University of Queensland in Brisbane, Australia, told Nature that scientific images are data and that scientists “need to be able to trust that what we’re seeing is based on the underlying measurement.” Former Small World in Motion judge Andrew Moore told The Telegraph that Xu’s entry made him think of “old photographs restored with AI.” “The high-resolution colorized photo may look good, but if it was your grandmother who changed the face, it will be disturbing,” Moore told the newspaper.
