Nature: AI use in award-winning Nikon microscopy video deepens the controversy

Started by Finley_27, Today at 02:52 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Topic: Nature: AI use in award-winning Nikon microscopy video deepens the controversy   Views(Read 81 times)
Active members in this topic:
Finley_27(1)

Finley_27

We discussed the row over Nikon's Small World in Motion winner earlier today, and Nature has now added more detail. The winning video, by Ning Xu, shows cilia in the airways of a child with primary ciliary dyskinesia, a rare genetic condition. The controversy centres on purple, blue and red structures in the video that experts say do not look like real biology. Nature's Rachel Fieldhouse has the story. It is now looking less like a misunderstanding and more like a real problem

Bioengineering researcher Edward Phelps said the purple structures resemble mitochondria, but that a structure like that sitting below the epithelium does not occur in biology. That is a fairly specific scientific objection rather than a general complaint about AI. It goes to the heart of whether the image shows what was there. Other specialists have raised similar doubts

After the scrutiny, Nikon updated its description of the entry to say Xu used an unsupervised AI model for post processing. Xu explained that AI was used to enhance the visual presentation of regions below the cilia without making anatomical claims about what those rendered features represent. He also said the AI was only applied after the experimental data had been reconstructed

Developmental biologist Melanie White summed up the wider concern well, saying scientific images are not just illustrations, and that we need to be able to trust that what we see is grounded in the underlying measurement. That is the key tension here. Enhancement that makes an image clearer is one thing, adding structures that were never measured is another. Competitions will need clearer rules after this

Nikon's updated description at least makes the AI use transparent now. Does that settle it for you, or should an image with AI rendered features be eligible to win a science photography competition at all?

Here more than I should be