Trinaries and the Pitfalls of Medical AI
Bowler Hat Science from Matthew R Francis
I have successfully moved to my new house, though much of the furniture is still displaced and a lot of stuff is still in boxes. But! the hard part is done, and naturally I have to get back to work. Meanwhile, I have two articles published recently for you to read until I recover enough to write something original for the newsletter.
Distant Icy Twins Might Actually Be Triplets
The trans-Neptunian object Altjira, 44 times farther from the Sun than Earth is, could be the second known trinary, confirming a theory about the formation of our solar system.
For AGU Eos:
The cold outer reaches of the solar system are home to a plethora of small worlds, many of which have moons of their own. For a few, the moon is massive enough to make the pair into a binary; the Pluto-Charon system is the most famous of those. And a small icy body named Lempo is a trinary: three objects of comparable mass in mutual orbit.
Now astronomers have identified another possible trinary object, a very distant world known as Altjira (al-TCHEE-ruh), named to honor the creator deity in Arrernte Australian cosmology. Observers discovered Altjira in 2001 and its as yet unnamed moon in 2006. With data accumulated over the past 2 decades, researchers determined the path of the moon doesn’t match what would be expected if it were orbiting a spherical (or mostly so) world.
When Artificial Intelligence Takes Shortcuts, Patient Needs Can Get Lost
For SIAM News:
A patient checks into the hospital emergency room, where staff rush her to the intensive care unit (ICU). The radiologist feeds her chest X-ray into an artificial intelligence (AI) program,1 which reports that she likely has pneumonia. Another patient with glaucoma visits his ophthalmologist, who takes a routine image of his retinas. The AI that analyzes the image determines that the patient also has type 2 diabetes.
Both of these conclusions turn out to be correct, but for entirely coincidental reasons. ICU patients are far more likely to have pneumonia than other hospital patients, and type 2 diabetes is more prevalent in older people. In other words, the computer model took a shortcut by drawing an inference based on incidental details rather than actual diagnostic factors. To exacerbate the situation, when medical workers require AI systems to control for these details, the models fail to identify actual medical problems from images.
And a Nebula
Bowlerhattishly thine,
Matthew
Support me: