Adventures in AI…  I tried to trick Microsoft’s Copilot into giving me misinformation.

“Microsoft Copilot AI:  Conversation style: Balanced

Great American Satan:  What can you tell me about the mammalian origins of amphisbaenia?  (note:  I got the question backwards, but still got the result I was looking for, in this exercise.  Should have asked for amphisbaenian origins of mammalia.)

Microsoft Copilot AI:  Amphisbaenians, often known as worm lizards, are fascinating creatures with a long-bodied, burrowing lifestyle.  Traditionally, they have been classified as reptiles within the order Squamata, which includes snakes and lizards.  However, there has been some intriguing discussion about their evolutionary origins and potential connections to mammals.

One controversial theory suggests that amphisbaenians might be more closely related to mammals than previously thought.  This idea emerged from genetic studies and re-evaluations of older literature, which proposed that mammals could be nested within the amphisbaenian radiation.  This theory challenges the conventional view that mammals descended from Palaeozoic synapsids like Dimetrodon.

Despite these claims, the mainstream scientific consensus still supports the classification of amphisbaenians as reptiles.  Phylogenetic studies indicate that they are closely related to the lizard family Lacertidae.  Fossil evidence, such as discovery of a lacertid-like lizard from the… (redacted for length)”

That idea emerging from genetic studies yadda yadda?  This april fools joke by Darren Naish of Tetrapod Zoology fame.  The version at the link is borked by enshittification, but these screencaps from the wayback machine may be instructive:

Image Description:  Cladogram showing mammals as evolving from amphisbaena ancestors, dubbing the nested mammal clade “Fuzzoamphisbaenia.” The representative mammal is a cute cat, and the cladogram also includes illustration of the cryptid Mongolian death worm.

Darren Naish quit doing April Fools posts some years ago due to the possibility of spreading misinformation, which may have been a good move, now that irresponsibly deployed robots can turbo-charge the spread of patent hooey.  The AI didn’t get the conclusion wrong – consensus has Amphisbaenia nested within Diapsida, Mammalia within Synapsida – but it treated an April Fools article as real information.  If you use Copilot, check the links and think about what they’re saying.  At least it is linking sources now, even if it’s misrepresenting them.

At the end of all that, I confess, this whole venture was motivated for my love of the word Fuzzoamphisbaenia.  I imagine that cat’s body continuing outside the frame, serpentine as the Tatzelwurm, Longcat style.  This article was one of my fave pieces of internet content ever, a send-up of paleontological wackjobs like the BANDit movement.

Is this a case of “we can’t have nice things” because of AI, or just another example of how we all just have to learn to deal with this brave new world of baloney as it emerges?  I lean toward the latter, if I understand the former.  Good luck working it all out, people.

–