I am suspicious of this story, however plausible it seems.
The source is given as a book; the Economist writer Shashank Joshi explicitly says that they describe the 'same story' in their article ("(I touched on the same story here: https://economist.com/technology-quarterly/2022/01/27/decept...)"). However, if you look at the book excerpt (https://twitter.com/shashj/status/1615716082588815363), it's totally different from the supposed same story (crawl or somersault? cited to Benjamin, or Phil? did the Marine cover his face, or did he cover everything but his face? all undetected, or not the first?).
Why should you believe either one after comparing them...? When you have spent much time tracing urban legends, especially in AI where standards for these 'stupid AI stories' are so low that people will happily tell stories with no source ever (https://gwern.net/Tanks) or take an AI deliberately designed to make a particular mistake & erase that context to peddle their error story (eg https://hackernoon.com/dogs-wolves-data-science-and-why-mach...), this sort of sloppiness with stories should make you wary.
The source is given as a book; the Economist writer Shashank Joshi explicitly says that they describe the 'same story' in their article ("(I touched on the same story here: https://economist.com/technology-quarterly/2022/01/27/decept...)"). However, if you look at the book excerpt (https://twitter.com/shashj/status/1615716082588815363), it's totally different from the supposed same story (crawl or somersault? cited to Benjamin, or Phil? did the Marine cover his face, or did he cover everything but his face? all undetected, or not the first?).
Why should you believe either one after comparing them...? When you have spent much time tracing urban legends, especially in AI where standards for these 'stupid AI stories' are so low that people will happily tell stories with no source ever (https://gwern.net/Tanks) or take an AI deliberately designed to make a particular mistake & erase that context to peddle their error story (eg https://hackernoon.com/dogs-wolves-data-science-and-why-mach...), this sort of sloppiness with stories should make you wary.