Welcome to the latest installment of “Five words from …” our series which highlights interesting words from interesting books! In this book, Arvind Narayanan and Sayash Kapoor, two of TIME’s 100 Most Influential People in AI 2023, explain how AI works (and why it often doesn’t), explore AI’s limits and risks, and outline where AI is a useful tool and where it’s not just empty hype but actually harmful.
“To appreciate how common it is for regular users to try to evade content moderation, consider algospeak: words or phrases that are widely understood and adopted by social media users as a way to avoid being mistakenly penalized by fickle content moderation algorithms.”
Neologism researcher Brianne Hughes has created a Wordnik list of “Algorithm Avoidant Inventions” to collect algospeak examples.
“One ambitious effort is the theory of cliodynamics by Peter Turchin, which applies mathematical models to populations.”
The ‘clio-‘ of ‘cliodynamics’ comes the name of the muse of history in Greek mythology.
“Researcher Lee Vinsel called this phenomenon criti-hype—criticism that tends up portraying technology as all powerful instead of calling out its limitations.”
Vinsel created the word ‘criti-hype’ in a 2021 Medium post titled “You’re Doing It Wrong: Notes on Criticism and Technology Hype“.
“In 2011, Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton decided to take a crack at the ImageNet competition using neural networks, which by then had been branded “deep learning” because of the key insight that having more layers (depth) improves accuracy.”
Neural networks, despite the name, are not intended to realistically model the behavior of actual neurons.
“Instead of removal, the post can be slapped with a warning, or, if it is a “borderline” policy violation, it might be silently shown to fewer users than it otherwise would. This is a notable development in the last few years known as downranking or demotion, or, colloquially, shadowbanning.”
Rich Kyanka, the creator of Something Awful, claims that the term ‘shadow ban’ was created on that forum. A 2018 explainer from Vice, “Where Did the Concept of ‘Shadow Banning’ Come From?” highlights similar practices, including ‘twit bit’, ‘bozo filter’, and ‘toading’.