And it's also an interesting proof of concept -- it's hard to imagine that this is the end of the story. Maybe it can be worked into a simple "check your conjectures" tool that would be of great help to mathematical researchers. 6/6— Timothy Gowers (@wtgowers) May 1, 2021

In other words, the ‘success’ of technologies like self-checkout machines is in large part produced by the human effort necessary to maintain the technologies ​ Their broader point is that automation doesn’t eliminate human labor; it often leads to its disguising and devaluation ​ Customers have to bridge the gap between the norms of human customer service and the company’s imposition of inhuman staff shortfalls without completely losing their patience or simply taking what they couldn’t find a reasonable way to purchase. ​ That is, they rely on the kinds of consideration from users that the companies themselves don’t practice. ​ They have prioritized user growth, data collection, and advertising over providing a safe, reliable service to users. It’s no wonder that some users will adopt the same attitude and the same goals: try to spread unwanted messages as far as possible.

that, more or less, for some reason the machine learning algorithm for autocorrect was learning something it never should have learned.