I agree, and the word I would use is competing machines. Every time me (or my algo) bid up a security price I remove that "cheapness" signal from a competitor,

I agree, and the word I would use is competing machines. Every time me (or my algo) bid up a security price I remove that "cheapness" signal from a competitor, increasing the noise they have to deal with and vice versa. ML hates noisy data. Returns are very noisy. ML has a much easier time with data generated from situations that are lacking competition, such as data from any process inside a company, e.g. consumer behavior. These are problems that want to be solved.