apparently the plan was to use humans to train an ML system but it never ended up getting good enough to take over
Once in SF when it was new for the novelty and once when I was in Seattle last year because it was close to where I was staying.
Like, there were genuinely a few hours between the moment I stepped out of the store and me getting the invoice in my phone. I don’t know what’s the point of over engineering this and not hiring cashiers (or even have self checkout).
(I mean I know the point is to save enough money to not hire humans but like, clearly that was a shit bet)
Back when i worked at Google (fuck them btw. it's become a horrible place to work), I was one of several linguists on a team where our whole thing was writing questions like "Which of these two ads is more relevant to this search query?" and then sending those out to a bunch of part-time contractors who worked for a vendor in order to collect training data for Google's search ad serving algorithm.
Almost all of the English language raters were in India or the Philippines because you can pay people in those countries way less than you'd have to pay people from the US.
Raters also got hijacked sometimes to directly enforce policy, like when YouTube had an "adpocalypse" due to Elsagate, or to judge if feed ads (for endless vertical scroll on mobile) contained adult or violent or drug-related content.
So I'm confident when I say that a ton of systems that are presented as being "automated" are built on top of cheap human labor, or just are cheap human labor.
Source: I applied for a part-time remote position a couple years ago thinking to get some side cash. Before I could be hired I had to read their extensive guidelines and then rate a bunch of results. All the automated feedback I got was good, but "mysteriously" I never got past that stage (I got a rejection, then an invite to continue, and then another rejection? Bewildering). Fast forward to me learning more about how these things are trained and now I'm a liiiittle 🤔
(I know I might've just not made it in because others were more qualified, but idk, something about the depth and scale of that assessment felt shifty to me. You shouldn't give a "skills assessment" to people based on something they have to read a whole manual for, that's training and should be either assumed as part of the skillset the potential hire already has, or included as part of onboarding after they're hired, depending on how much you want your entry level salary to be. So even if it wasn't a free labor scam, it was definitely unethical and I dodged a bullet there.)
