To that I say, good—if you can't support it, you shouldn't ship it. At least when it comes to taking people's money or promising to revenue-share (like YouTube creators).

In any case, Alphabet's net earnings in 2021 were $76,000,000,000. They could hire 760,000 people at $100k cost per year and still break even. It wouldn't take anywhere near that number to introduce humans into critical loops like copyright strikes and ad revenue customer support, so they could easily–easily—afford it. I think they just don't want to, because they'd rather make all the money and they don't institutionally care about people ground up in the cracks.

But in this case, those problems need to be disclosed upfront. Google should have a big banner upon signup that says "this is a fully automated service and you have no recourse if the automation goes wrong", to warn people against relying on them for anything critical.

They probably do, but it's somewhere in 10MB document written in legalese indistinguishable from Klingon.

>the cost of serving the sheer number of people Google serves

That cost is mostly imaginary. First, it’s already tiny - computers are cheap - and second, 90% of that cost is for stuff that only matters to Google; you could make that part disappear without hurting user experience.

We're talking explicitly about the cost of putting humans in the loop as content moderators / customer-service representatives / ombudsmen / etc. That cost, industry-wide, grows at O(N) with the number of customers; as some constant fraction of people will always call with their problems rather than attempting to engage with automated support systems.

So? That doesn’t justify shifting that costs to society. You’re assuming corporations have to be allowed to do whatever they want, because otherwise they would earn less money, and you’re assuming that’s not an acceptable outcome.