I wonder if Google should be using signals outside of YouTube to help their ML detection efforts. This example seems like a good one…if an account has a paid YouTube Music account that’s been in good standing for months/years, that seems like it should be taken into account. How many spammers would be willing to nurse an account through months/years of paid service only to have it pay off in a brief attempt at scamming people? If you added in listening data, you’d get an even clearer signal that the account is at least attempting to stay within the rules.

It feels like Google has it backwards…they’re siloing the fraud detection and banning across different products. If they switched that and did fraud detection across product boundaries and more targeted bans, we’d see far fewer of these AI-run-amok pleas to publicly shame them into getting a human to look into an obvious mistake.

Might not be worth making the accounts themselves, but would incentivize breaking into legitimate accounts, which might be worse optically

If Google were a solo schlub in their basement writing code because they didn't have time to do manual review, that would be understandable, even excusable.

But last I checked, they were the #5 largest company in the world.

You'd think that an issue that's obvious to literally everyone who uses the platform would also be obvious to even a handful of manual reviewers, and they could take the fight to the spammers a little more aggressively. Just a smidge. Like, one full-time reviewer, since it appears on my feed with devastating regularity, I'd even apply for the job myself, if it meant I could keep that scourge out of my own recommendations.

I'm left to conclude that maybe the #5 largest company in the world can't afford one full-time reviewer?

More likely they outsource the review job to a contractor who operates out of whichever country currently has the most lax labor laws. The contractor hires only the cheapest labor they can find and sets a quota like reviewing at least 360 complaints every hour for their 12 hour shift. Also, if the workers allow an actual bad faith objection to go through they get penalized.

I recently got spam in my gmail inbox, something related to a bitcoin opportunity.

Unknown sender, body full of red flag keywords, and an attached .htm file with an obfuscated javascript redirect to the scam domain.

I was honestly impressed they had achieved this. Evolutionary pressure for virii and scams is a real thing, I guess.