Are We Still Providing Useful Training Data?

"I seriously doubt humans can do a better job of identifying any of those images than Google's already trained models can," he says. "It feels kind of like taking those multiple choice exams, you know, when your teacher has absolutely not figured out the value of unambiguously correct and incorrect answers" He adds that Cloudflare is also talking about theirs because they're trying to keep bots out of websites as part of their business model. 'You have ultimately a finite set of resources and you're being forced to expend more of them'