This particular issue (AI performance on captchas) is really quite fascinating. It's an arms race, but the problem is, only one side can win. Google is claiming they have improved their system in some (understandably) unspecified way, but there's only so far this can go. Captchas need to detect whether someone is human, but it has to work for everyone, ideally, even those with disabilities. Any simple task a human can do will eventually be able to be automated. Tasks that aren't currently feasible to be automated, say some natural language processing tasks, have another problem: scalability. To prevent simple databases of problems -> solutions, the problems need to be generated en masse, and for cheap, which means a computer needs to generate the solutions in the first place. And of course, paying people to just do captchas all day already happens.
The street address/book scan approach that Google uses is interesting in that the exact solution is not known, so they presumably have to be somewhat forgiving in accepting answers (as their machine learning might have gotten it wrong). Perhaps this is what their "risk analysis" refers to--whether their response seems "human" enough according to their data, not necessarily whether it's correct.
I don't see a way around this problem for free services that still preserves privacy (so directly using some government-issued ID is off the table). Maybe some Persona-like digital signature system, where a person can go to a physical location with a government ID, and get a signature that says "Trusted authority X affirms that person Y is in fact a person". Obviously this still has problems, as you need to trust X, and it's a big pain in the ass.
There are parallels to the realm of passwords, which are also becoming obsolete (not that there's a good replacement...). Anything that a human can feasibly remember for a bunch of sites is becoming easier and easier for computers to guess.
So basically, computers are taking over the world, and we can't do anything to stop it. God help us all.
The street address/book scan approach that Google uses is interesting in that the exact solution is not known, so they presumably have to be somewhat forgiving in accepting answers (as their machine learning might have gotten it wrong). Perhaps this is what their "risk analysis" refers to--whether their response seems "human" enough according to their data, not necessarily whether it's correct.
I don't see a way around this problem for free services that still preserves privacy (so directly using some government-issued ID is off the table). Maybe some Persona-like digital signature system, where a person can go to a physical location with a government ID, and get a signature that says "Trusted authority X affirms that person Y is in fact a person". Obviously this still has problems, as you need to trust X, and it's a big pain in the ass.
There are parallels to the realm of passwords, which are also becoming obsolete (not that there's a good replacement...). Anything that a human can feasibly remember for a bunch of sites is becoming easier and easier for computers to guess.
So basically, computers are taking over the world, and we can't do anything to stop it. God help us all.