Hacker News Reader

Comment by its_bbq | Hacker News Reader

Former Jigsawyer here. I think this article is pretty fair to Perspective given that it was never meant to be used in a fully automated way, just as a first pass to help forum moderators.

It's very difficult when you blur the lines of code and ethics, as real world ethical judgements aren't necessarily consistent or well defined in a way which is easily translatable, even by a large ML model. Jigsaw is a great example of this -- right across the aisle (pre-pandemic) from Perspective is a team fighting internet censorship. Obviously Perspective's "censorship" is different in quality from the Great Firewall, but it shows the hairiness of the problems.

All this is to say the people at Jigsaw are some of the most brilliant people I've ever met and I'm glad they're out there working on difficult problems.

View on HN

> I think this article is pretty fair to Perspective given that it was never meant to be used in a fully automated way, just as a first pass to help forum moderators.

Given how eager Google itself is to automate all your interaction with them (and how hard it is to get access to anything resembling actual human judgment), did they really think it wouldn't be used that way?

Or did they just not care?

Also, I would have hoped the project's close ties to the US State Department would have worried some of you brilliant people a little more.

The close ties to the state department made me uncomfortable while I was there, but I never saw anything that looked like they had any say in anything we did. But I never worked on Perspective.

Yeah, we love what Jigsaw's building! This is all with the hope of improvement and collaboration.

We deal with these hairy problems a lot too. Even before ML proper, getting the definitions right is very tricky. Should a comment that's polite and positive on its own, but supportive of a toxic parent post ("I love Nazis!" -> "I agree!"), be treated as toxic? Is "toxic counterspeech" equally toxic? What about a comedian making fun of an actor's nose, and does it depend on whether the joke is to that actor directly vs. merely referencing them in the third person? etc.

For these reasons, I actually personally like the fact that the Jigsaw annotation guidelines are very high level (as opposed to long and prescriptive) -- it lets the data capture the spectrum of "human preferences" on its own (at least, it does when you can trust that the annotators are able and trying to do a good job).

Reply on news.ycombinator.com