3 ms·

> This tells me no humans are in the loop on any of this. At that scale, it's very hard to achieve. 500M tweet a day. Let's suppose 1% of the tweets are flagge

> This tells me no humans are in the loop on any of this. At that scale, it's very hard to achieve. 500M tweet a day. Let's suppose 1% of the tweets are flagged by users or by keyword, that would be 5M of tweets a day to review. Let suppose that one person can review 1000 tweets a day, that would be 5000 people working full time. And even with humans, there'll always be some flagging that will be contested. Besides, different countries have different legislation regarding what can be be published on the web. It's a hard problem to solve, even with the best intent. I don't think Musk is going to fix this. I think it'll be more about reinstating people like Trump on the platform.

pnt12 4y ago
Where I'm from, minimum wage is around 15k per year. So hiring 5k people would cost 75M in salaries, and let's just double it for other operation costs, 150M. People don't work 24/7, so we must quadruple it for triple shifts + weekends. That leaves us at around 600M: yes, that's terribly expensive. 1 But maybe we don't need to review all the reports! What if we rank them? Correct reports grand a user positive 'report karma', and they climb higher. Posts that are 'trending' in reports should also climb higher. New posts take priority over old posts, ie let's police new content, not everything that has been ever written. And of course, we can also use modern techniques, ie AI to score the reports. If Twitter did that and could review the top 10% of reports, at 60M per year, would that be worth it? If successful, I think so. 1: this number is biased because it assumes all reports are independent, but I suspect a significant amount of reports would concern a low number of posts, eg controversial or political posts from accounts with a large follow count.

3 ms·

> This tells me no humans are in the loop on any of this. At that scale, it's very hard to achieve. 500M tweet a day. Let's suppose 1% of the tweets are flagge

by yodsanklai 4y ago

pnt12 4y ago
Where I'm from, minimum wage is around 15k per year. So hiring 5k people would cost 75M in salaries, and let's just double it for other operation costs, 150M. People don't work 24/7, so we must quadruple it for triple shifts + weekends. That leaves us at around 600M: yes, that's terribly expensive. 1 But maybe we don't need to review all the reports! What if we rank them? Correct reports grand a user positive 'report karma', and they climb higher. Posts that are 'trending' in reports should also climb higher. New posts take priority over old posts, ie let's police new content, not everything that has been ever written. And of course, we can also use modern techniques, ie AI to score the reports. If Twitter did that and could review the top 10% of reports, at 60M per year, would that be worth it? If successful, I think so. 1: this number is biased because it assumes all reports are independent, but I suspect a significant amount of reports would concern a low number of posts, eg controversial or political posts from accounts with a large follow count.