I was searching for articles on Google Translate with SearX (serx.cf) and then, when I open the Mastodon Federated timeline (mastodon.la), guess what, a post about Google Translate! Talking about how humans should translate and not machines, the exact same topic. Can’t believe I haven’t escaped this bullshit with Mastodon. Live and learn.
But this is survivorship bias. You remember it because it seems suspicious.
Think of how many things you search per day before checking mastodon and how many toots you see when you check mastodon. If you search 10 topics a day. Now you were searching for a tech related topic and many mastodon posts are about technology (it seems to be the most common niche). Let’s say that 1 in 10k toots are likely about Google Translate as a fairly popular tech topic. Now if you see 20 toots every time you open your timeline and you check your timeline 5 times a day you have about 10*20/10000 chance of seeing a “coincidental” toot on a given day. that is a 2% chance. If you have used mastodon for a month that is a 55% chance of this occurring.
These are very rough numbers but you can see how it works out. Maybe you consider the last 5 days of search suspicious so now your daily chance is 5x higher. Or maybe only 1 in 100k federated toots are about Google Translate so it is 10x lower. But the point is that while it feels very unlikely that this specific instance of this problem was incredibly unlikely, the truth is that it is easy to forget the many times where this didn’t occur and the multiple ways that it could have occurred.
If you start to see this much more frequently then that then you may want to start getting suspicious, but it is also important to consider the numbers.
Google translate working isn’t exactly Firefox working with Meta. That is, it isn’t anywhere near a hot topic. Your 2% chance is bullshit, this is much more unlikely.
There is no persuading someone who entered the discourse convinced he was right and who was unwilling to accept any other answer.
I thought I left this behind with Twitter, but hey!, here we are.
Yup. I figured I would try once with some rough framework for estimating how likely or unlikely it really is but at this point it definitely isn’t worth trying to help someone who doesn’t want to be helped.
Some people just want to feel so special that they’re the only ones who know THE TRUTH and that THEY really are out to get them.
I’m convinced because this is obviously unlikely. Other answers given: scrape a highly complex code for seeing how the federated timeline works, and shitty basic mathematics that never apply to real world probabilities. I could do without those. What would be nice: someone involved in the project pointing out the error in my thinking through the code, or actual mathematics.
That’s pretty much exactly what I said. Just with more self-congratulating bullshit attached.
So, the fedi tracker in Mastodon says there are 3 million mastodon accounts. Lets say there are 5000 posts each minute. That is, 1 in 600 users are posting every minute. The machine is picking 3 out of those 5000 posts to show to me. That is somewhere near 0,05% chance of me seeing a post that happens each minute. I perfectly trust someone in 3000000 people is thinking about the same topic as me, even though that is a weird coincidence that’s worth thinking about. What I don’t believe is the machine picking exactly that post to show me in the federated timeline without manipulation (adjustments to tracking).