No. The berets Ragstock carried back in the 80’s and 90’s were from Germany, the liner tags made that clear. Those were the only stores you could get such a thing back then in the Twin Cities. Not even Sun’s Rock-n-Roll Items carried them.
3 Likes
I’m seeing too many articles like these:
In too many instant arrests, no zoom in is even attempted. For some groups, LEOs don’t seem to care about accurate identification, since casting a wide net and sorting out the details later can still increase their closed case count. I’m not even gonna rehash letting the insurrectionists go home, while BLM protesters were rounded up very quickly and large numbers faced harsh penalties.
5 Likes
aside: I’d just like to mention that I love it when a song is good and also sounds good done as a cover in a different style.
(REM - Stipe + Zevon)
/aside
4 Likes
That tracks.
Many years ago I was made aware of the “false positive” problem [1] that applies to things like facial recognition software: if the particular people you’re looking for are rarer than the software’s false positive rate, then you’ve got a less-than-even chance of having the right person when the software says “identified”.
Example: in a city where the cameras scan 50,000 people a day, you’re scanning for 100 suspects in particular, who you have sound reasons to believe are in the crowd somewhere.
Facial recognition software often brags about “99.5% accuracy” [2]. Let’s take that at its word (there’s no reason to, but for this example we will).
50,000 people scanned, 0.5% error rate = 250 mistakes.
100 suspects. Let’s round 99.5% up, so all of them are detected.
Given that detection has occurred, there is a (100 / (100 + 250)) = about 29% chance you’ve got the right person.
99.5% might be too good to be true, btw. And 50,000 is a small number for this kind of work. So it’s probably worse than that.
The points @PsiPhiGrrrl has raised illustrate this: whether LEOs say “let’s be careful not to persecute innocent people” or “round 'em up and let the courts sort it out” depends on whether the suspect’s skin is lighter or darker than a hamburger bun.
ETA:
[1] The “false positive problem” was first explained to me by my probability prof, whose distaste for authoritarians was personal. In 1940 when he was a baby, his family fled from France to Romania. As a “numbers professional”, he absolutely hated people using numbers like “99.5% accurate” and not mentioning what this means in practice when you’re looking for a rare thing. And as someone whose family fled a persecution nightmare, he hated the agendas behind that number-juggling.
[2] The quoted accuracy rates vary from 99.95% under “ideal conditions” - identical lighting and background to the reference image - to less than 85% when you complicate the problem with e.g. different background, uneven lighting, partially obscured by umbrellas, hats, other faces, etc.
7 Likes
Yup. Pretty easy with anything written by Prince, though - you’d have to go out of your way and put in some real effort to make a bad version of it.
2 Likes
This topic was automatically closed after 5 days. New replies are no longer allowed.
