Die von der britischen Polizei eingesetzte Gesichtserkennungstechnologie ist voreingenommen, räumt das Innenministerium ein
https://www.ft.com/content/c49a4524-e4bd-42f1-aabb-d7b0e50dd066
Von 457655676
Die von der britischen Polizei eingesetzte Gesichtserkennungstechnologie ist voreingenommen, räumt das Innenministerium ein
https://www.ft.com/content/c49a4524-e4bd-42f1-aabb-d7b0e50dd066
Von 457655676
6 Kommentare
Paywall free link [http://archive.today/ExU4b](http://archive.today/ExU4b)
While I’m sure there are technical reasons as to why the system is „Biased“ I pulled this out from the article:
>
Testing at the National Physical Laboratory found the software correctly identified Asian suspects 98 per cent of the time, falling to 91 per cent for white suspects and 87 per cent for Black suspects.
So out of 100 people it picks out it gets 13 wrong for Black suspects and 9 white suspects. Key word suspects. I’m sure after getting stopped by the police and identifying yourself that’d realise it wasn’t you and let you go. while inconvenient if its catching 87 or 91 actual criminals its got to be worth it.
I’m totally shocked at this.
Totally.
It’s also yet another fucking reason why the drive to AI and surveillance is fucking bad – these systems are **inherently** biased by the data they are trained on, and they aren’t good at picking up new and emerging trends.
Nothing like changing the narrative from a privacy issue to a race issue
AI is only as biased as the person feeding it information.
This is important (and unsurprising, if it’s been trained on UK data, when the vast majority of people are white). Police need to understand that a facial recognition match on its own isn’t an identification, and they should watch that person and see if there is other information to match the identification rather than doing a stop solely on that basis. They should do that anyway tbh, and from the article it sounds like that is already the case.