Google’s Photo App Still Can’t Find Gorillas. And Neither Can Apple’s: Eight years after a controversy over Black people being mislabeled as gorillas by image analysis software — and despite big advances in computer vision — tech giants still fear repeating the mistake.

submitted by neolib from (nytimes.com)

sorted by:
top
newinsightfulfun
you are viewing a single comment's thread.

view the rest of the comments →

[–]Alienhunter 2 insightful - 2 fun2 insightful - 1 fun3 insightful - 2 fun -  (2 children)

Lol. Talk about hard coding AI to fail.

Thing is there's definitely truth to the idea that AI algorithms reflect the bias of people who make them. It's mostly based on sample training data rather than this knee jerk racism reaction though. Not limited to image recognition.

Google maps is a good example, obviously developed by Americans and largely tested in America. Try using it in a foreign country and it can be really garbage at the directions it gives. Is it racist ? Hardly, google fed it's map data into it based on US standards and it's not their fault that road standards and rules and the like change from country to country. The map data for a one lane road with 1.5 meter clearance is indeed the same as an American road with 2 lanes and 5 meter clearance. The really stupid people are the users who prefer to follow google directions over road signs or common fucking sense.

[–]Dragonerne 3 insightful - 1 fun3 insightful - 0 fun4 insightful - 1 fun -  (1 child)

as someone who develops AI, if you think racism in AI is only an outcome of "training data" then you are mistaken.

These antiwhites artificially forces the AI to be antiwhite in its bias by changing how it values different things. They actively boost content that is associated with nonwhites positively and deboost negative content about nonwhites in the training.

The AI will still be accurate, it will just be extremely biased against whites, and outright lying in some cases.

[–]Alienhunter 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (0 children)

They're interfering with constantly when it pops out responses that they don't like. It's all gonna be a massive mess.

Problem is people who outright trust AI like it's some infallible magic solution to shit. Have had to deal with this before in really unimportant situations, GPS says turn left when the signs and the rest all say turn right, but google must be right eh, nah.

You're gonna get a bunch of people folloeinf the rule book unwilling to risk their own necks because the AI said something obviously false. It's gonna be a huge shit storm.

If the stuff like face recognition even works. Seems like stuff like face paint can mess with it, and I don't even want to know how many false positives or negatives it gives when it rolls out on a large scale.

I went to an amusement park with my friend a couple years back that had rolled out face recognition tech in lieu of tickets, my friend gave this dumbass goofy smile at the initial registration, not realizing he'd had to give the same dumbass goofy smile every single time from then on. It was fucking hilarious for something so unimportant, but it's gonna be useless for anything important.