Culture Google Photos recognize black couple as gorillas, google fixes the problem but in a way no one likes

The incident actually happened in 2015:

Screenshot_2018-01-16-10-35-57-1.png


With google apologizing and saying that they would fix their algorithm then.

But now new information has come to light that google simply binned the gorilla tag instead of trying to properly fix it.

Binned as in nothing will tag as a gorilla.

Articles about it are popping up everywhere.

Screenshot_2018-01-16-10-47-10-1.png Screenshot_2018-01-16-10-47-02-1.png Screenshot_2018-01-16-10-46-53-1.png Screenshot_2018-01-16-10-46-48-2.png Screenshot_2018-01-16-10-46-48-1.png Screenshot_2018-01-16-10-46-42-1.png
 
Remember when Apple and Samsung et all found issues with their phones and webcam auto-focuses because most of the testing had been done by Foxconn employees and such? So they didn't just didn't recognize black people due to the dark pigment of the skin and assumed it was a lighting issue?

Remember how it turned out to be like, a 10 second software fix and a making sure prototype tests took place stateside again to check it'd work for people?

Makes you wonder why a fucking amerifat corp can't do something as easily bloody simple.
 
Remember when Apple and Samsung et all found issues with their phones and webcam auto-focuses because most of the testing had been done by Foxconn employees and such? So they didn't just didn't recognize black people due to the dark pigment of the skin and assumed it was a lighting issue?

Remember how it turned out to be like, a 10 second software fix and a making sure prototype tests took place stateside again to check it'd work for people?

Makes you wonder why a fucking amerifat corp can't do something as easily bloody simple.
Because Google became so hyper focused on diversity they forgot that employees need to have some kind of ability.
 
And what did you expect? An Algorithm is just that: a mathematical resolution based on several data inputs. You have two animals: One is human, the other one an ape. Both are the same color, similar shape and size.. Both belong to the same race of hominids and one is a descendant of the other. Do you really think a machine will know the difference based on your social sensitivities?
 
:c tfw when even google algorithms think of blacks as apes :heart-empty:

And they wonder why Tay ended up a Nazi within hours.

Of course they can't fix it. No one who knows how to code actually works at Google anymore as the Demore lawsuit showed.

And on top of that, most of them are tumblrinas. If you've ever looked at tumblr art, you might have noticed they always draw black people as gorillas.
 
Let's ignore the fact this is a lazy fix.

What does Google going to do when its algorithm finds the creature we call "gorilla" and not having a word for it chooses to associate it with black people?

Congratulations you just put in way more thought than the analysts who made this fix and they didn't even have to pay you 200k a year.
 
Of course they can't fix it. No one who knows how to code actually works at Google anymore as the Demore lawsuit showed.
I know you're just meming and while I do think Damore's complaints have some merit, Google employees really do not know how to fix this shit properly. And no group of engineers would know it either.

The recognition algorithm uses neural networks and its results are multiple huge tables of numbers that mean nothing to humans. Fixing those would be like asking a neurosurgeon to fix little Jimmy's brain to stop calling his classmate Susie a cunt.

This is going to be more and more of a problem as soon as more tasks will be performed by machine learning projects. Imagine that self-driving cars have a tendency to ram into any group of exactly four children. The only way to fix that in a timely manner is to disable self-driving capabilities in all cars entirely and then retrain the neural network, while the engineers cross their fingers that maybe, just maybe, few days of a server farm working at its full capacity will teach the cars to not kill children.
 
Back