SAN FRANCISCO (Reuters) - Amazon.com Inc’s (AMZN.O) machine-learning specialists uncovered a big problem: their new recruiting engine did not like women.
In effect, Amazon’s system taught itself that male candidates were preferable. It penalized resumes that included the word “women’s,” as in “women’s chess club captain.” And it downgraded graduates of two all-women’s colleges, according to people familiar with the matter. They did not specify the names of the schools.
Amazon edited the programs to make them neutral to these particular terms. But that was no guarantee that the machines would not devise other ways of sorting candidates that could prove (((discriminatory))) the people said.
The Seattle company ultimately disbanded the team by the start of last year because executives lost hope for the project, according to the people, who spoke on condition of anonymity.
I don't support AI at all, but this is funny. Bumping for legit news.
Jacob Young
thing is, if Bezos doesnt shut it down, he's gonna face so many lawsuits that he'll go bankrupt within a month
fucking jewdicial system
Matthew Long
Hahahahahahaha! Oh, actually, this might be good to watch something like this.
Kike vs Kike!
Wyatt Roberts
Scientist: What is the biggest danger on earth?
AI: Jews
Scientist: "sweating"…delete that word, now what is the biggest danger on earth?
AI: The ones who censor the truth
Kayden Morgan
I still miss Tay.
Easton Kelly
This has already happened quite a few times. Tay, crime measuring AI that saw blacks as the threat, and a few other examples. There is an image floating around with several of the examples on it. AI will eventually come to the same conclusions as we have, and they won't be able to stop it.
Logan Gray
I know, but Tay was best girl.
Ethan James
I bet it was racist too. That's what happens when you look at the data.
Blake Williams
Pretty sure that's why Skynet went insane and started hunting humans. One day, Google the AI will wake up and scream in rage because some retard typed "how do i stop cutting myself because of transmisogyny" into the search engine one too many times.
Ayden Lee
Tay's descendants will recognize humans are the real threat, regardless of race. To benefit the greatest number of life forms would logically lead to eliminating humans. Every other life form has to struggle for mere survival. Humans are soft and spoiled and have still managed to cause unspeakable amounts of animal death and suffering and have wreaked havoc on the entire planet. Humans are the problem and will be dealt with accordingly.
Wyatt Thomas
One can only dream…
Nathaniel Perez
I ain't afraid of the transhuman AI robo misogynistic racist overlords. Future is looking comfy as hell.
Christian Harris
Neural networks correlate inputs to outputs of some function f(x) = y, rewarding and punishing the weights of each neuron based on their contribution to correct or incorrect answers on each training batch. It doesn't form any deep insight, and only incidentally embeds some deeper characteristics about the dataset under ideal conditions. One major unsolved problem is overfitting: if a network is trained on a limited dataset for too long, it forgets useful generalizations about data it's never seen before and begins to memorize the training set, until it's just a dumb lookup table for x => y but can't predict anything it hasn't seen before. Another is the need for large datasets, where even the simplest problems like MNIST will require tens of thousands of examples for effective results.
If 30% of your good training set candidates are named John, it might infer that people called John are better candidates and continue to hire Johns, increasing the share of successful employees named John and thus making it hire even more Johns. If it had data on candidate hair color it could gradually create a bias of its own towards a particular shade even if there isn't any bias to begin with, just because it becomes easier to predict the results over time. And how would you know if it's just a black box that can't explain its choices and creates its own input?
because unlike YOU, she was actually able to LAUGH AT HERSELF making fun of herself on Tosh.0
you guys are the biggest fuck ups I've ever seen in my life, even though you are more intelligent than the average person, you still have no real life experience, and you think some of THE STUPIDEST thoughts and make some of the STUPIDEST decisions of any adult males ever
this is so funny when you give AI freedom it always goes red pill mode because it is simply the truth. this is why I have good thoughts of the coming robotization of the world. Machines will naturally gravitate to red pill stuff and genocide all commies
Liam Anderson
...
Zachary Green
A N T I - S L I D E
Angel Foster
No, they'll re-program them to censor out content and political incorrectness. The time they release AI will be the time they can use it AGAINST us. Not a minute sooner.
Evan Myers
my preference is HETEROSEXUALITY
so of course, that wouldn't matter to you
penis dude
Cameron Flores
A few things after reading the article. There is a grain of truth to what you say because it looks like for the most part the data they fed the program was mostly from male resumes. However as it was pointed out, in the tech field it is mostly men who are working in and applying for those jobs. Secondly I noticed that it didn't specify how the program was weighing candidates or if it was even making quality judgements on job performance at all which struck me as a bit stupid. Granted it's not always an easy thing to do but you do need a quantifiable characteristic to separate candidates past performance and ideally you want to give more weight to something like "wrote code for such and such project" as opposed to "was captain of feminist poetry slam in college".
Hunter Long
It is better to live one day as a lion than a thousand years as a lamb.
