Amazon forced to scrap machine learning recruitment

However, by 2015, the company realised the tool was consistently ranking resumes that featured the word ‘women’ poorly, compared to those that didn’t.

Any resume that mentioned involvement in women’s clubs or sports teams, for example, were downgraded. According to Reuters, it also downgraded graduates of two women’s colleges.

Any artificial intelligence or machine learning technology is only as smart as the information it’s learning from. Amazon’s tool was designed to analyse hiring patterns over a ten-year period — a period when the vast majority of hires, especially engineers and data scientists, were male.

Attached: amazonlogo.png (500x500, 14K)

Other urls found in this thread:

smartcompany.com.au/startupsmart/news-analysis/unconscious-bias-amazon-forced-scrap-machine-learning-recruitment-tool-didnt-like-women/
youtube.com/watch?v=TlO2gcs1YvM
twitter.com/NSFWRedditImage

smartcompany.com.au/startupsmart/news-analysis/unconscious-bias-amazon-forced-scrap-machine-learning-recruitment-tool-didnt-like-women/

HAHA REALLY MAKES THE OLD BRAIN CELLS ITCH HUH

LIBTARDS BTFO

Attached: 1507898839378.jpg (896x960, 125K)

Like Tay, and this, demonstrates: reality has a right-wing bias.

I'm confused as to how the machine learning worked. Did they assume that they already had the best hiring practices, and trained a machine to hire people with similar resumes to everyone already hired? Or did they feed it information with correlations between resumes and performance, and then have it learn that way?

MFW skynet is on our side

Attached: 1474345301177.png (600x800, 347K)

Obviously there was a patriarchal conspiracy to feed it bad data that indicates women aren't as competent

Makes sense. There's also no reason to hire white applicants given enough Jewish and Asian candidates.
No wonder the wealthiest states are so open to immigration.

Probably better off moving to China if you want to actually be able to use A.I. without triggering wahmen and bugmen.

Shouts out to my dude with the only non retarded response.

Attached: enhanced-15865-1458694668-9.png (518x431, 562K)

And yeah, it was trained to analyze hiring patterns, so Jow Forums circlejerkers can btfo

>Amazon’s tool was designed to analyse hiring patterns over a ten-year period — a period when the vast majority of hires, especially engineers and data scientists, were male.
The explanation is all right here in this last part. Yes, "reality" does have a right-wing bias, because for the past ten years, the political right has owned "reality."

Simple Entropy. It's easy to break things down. Harder to make them.
An AI will always try to cause destruction because it's easier than being nice.
Being nice requires effort. Being an asshole is easy.

Maybe the left should've been busier cultivating technical skills instead of trying to get men that dress like women into women's bathrooms.

>Being nice requires effort. Being an asshole is easy.
Imagine being this much of an NPC you said something like this unironically.

lmoa b&r

Niggers aren't nice. Racists aren't nice.
It's simple really.
Kicking a sandcastle is easier than making a sandcastle.
To be nice you have to say a sentence and all. To be an asshole you have to say nigger and that's it.

You say that as if the alleged lack of technical skills on the left is the reason they were never in power. As opposed to the real reason: corporate lobbying and underground deals.

>it's harder to be a pushover than to speak your mind

Spoken like a true NPC

>You say that as if the alleged lack of technical skills on the left is the reason they were never in power
... are you familiar with the 1900's through 1940's? You know, the era of overt leftism and when we lost centuries worth of exponential growth due to leftism.

You mean the Era that made our greatest tech boom possible exist though proper funding plans? Yeah, was a good time. Hard times breed strong men. Then when things got easy the conservative boomers where born and squandered all the gifts left by the greatest gen in recent history.

Yeah being a creator is harder than being a destroyer. The same way your posts are shorter than mine.

Low hanging fruit is always easy to pick. That's why it's hanging so low.

this tbqh but modern leftists are not better then boomers fuck both I say

Being a creator vs destroyer is different than being nice vs asshole, don't move the goalpost user. And yes, being able to say more in less words is something any competent creator should strive for.

Okay. If you wanna convince me to join your fascist blood cult you should do better than that kid. Because you've done nothing but insult me and that's not very mature. Unless you're shilling that reality is immature and lacks restraint.

But that "greatest tech boom", as you call it, was built upon the industries and products that capitalists like Rockefeller, Carnegie, Vanderbilt, Ford, Westinghouse, Edison, Morgan built.

Wow, user, you're not being very nice, why don't you try to be a little nicer, huh?

Not him but that was unavoidable. In modern society, only capitalists can build anything. It's not because they're better or their philosophy is correct, it's because they own everything. And the fact that they own everything is also not because they're better or their philosophy is correct, it's because they gang up to cheat the rest of us on a regular basis.

>AI gives no fucks about diversity/is objective by design
>analyzes male and female applicants
>determines female applicants are, on average, detrimental to productivity
>libfags and feminists accuse AI of being sexist
Reminds me of Google and the gorilla incident.

I'm trying to get my points across

Attached: yellow girl.jpg (1196x2125, 746K)

Humans aren't robots. You can't focus on only the productive element. Women are important to keep the workplace running smoothly. Like engine oil.

So you're telling me that sometimes to get a point across, you have to be "not nice"?
Imagine that.

Attached: 1538063282836.png (218x241, 5K)

>it's because they own everything
Is this a joke?
How did Apple start up if HP, Intel, IBM owned everything back then?

I wasn't mean to you though. You were just being juvenile and I told you that you were.

Attached: 1544937269985.jpg (640x904, 115K)

You're referring to the military Industrial complex. I'm referring to companies like Motorola, and HP being founded by a couple dudes and an idea along with the tech giants that grew from late 19th & early 20th century companies like IBM.

>Women are important
They weren't 60 years ago.

Wtf are you rambling on about?

Attached: women.jpg (1200x673, 80K)

The same way HP, Intel, IBM started up: they exploited unethical business practices that were not yet illegal, and they couldn't be punished for it because it's unconstitutional to punish crimes that were not crimes at the time they were committed.

Ok. You don't know what you are talking about. Thanks for trying though.

>turning away from the truth because it contradicts your ideals
lol

They still aren't.
Women simply don't care about technology. No amount of outreach is going to change the zeitgeist and make women think that technology is interesting.

All of these bluehairs and other assorted dangerhairs are trying to work against the greatest cultural force known to man (the trend for women to do women-things and for men to do men-things) and they're going to drown in the million years of built up history.

Greatest tech boom was post Ww2 50s through today. Hasn't ended yet. The only fast development that happened at the turn of the century was poo tier countries catching up to the west overnight due to gibs me dat in exchange for bodys to throw at war.

but what about women who care about technology though

Last year I worked on an NLP auto CV analyzer for a company.

Ours wasn't targeting the CVs of the candidates that got hired, but the CVs of the best employees given by their performance reviews. The target company was one of the biggest of the country so we had all sort of employee backgrounds.

The three most influential traits (from most important to least):

>Where do applicants live (out of the gate, people that live on poor sectors make for poor employees, and that matters more than education).

>Name (poor sounding names make worst employees).

>Name (Gender) (women make worst employees in general according to our program).

We really only had one name parameter, but on tests we observed that women names where doing worst than male names and that names associated with poorer people where also doing worse than normal names. However while there was a bias for males over woman it wasn't that big of deal. In general males and females are mostly equality incompetent.

Our program worked for CVs of non-technical people, it was mostly for office work. We made a smaller demo for technical applicants. On it the most important quality was education (names of the schools do matter), estimated age and experience followed by the previous once on the same order, but in this case the bias towards males (while it existed) it was even slimmer. Not really enough to really be concerned or to discard women applicants just because they were women.

no one's complaining about them
but I do complain about people in general who get in to just get the job done and go home since they tend to not know anything

>Math concludes you should hire less women
>Math is sexist and wrong
>Proceed to deny reality and just scrap math

I don't argue opinion, and you are obviously arguing opinion. So, again, you don't know what you are talking about. Thanks for trying.

>>Math is sexist and wrong
It is when the parameters it's given are sexist and wrong.
>inb4 but those parameters are reality
Sometimes reality is sexist and wrong. That's not an inherent property of reality through, it's what happens when *people* who are sexist and wrong *own* reality.

>pointing out the undeniable crimes of the megacorps is """"obviously arguing opinion""""
top zoz

Attached: oh boh muhmroh .png (366x275, 103K)

>Posts picture of an underdressed, underaged looking, sexualized cartoon.

Yeah, you're right, I'm the juvenile one.

>Women simply don't care about technology.
Obviously false. There might be a measurable disparity in their representation in the field. Unsurprising given their disparate representation in all sectors of all contemporary societies.

>dangerhairs are trying to work against the greatest cultural force known to man (the trend for women to do women-things and for men to do men-things) and they're going to drown in the million years of built up history.
True. More power to em.

In all seriousness this is a widespread issue. A few years ago this book came out called "Weapons of Math Destruction" which was widely circulated in the industry. Which argues that basically every use of machine learning needs to be carefully controlled or outright banned.

There has been a sucessful effort to ban most uses of machine learning in the EU. At least in any use that might in any way judge a human being. It's questionable whether even spam filters are legal anymore, though it's unlikely to be enforced that autistically.

In the US they are trying to get rid of it as well. And existing law already makes it very difficult to use it without getting sued. E.g. a fire department in the US has been sued multiple times for millions of dollars because they used a "merit based test" that just so happened that white people did better at.

There is a billion dollar industry in the US of suing any company that doesn't have perfect race or gender ratios (which of course no one does.)The tech industry gets away with it by paying tons of money to diversity industry groups to effectively bribe them. If you get sued in court for racism, the only defense corporations have is too prove how not racist they. By showing all their diversity programs and how much money they donated to diversity industry causes. They have to hire all these "diversity officers" which lead to many of the problems this board complains about.

But if they didn't do it they'd be sued out of existence. Google has like 1% black employees - and they are all like black Indian or "Barrack black". So they try to focus on feminism as much as they can, because they do much better there.

Attached: Cathy O'Neil Compressed.jpg (2880x1620, 471K)

>There is a billion dollar industry in the US of suing any company that doesn't have perfect race or gender ratios
oy vey

You didn't point to any crimes.
You wrote explicitly your opinion about what you consider crimes.
Post facts or fuck off.

>outsourcing

>If you have a PFT and a white guy wins, the test must be adjusted until he loses or you get sued.
Facts of life.

A large part of this book and many of the propaganda against machine learning is based on a single propublica study. The study found that a prison was using machine learning in a way that discriminated against black inmates.

Except it turned out to be entirely false. Their study was "not statistically significant" in their very own words. And they published it anyway. It was then widely circulated in the media and the tech industry about how horrible algorithms are.

The state had massively reduced the prison population by releasing inmates they predicted were not likely to commit crimes again. And it was a hugely sucessful program that saved millions of dollars and saved prisoners of nonviolent crimes. The algorithms did significantly better at predicting than human judges, probation officiers, or even trained psychologists. And it wasn't any fancy deep learning stuff, just some basic statistics.

Attached: 2597982180998295809.jpg (500x455, 48K)

>Except it turned out to be entirely false
You say that like it matters.
Typical white guy thinking.

Another example found that "word vectors" were "gender biased". And algorithms were predicting that "he" was more likely to appear in the same sentences as "doctor", and "she" more likely to appear in the context of "nurse".

So there's now an entire field of research on how to stop this terrible crime. That people are being paid lots of money to do. And you actually have to "degender" your word vectors before you use them in any commercial application. Which generally makes them worse, of course. It's like doing brain damage to a poor little neural net.

Someone found that the word vectors were actually highly predictive of the gender ratio of the actual professions. E.g. you could predict how many men were in a profession by the gender of the word vector. It was just random or just stereotypes.

And this isn't meaningless because a lot of languages are heavily gendered. Or a machine using the wrong gender word when translating a piece of text can be embarrassing. And degendering the algorithm really can really damage the accuracy of these things.

Attached: word vector bias predicts occupation gender ratio.png (399x333, 27K)

>typing on a calculator is hard

Typical nigger still crying about muh cotton fields.

youtube.com/watch?v=TlO2gcs1YvM

Saw this the other day, then saw a drone swarm exhibition on the BBC weekly tech show. Shit is already pretty much possible

>being unable to detect a shitpost

It takes one to know one.

>The Jow Forumstard literally believes an episode from a Netflix series is real
No wonder you're all Republicans. Gullible retard

The left literally owns everything asswipe. You ARE in power. You're the people whom are being rebelled against.

You're just ignoring the foundations of modern programmable computers like the Z3, along with jet and rocket technology then? Nice.

Worry not, eventually the AI will be told to just add +100 to all women scores, problem solved!

>Name (poor sounding names make worst employees).
What the hell defines a poor name? Is Christopher a more wealthy name than Thomas or Edward?

i got you homie

Attached: skynetWithTerminator.jpg (890x501, 160K)

My assumption would be that they just trained it for initial screening. A lot of engineers waste time on initial phone interviews with dead ends. I'd say more than half is pointless. In that case if you can reduce the number of pointless phonescreens by even 10% you are saving thousands upon thousands of hours of engineer time.

My assumption would be they just collected data on which phonsecreens led to nowhere and tried matching that with CVs in their machine learning setup.

Source: Used to work for amazon, used to do interviews.

imagine actually believing this

My post was neither a nigger nor a shitpost.

It was a jew post.

You act like the fact that pro publica studies are false MATTERS. It doesn't. The purpose behind these studies has absolutely nothing to do with the truth. It has to do with providing intellectual justification for Jews doing what they WANT: promoting blacks and hamstringing whites.

The study is wrong? Who the fuck cares. You're "racist." Fuck you.

We'll make a new study, white guy. Same as the old one.

I literally can't comprehend how someone can be deluded enough to continue thinking that "conservatives" (euphemism for white men) are in control of the West
>implying that white people controlling white nations could even possibly be controversial
>implying if "conservatives" lose any more ground then white people won't soon be exterminated as in South Africa

>he doesnt recognize the eternal trap
your newfaggotry is showing. to Jow Forums

Pablo probably didn't get as good an education as Christopher
Maybe Juan wasn't properly nurtured when he was 3 years old, so his brain didn't develop properly
Maybe Deshaun has a shopping list sized criminal record

if you're responsible for hiring new employees at a tech companydo you thing the one with the name la-ah, leroy and laquitalaquanta are going to be as good as eckbert?

Should have added a global script in the AI that always made it where men and women would always be equal no matter what.

Imagine being such a dumb nigger that drawing parallels between hiring patterns that require as much "effort" (which doesn't exist as far as AI is concerned) no matter how theyre based, and kicking sandcastles is a valid comparison

ill
cringe

>necessary
>important

You're a fucking joke, most tech companies started from nothing

>representation
Who pays you to talk like a fag?

>he thinks attending Harvard is cheap

Imagine denying the "liberal world order" (NWO) exists

Back to the CIA HQ for you invaderfag

>MUH HURRRVARD
Bill Gates dropped out and used a $50,000 grant from his parents and a nerd friend to start
Steve Jobs dropped out and sold his hippy van
MUH HARVARD
MUH EDUCASHAN

You liberals are a fucking joke get out of my face

Yeah, and it's run by freaking corporations (i.e., capitalists). It's the CAPITALIST world order, you idiot.

Dumbass, you just admitted they didn't start from nothing.
Also, you forgot Zuckerberg.

If you parents make less than 100k a year you go to Harvard for completely free, regardless of race or scholarships.

If you're white, you don't go to Harvard anymore.

>Almost every big project engineering were men

>he thinks Gates and Zuckerberg's parents made under 100k a year

Yeah sure but without justice we become like china. No soul.

Attached: 1546307527.png (1075x661, 34K)

>0 capitalist nations in the world
>every major company supports socialist politicians
really opens up the noodle

>Amazon’s tool was designed to analyse hiring patterns over a ten-year period
The training data was their own hiring data. They basically outed themselves as biased

Are you aware that jews are not white?
Are you aware of how ridiculous it is that ANY non-whites should be allowed access to white prestige-granting institutions?

The real percentage of non-jewish white people at Harvard is less than 10%.

And since shit skins and jews will devour the prestige of this institution, it is up to white men to create the new prestigious institution.

Fuck you sir.