How I'm fighting bias in algorithms | Joy Buolamwini

Prevodilac: Ivana Krivokuća Lektor: Tijana Mihajlović

Hello, I'm Joy, a poet of code,

Zdravo, ja sam Džoj, pesnikinja kodova,

na misiji da zaustavim neviđenu silu u usponu,

on a mission to stop an unseen force that's rising,

a force that I called "the coded gaze,"

silu koju nazivam „kodirani pogled“,

my term for algorithmic bias.

što je moj termin za algoritamsku pristrasnost.

Algorithmic bias, like human bias, results in unfairness.

Algoritamska pristrasnost, kao i ljudska, ima za posledicu nepravednost.

However, algorithms, like viruses, can spread bias on a massive scale

Međutim, algoritmi, poput virusa, mogu raširiti pristrasnost u ogromnoj meri

at a rapid pace.

velikom brzinom.

Algorithmic bias can also lead to exclusionary experiences

Algoritamska pristrasnost može dovesti i do izloženosti isključivanju

and discriminatory practices.

i prakse diskriminacije.

Let me show you what I mean.

Dozvolite da vam pokažem šta hoću da kažem.

(Video) Joy Buolamwini: Hi, camera. I've got a face.

(Video) Džoj Buolamvini: Zdravo, kamero. Imam lice.

Can you see my face?

Možeš li da vidiš moje lice?

No-glasses face?

Lice bez naočara?

You can see her face.

Možeš videti njeno lice.

A moje?

What about my face?

I've got a mask. Can you see my mask?

Imam masku. Možeš li da vidiš moju masku?

Joy Buolamwini: So how did this happen?

Džoj Buolamvini: Pa, kako se ovo dogodilo?

Why am I sitting in front of a computer

Zašto sedim ispred kompjutera

in a white mask,

sa belom maskom,

trying to be detected by a cheap webcam?

pokušavajući da me prepozna jeftina kamera?

Well, when I'm not fighting the coded gaze

Kada se ne borim protiv kodiranog pogleda

as a poet of code,

kao pesnikinja kodova,

I'm a graduate student at the MIT Media Lab,

postdiplomac sam medijske laboratorije MIT-a

and there I have the opportunity to work on all sorts of whimsical projects,

i tamo imam priliku da radim na raznim neobičnim projektima,

including the Aspire Mirror,

uključujući „Ogledalo aspiracije“, projekat koji sam sprovela

a project I did so I could project digital masks onto my reflection.

tako da mogu da projektujem digitalne maske na svoj odraz.

So in the morning, if I wanted to feel powerful,

Tako bih ujutru, ako želim da se osećam snažno,

I could put on a lion.

mogla da stavim lava.

Ako bih htela da podignem raspoloženje, možda bih dobila citat.

If I wanted to be uplifted, I might have a quote.

So I used generic facial recognition software

Koristila sam generički softver za prepoznavanje lica

to build the system,

da bih napravila sistem,

ali sam otkrila da ga je teško testirati ukoliko ne nosim belu masku.

but found it was really hard to test it unless I wore a white mask.

Unfortunately, I've run into this issue before.

Nažalost, već sam ranije nailazila na ovaj problem.

When I was an undergraduate at Georgia Tech studying computer science,

Kada sam bila na osnovnim studijama na Tehnološkom institutu u Džordžiji,

gde sam studirala informatiku,

I used to work on social robots,

radila sam na društvenim robotima,

and one of my tasks was to get a robot to play peek-a-boo,

a jedan od mojih zadataka je bio da navedem robota da se igra skrivanja,

a simple turn-taking game

jednostavne igre menjanja uloga

where partners cover their face and then uncover it saying, "Peek-a-boo!"

u kojoj partneri pokrivaju lice, a zatim ga otkriju i kažu: „Uja!“

The problem is, peek-a-boo doesn't really work if I can't see you,

Problem je što igra skrivanja ne funkcioniše ako ne mogu da vas vidim,

and my robot couldn't see me.

a moj robot nije mogao da me vidi.

But I borrowed my roommate's face to get the project done,

No, pozajmila sam lice svoje cimerke da bih završila projekat,

submitted the assignment,

predala sam zadatak,

and figured, you know what, somebody else will solve this problem.

i mislila sam: „Znate šta, neko drugi će rešiti ovaj problem.“

Not too long after,

Nedugo potom,

I was in Hong Kong for an entrepreneurship competition.

bila sam u Hongkongu na takmičenju preduzetnika.

The organizers decided to take participants

Organizatori su rešili da povedu učesnike

on a tour of local start-ups.

u obilazak lokalnih startapova.

One of the start-ups had a social robot,

Jedan od startapova imao je društvenog robota

and they decided to do a demo.

i rešili su da naprave demonstraciju.

The demo worked on everybody until it got to me,

Demonstracija je radila kod svih dok nisu stigli do mene

and you can probably guess it.

i možete verovatno pretpostaviti šta se dogodilo.

It couldn't detect my face.

Nije mogao da prepozna moje lice.

I asked the developers what was going on,

Pitala sam programere šta se dešava

and it turned out we had used the same generic facial recognition software.

i ispostavilo se da smo koristili isti generički softver prepoznavanja lica.

Halfway around the world,

Preko pola sveta,

I learned that algorithmic bias can travel as quickly

saznala sam da algoritamska pristrasnost može putovati toliko brzo

as it takes to download some files off of the internet.

koliko treba da se skine nešto fajlova sa interneta.

So what's going on? Why isn't my face being detected?

Pa, šta se dešava? Zašto se moje lice ne prepoznaje?

Well, we have to look at how we give machines sight.

Pa, moramo pogledati kako mašini dajemo vid.

Kompjuterski vid koristi tehnike mašinskog učenja

Computer vision uses machine learning techniques

to do facial recognition.

da bi prepoznao lica.

So how this works is, you create a training set with examples of faces.

To funkcioniše tako što napravite komplet za vežbanje sa primerima lica.

This is a face. This is a face. This is not a face.

Ovo je lice. Ovo je lice. Ovo nije lice.

And over time, you can teach a computer how to recognize other faces.

Vremenom možete naučiti kompjuter kako da prepoznaje druga lica.

However, if the training sets aren't really that diverse,

Međutim, ako kompleti za vežbanje baš i nisu tako raznovrsni,

any face that deviates too much from the established norm

svako lice koje previše odstupa od uspostavljene norme

biće teže da se prepozna,

will be harder to detect,

which is what was happening to me.

a to je ono što se događa sa mnom.

But don't worry -- there's some good news.

Ali ne brinite, ima dobrih vesti.

Training sets don't just materialize out of nowhere.

Kompleti za vežbanje ne dolaze tek tako niotkuda.

We actually can create them.

Možemo ih stvoriti.

So there's an opportunity to create full-spectrum training sets

Postoji mogućnost za stvaranje kompleta za vežbu celokupnog spektra

that reflect a richer portrait of humanity.

koji odražavaju bogatiji portret čovečanstva.

Now you've seen in my examples

Videli ste u mojim primerima

da sam preko društvenih robota

how social robots

was how I found out about exclusion with algorithmic bias.

saznala za isključivanje kroz algoritamsku pristrasnost.

But algorithmic bias can also lead to discriminatory practices.

Ali algoritamska pristrasnost može dovesti i do prakse diskriminacije.

Across the US,

Širom SAD-a,

police departments are starting to use facial recognition software

policijske uprave počinju da koriste softver za prepoznavanje lica

in their crime-fighting arsenal.

u svom arsenalu za borbu protiv kriminala.

Georgetown Law published a report

Zakon Džordžtauna je objavio izveštaj

showing that one in two adults in the US -- that's 117 million people --

koji pokazuje da se jednoj od dve odrasle osobe u SAD-u -

to je 117 miliona ljudi -

have their faces in facial recognition networks.

lice nalazi u mrežama za prepoznavanje lica.

Police departments can currently look at these networks unregulated,

Odeljenja policije trenutno mogu da neregulisano pregledaju ove mreže,

using algorithms that have not been audited for accuracy.

pomoću algoritama kojima nije proverena tačnost.

Yet we know facial recognition is not fail proof,

Znamo da prepoznavanje lica nije bez mane,

and labeling faces consistently remains a challenge.

a naznačavanje lica stalno ostaje izazov.

You might have seen this on Facebook.

Možda ste to videli na Fejsbuku.

My friends and I laugh all the time when we see other people

Moji prijatelji i ja se uvek smejemo kad vidimo druge ljude

mislabeled in our photos.

koji su pogrešno označeni na našim fotografijama.

But misidentifying a suspected criminal is no laughing matter,

Ali pogrešno identifikovanje osumnjičenog zločinca nije za smejanje,

nor is breaching civil liberties.

kao ni kršenje građanskih sloboda.

Machine learning is being used for facial recognition,

Mašinsko učenje se koristi za prepoznavanje lica,

but it's also extending beyond the realm of computer vision.

ali se takođe proteže i van dometa kompjuterskog vida.

In her book, "Weapons of Math Destruction,"

U svojoj knjizi „Oružja za matematičko uništenje“,

data scientist Cathy O'Neil talks about the rising new WMDs --

naučnica u oblasti podataka Keti O'Nil govori o usponu novih RMD-a,

widespread, mysterious and destructive algorithms

rasprostranjenih, misterioznih i destruktivnih algoritama

that are increasingly being used to make decisions

koji se sve više koriste za donošenje odluka

that impact more aspects of our lives.

koje utiču na sve više aspekata našeg života.

So who gets hired or fired?

Koga će zaposliti ili otpustiti?

Da li ćete dobiti taj kredit? Da li ćete dobiti osiguranje?

Do you get that loan? Do you get insurance?

Are you admitted into the college you wanted to get into?

Da li ste primljeni na fakultet u koji ste želeli da upadnete?

Do you and I pay the same price for the same product

Da li vi i ja plaćamo istu cenu za isti proizvod

purchased on the same platform?

kupljen na istoj platformi?

Law enforcement is also starting to use machine learning

Sprovođenje zakona takođe počinje da koristi mašinsko učenje

for predictive policing.

za prediktivni rad policije.

Some judges use machine-generated risk scores to determine

Neke sudije koriste mašinski generisane procene rizika

how long an individual is going to spend in prison.

da bi odredile koliko vremena će neki pojedinac provesti u zatvoru.

So we really have to think about these decisions.

Zato zaista treba da razmislimo o ovim odlukama.

Are they fair?

Jesu li pravedne?

And we've seen that algorithmic bias

A videli smo da algoritamske predrasude

doesn't necessarily always lead to fair outcomes.

ne dovode nužno uvek do pravednih ishoda.

So what can we do about it?

Šta možemo da uradimo u vezi sa time?

Well, we can start thinking about how we create more inclusive code

Pa, možemo početi da razmišljamo o tome kako da stvorimo inkluzivniji kod

and employ inclusive coding practices.

i da upotrebimo inkluzivne postupke kodiranja.

It really starts with people.

To zapravo počinje od ljudi.

So who codes matters.

Zato je bitno ko kodira.

Are we creating full-spectrum teams with diverse individuals

Da li kreiramo timove celokupnog spektra sa različitim pojedincima

who can check each other's blind spots?

koji mogu da jedno drugome ispitaju stvari za koje su slepi?

On the technical side, how we code matters.

Sa tehničke strane, bitno je kako kodiramo.

Are we factoring in fairness as we're developing systems?

Da li uzimamo u obzir pravičnost dok razvijamo sisteme?

And finally, why we code matters.

I najzad, bitno je zašto kodiramo.

We've used tools of computational creation to unlock immense wealth.

Koristili smo alate računarskog stvaranja da bismo otključali ogromno bogatstvo.

We now have the opportunity to unlock even greater equality

Sada imamo priliku da otključamo još veću ravnopravnost

if we make social change a priority

ako učinimo društvene promene prioritetom,

and not an afterthought.

a ne da ih naknadno promišljamo.

And so these are the three tenets that will make up the "incoding" movement.

Dakle, ovo su tri principa koji će sačinjavati pokret „inkodiranja“.

Who codes matters,

Bitno je ko kodira,

how we code matters

kako kodiramo

and why we code matters.

i zašto kodiramo.

So to go towards incoding, we can start thinking about

Stoga, da bismo išli u pravcu inkodiranja, možemo početi da razmišljamo

building platforms that can identify bias

o izgradnji platforma koje mogu da identifikuju pristrasnost

by collecting people's experiences like the ones I shared,

prikupljanjem iskustava ljudi poput onih koje sam podelila,

but also auditing existing software.

ali i pregledom postojećeg softvera.

We can also start to create more inclusive training sets.

Takođe možemo početi da stvaramo inkluzivnije komplete za vežbanje.

Imagine a "Selfies for Inclusion" campaign

Zamislite kampanju „Selfiji za inkluziju“

where you and I can help developers test and create

u kojoj vi i ja možemo pomoći programerima da testiraju i naprave

more inclusive training sets.

inkluzivnije komplete za vežbanje.

And we can also start thinking more conscientiously

Takođe možemo početi da savesnije razmišljamo

about the social impact of the technology that we're developing.

o društvenom uticaju tehnologije koju razvijamo.

To get the incoding movement started,

Da bih otpočela pokret inkodiranja,

I've launched the Algorithmic Justice League,

pokrenula sam Ligu za algoritamsku pravdu,

where anyone who cares about fairness can help fight the coded gaze.

gde svako ko se brine o pravičnosti

može pomoći u borbi protiv kodiranog pogleda.

On codedgaze.com, you can report bias,

Na codedgaze.com možete prijaviti pristrasnost,

request audits, become a tester

zatražiti proveru, postati tester

and join the ongoing conversation,

i pridružiti se aktuelnom razgovoru,

#codedgaze.

#codedgaze.

So I invite you to join me

Pozivam vas da mi se pridružite

u stvaranju sveta u kome tehnologija radi za sve nas,

in creating a world where technology works for all of us,

not just some of us,

a ne samo za neke od nas,

a world where we value inclusion and center social change.

sveta u kome cenimo inkluziju i stavljamo u središte društvene promene.

Thank you.

Hvala.

(Applause)

(Aplauz)

But I have one question:

Ali imam jedno pitanje.

Will you join me in the fight?

Hoćete li mi se pridružiti u borbi?

(Laughter)

(Smeh)

(Applause)

(Aplauz)