Machine intelligence makes human morals more important | Zeynep Tufekci

Prevodilac: Milenka Okuka Lektor: Mile Živković

So, I started my first job as a computer programmer

Dakle, počela sam da radim kao kompjuterska programerka

in my very first year of college --

kad sam bila prva godina na fakultetu -

basically, as a teenager.

u suštini kao tinejdžerka.

Soon after I started working,

Ubrzo nakon što sam počela da radim,

writing software in a company,

da pišem softvere za firmu,

a manager who worked at the company came down to where I was,

menadžer koji je radio u firmi se spustio do mene

and he whispered to me,

i prošaputao mi je:

"Can he tell if I'm lying?"

"Zna li da lažem?"

There was nobody else in the room.

Nije bilo više bilo koga u prostoriji.

"Can who tell if you're lying? And why are we whispering?"

"Zna li ko da lažete? I zašto šapućemo?"

The manager pointed at the computer in the room.

Menadžer je pokazao na kompjuter u prostoriji.

"Can he tell if I'm lying?"

"Zna li da lažem?"

Well, that manager was having an affair with the receptionist.

Pa, taj menadžer je imao aferu sa recepcionerom.

(Laughter)

(Smeh)

And I was still a teenager.

A ja sam još uvek bila tinejdžerka.

So I whisper-shouted back to him,

Pa sam mu polušapatom odgovorila:

"Yes, the computer can tell if you're lying."

"Da, kompjuter zna kad mu lažete."

(Laughter)

(Smeh)

Well, I laughed, but actually, the laugh's on me.

Pa, smejala sam se, ali zapravo šala je na moj račun.

Nowadays, there are computational systems

Ovih dana imamo kompjuterske sisteme

that can suss out emotional states and even lying

koji mogu da prozru emocionalna stanja, čak i laganje,

from processing human faces.

obradom ljudskih lica.

Advertisers and even governments are very interested.

Oglašivači, čak i vlade su veoma zainteresovane za to.

I had become a computer programmer

Postala sam kompjuterska programerka

because I was one of those kids crazy about math and science.

jer sam bila jedno od dece koja su luda za matematikom i naukom.

But somewhere along the line I'd learned about nuclear weapons,

No, nekako sam usput saznala za nuklearno oružje

and I'd gotten really concerned with the ethics of science.

i postala sam zaista zabrinuta zbog naučne etike.

I was troubled.

Mučilo me je to.

However, because of family circumstances,

Međutim, zbog porodičnih okolnosti,

I also needed to start working as soon as possible.

takođe je bilo potrebno da počnem da radim što pre.

So I thought to myself, hey, let me pick a technical field

Pa sam pomislila u sebi, hej, hajde da izaberem tehničku oblast

where I can get a job easily

gde mogu lako da se zaposlim

and where I don't have to deal with any troublesome questions of ethics.

i gde ne moram da se bavim bilo kakvim mučnim etičkim pitanjem.

So I picked computers.

Pa sam odabrala kompjutere.

(Laughter)

(Smeh)

Well, ha, ha, ha! All the laughs are on me.

Pa, ha, ha, ha! Šala je skroz na moj račun.

Nowadays, computer scientists are building platforms

Ovih dana kompjuterski naučnici prave platforme

that control what a billion people see every day.

koje upravljaju onim što milijarde ljudi gledaju svakodnevno.

They're developing cars that could decide who to run over.

Razvijaju automobile koji mogu da odluče koga da pregaze.

They're even building machines, weapons,

Čak grade mašine, oružja,

that might kill human beings in war.

koja mogu da ubijaju ljude u ratu.

It's ethics all the way down.

U potpunosti se radi o etici.

Machine intelligence is here.

Mašinska inteligencija je tu.

We're now using computation to make all sort of decisions,

Trenutno koristimo kompjutere da donesemo razne odluke,

but also new kinds of decisions.

ali i nove tipove odluka.

We're asking questions to computation that have no single right answers,

Postavljamo kompjuterima pitanja koja nemaju jedan pravi odgovor,

that are subjective

koja su subjektivna,

and open-ended and value-laden.

otvorena i krcata vrednostima.

We're asking questions like,

Postavljamo ovakva pitanja:

"Who should the company hire?"

"Koga da firma zaposli?"

"Which update from which friend should you be shown?"

"Koje ažuriranje od kog prijatelja treba da bude vidljivo?"

"Which convict is more likely to reoffend?"

"Koji osuđenik je skloniji novom prestupu?"

"Which news item or movie should be recommended to people?"

"Koji novinski čalanak ili film treba preporučiti ljudima?"

Look, yes, we've been using computers for a while,

Gledajte, da, već neko vreme koristimo kompjutere,

but this is different.

ali ovo je drugačije.

This is a historical twist,

Ovo je istorijski preokret

because we cannot anchor computation for such subjective decisions

jer ne možemo da usidrimo kompjutere kod sličnih subjektivnih odluka

the way we can anchor computation for flying airplanes, building bridges,

na način na koji usidravamo kompjutere koji upravljaju avionima, grade mostove,

going to the moon.

idu na mesec.

Are airplanes safer? Did the bridge sway and fall?

Jesu li avioni bezbedniji? Da li se most zaljuljao i pao?

There, we have agreed-upon, fairly clear benchmarks,

Tu imamo uspostavljene prilično jasne repere

and we have laws of nature to guide us.

i tu su zakoni prirode da nas vode.

We have no such anchors and benchmarks

Nemamo slična težišta i repere

for decisions in messy human affairs.

za odluke koje se tiču haotičnih ljudskih odnosa.

To make things more complicated, our software is getting more powerful,

Da bi stvari bile još složenije, naši softveri postaju sve moćniji,

but it's also getting less transparent and more complex.

ali takođe postaju manje transparentni i složeniji.

Recently, in the past decade,

U skorije vreme, u poslednjoj deceniji,

complex algorithms have made great strides.

složeni algoritmi su poprilično napredovali.

They can recognize human faces.

Mogu da prepoznaju ljudska lica.

They can decipher handwriting.

Mogu da dešifruju rukopis.

They can detect credit card fraud

Mogu da zapaze prevaru kod kreditnih kartica

and block spam

i da blokiraju spamove

and they can translate between languages.

i da prevode s jezika na jezik.

They can detect tumors in medical imaging.

Mogu da zapaze tumore na medicinskim snimcima.

They can beat humans in chess and Go.

Mogu da pobede ljude u šahu i gou.

Much of this progress comes from a method called "machine learning."

Veliki deo ovog napretka potiče od metoda nazvanog "mašinsko učenje".

Machine learning is different than traditional programming,

Mašinsko učenje se razlikuje od tradicionalnog programiranja,

where you give the computer detailed, exact, painstaking instructions.

gde dajete kompjuteru detaljne, tačne, minuciozne instrukcije.

It's more like you take the system and you feed it lots of data,

Pre se radi o odabiru sistema i pohranjivanju podataka u njega,

including unstructured data,

uključujući nestrukturirane podatke,

like the kind we generate in our digital lives.

poput onih koje stvaramo u digitalnim životima.

And the system learns by churning through this data.

A sistem uči, pretresajući podatke.

And also, crucially,

Suštinsko je takođe

these systems don't operate under a single-answer logic.

da se ovi sistemi ne vode logikom samo jednog odgovora.

They don't produce a simple answer; it's more probabilistic:

Ne proizvode jednostavne odgovore; više se radi o verovatnoći:

"This one is probably more like what you're looking for."

"Ovo je verovatno sličnije onome što tražite."

Now, the upside is: this method is really powerful.

Sad, pozitivno je: ovaj metod je zaista moćan.

The head of Google's AI systems called it,

Glavni u Guglovom sistemu za VI je to nazvao:

"the unreasonable effectiveness of data."

"nerazumna efikasnost podataka."

The downside is,

Negativno je:

we don't really understand what the system learned.

ne razumemo zaista šta je sistem naučio.

In fact, that's its power.

Zapravo, to je njegova moć.

This is less like giving instructions to a computer;

Ovo manje liči na davanje uputstava kompjuteru;

it's more like training a puppy-machine-creature

više liči na dresiranje bića - mehaničko kuče,

we don't really understand or control.

koje zaista ne razumemo, niti kontrolišemo.

So this is our problem.

Dakle, to je naš problem.

It's a problem when this artificial intelligence system gets things wrong.

Problem je kad ovaj sistem veštačke inteligencije nešto pogrešno shvati.

It's also a problem when it gets things right,

Takođe je problem kad nešto dobro shvati.

because we don't even know which is which when it's a subjective problem.

jer čak ni ne znamo šta je šta kod subjektivnog problema.

We don't know what this thing is thinking.

Ne znamo o čemu ova stvar razmišlja.

So, consider a hiring algorithm --

Dakle, uzmite u obzir algoritam za zapošljavanje -

a system used to hire people, using machine-learning systems.

sistem koji se koristi pri zapošljavanju, koji koristi sisteme mašinskog učenja.

Such a system would have been trained on previous employees' data

Sličan sistem je obučavan na podacima prethodnih zaposlenih

and instructed to find and hire

i naučen je da pronalazi i zapošljava

people like the existing high performers in the company.

ljude poput postojećih najučinkovitijih u firmi.

Sounds good.

Zvuči dobro.

I once attended a conference

Jednom sam bila na konferenciji

that brought together human resources managers and executives,

koja je spojila menadžere iz kadrovske službe i direktore,

high-level people,

ljude s visokih pozicija,

using such systems in hiring.

koristeći ove sisteme zapošljavanja.

They were super excited.

Bili su veoma uzbuđeni.

They thought that this would make hiring more objective, less biased,

Smatrali su da bi zbog ovoga zapošljavanje bilo objektivnije, nepristrasnije,

and give women and minorities a better shot

i da bi žene i manjine imale više šanse,

against biased human managers.

nasuprot pristrasnim ljudskim menadžerima.

And look -- human hiring is biased.

I, gledajte - zapošljavanje ljudi je pristrasno.

I know.

Znam.

I mean, in one of my early jobs as a programmer,

Mislim, na jednom od mojih prvih poslova kao programerke,

my immediate manager would sometimes come down to where I was

moja nadređena menadžerka bi ponekad prišla mestu na kom sam,

really early in the morning or really late in the afternoon,

veoma rano ujutru ili veoma kasno poslepodne,

and she'd say, "Zeynep, let's go to lunch!"

i rekla bi: "Zejnep, pođimo na ručak!"

I'd be puzzled by the weird timing.

Zbunilo bi me neobično vreme.

It's 4pm. Lunch?

Četiri je popodne. Ručak?

I was broke, so free lunch. I always went.

Bila sam švorc, pa sam uvek išla na besplatan ručak.

I later realized what was happening.

Kasnije sam shvatila o čemu se radilo.

My immediate managers had not confessed to their higher-ups

Moji nadređeni menadžeri nisu priznali svojim nadređenim

that the programmer they hired for a serious job was a teen girl

da je programer kog su zaposlili za ozbiljan posao bila tinejdžerka

who wore jeans and sneakers to work.

koja je nosila farmerke i patike na posao.

I was doing a good job, I just looked wrong

Bila sam dobar radnik, samo pogrešnog izgleda

and was the wrong age and gender.

i bila sam pogrešnih godina i roda.

So hiring in a gender- and race-blind way

Pa zapošljavanje na rodno i rasno nepristrasan način

certainly sounds good to me.

izvesno da mi zvuči dobro.

But with these systems, it is more complicated, and here's why:

Ali uz ove sisteme, složenije je, a evo zašto:

Currently, computational systems can infer all sorts of things about you

trenutno kompjuterski sistemi mogu da zaključe razne stvari o vama

from your digital crumbs,

iz vaših digitalnih tragova,

even if you have not disclosed those things.

čak iako to niste obelodanili.

They can infer your sexual orientation,

Mogu da zaključe vašu seksualnu orijentaciju,

your personality traits,

vaše lične osobine,

your political leanings.

vaša politička naginjanja.

They have predictive power with high levels of accuracy.

Imaju moć predviđanja sa visokim stepenom tačnosti.

Remember -- for things you haven't even disclosed.

Zapamtite - za ono što čak niste ni obelodanili.

This is inference.

To je zaključivanje.

I have a friend who developed such computational systems

Imam prijateljicu koja je razvila sličan kompjuterski sistem

to predict the likelihood of clinical or postpartum depression

za predviđanje verovatnoće kliničke ili postporođajne depresije

from social media data.

iz podataka sa društvenih mreža.

The results are impressive.

Rezultati su bili impresivni.

Her system can predict the likelihood of depression

Njeni sistemi mogu da predvide verovatnoću depresije

months before the onset of any symptoms --

mesecima pre nastupa bilo kakvih simptoma -

months before.

mesecima ranije.

No symptoms, there's prediction.

Bez simptoma imamo predviđanje.

She hopes it will be used for early intervention. Great!

Ona se nada da će biti korišćeni za rane intervencije. Sjajno!

But now put this in the context of hiring.

Sad ovo stavite u kontekst zapošljavanja.

So at this human resources managers conference,

Pa sam na ovoj konferenciji menadžera iz kadrovske

I approached a high-level manager in a very large company,

prišla visokoprofilnoj menadžerki iz prilično velike firme,

and I said to her, "Look, what if, unbeknownst to you,

i rekla sam joj: "Pazi, šta ako bi, bez tvog znanja,

your system is weeding out people with high future likelihood of depression?

ovaj sistem iskorenjivao ljude sa velikim izgledima za depresiju u budućnosti?

They're not depressed now, just maybe in the future, more likely.

Trenutno nisu depresivni, ali je veća verovatnoća da će biti u budućnosti.

What if it's weeding out women more likely to be pregnant

Šta ako iskorenjuje žene s većom verovatnoćom da zatrudne

in the next year or two but aren't pregnant now?

u narednih godinu ili dve, ali trenutno nisu trudne?

What if it's hiring aggressive people because that's your workplace culture?"

Šta ako zapošljava agresivne ljude jer je to kultura na vašem radnom mestu?"

You can't tell this by looking at gender breakdowns.

Ovo ne možete da vidite, posmatrajući rodnu nejednakost.

Those may be balanced.

Ona bi mogla da bude u ravnoteži.

And since this is machine learning, not traditional coding,

A kako se radi o mašinskom učenju, a ne tradicionalnom programiranju,

there is no variable there labeled "higher risk of depression,"

tu nemamo varijablu s oznakom "veći rizik od depresije",

"higher risk of pregnancy,"

"veći rizik za trudnoću",

"aggressive guy scale."

"skala agresivnih muškaraca".

Not only do you not know what your system is selecting on,

Ne samo da ne znate na osnovu čega vaš sistem bira,

you don't even know where to begin to look.

čak ne znate ni gde da gledate.

It's a black box.

To je crna kutija.

It has predictive power, but you don't understand it.

Ima moć predviđanja, ali je vi ne razumete.

"What safeguards," I asked, "do you have

"Koja vam je zaštita", pitala sam, "koju imate

to make sure that your black box isn't doing something shady?"

kojom se starate da crna kutija ne obavlja nešto sumnjivo?"

She looked at me as if I had just stepped on 10 puppy tails.

Pogledala me je kao da sam nagazila na 10 kučećih repića.

(Laughter)

(Smeh)

She stared at me and she said,

Buljila je u mene i rekla:

"I don't want to hear another word about this."

"Ne želim da čujem ni reč više o ovome."

And she turned around and walked away.

Okrenula se i otišla.

Mind you -- she wasn't rude.

Pazite - nije bila nepristojna.

It was clearly: what I don't know isn't my problem, go away, death stare.

Jasno se radilo o ovome: ono što ne znam nije moj problem, nestani, prazan pogeld.

(Laughter)

(Smeh)

Look, such a system may even be less biased

Vidite, sličan sistem bi mogao čak da bude na neki način

than human managers in some ways.

i manje pristrasan od ljudskih menadžera.

And it could make monetary sense.

I mogao bi da ima finansijsku prednost.

But it could also lead

Ali bi takođe mogao da dovede

to a steady but stealthy shutting out of the job market

do stabilnog, ali prikrivenog isključivanja sa tržišta rada

of people with higher risk of depression.

ljudi s većim rizikom od depresije.

Is this the kind of society we want to build,

Da li je ovo oblik društva koji želimo da gradimo,

without even knowing we've done this,

a da pri tom ne znamo da smo to uradili

because we turned decision-making to machines we don't totally understand?

jer smo prepustili donošenje odluka mašinama koje u potpunosti ne razumemo?

Another problem is this:

Drugi problem je sledeće:

these systems are often trained on data generated by our actions,

ovi sistemi su često obučavani na podacima koje proizvode naša delanja,

human imprints.

na ljudskom otisku.

Well, they could just be reflecting our biases,

Pa, oni bi prosto mogli da odražavaju naše predrasude,

and these systems could be picking up on our biases

i ovi sistemi bi mogli da pokupe naše predrasude

and amplifying them

i da ih naglase

and showing them back to us,

i potom da nam ih pokažu,

while we're telling ourselves,

dok mi govorimo sebi:

"We're just doing objective, neutral computation."

"Samo izvodimo objektivne, neutralne proračune."

Researchers found that on Google,

Istraživači su otkrili da na Guglu

women are less likely than men to be shown job ads for high-paying jobs.

ženama mnogo ređe nego muškarcima prikazuju oglase za dobro plaćene poslove.

And searching for African-American names

A pretraga afroameričkih imena

is more likely to bring up ads suggesting criminal history,

često sa sobom povlači oglase koji nagoveštavaju kriminalnu prošlost,

even when there is none.

čak i kad ona ne postoji.

Such hidden biases and black-box algorithms

Slične prikrivene predrasude i algoritmi nalik crnoj kutiji,

that researchers uncover sometimes but sometimes we don't know,

koje istraživači povremeno otkrivaju, ali ponekad to ne uspeju,

can have life-altering consequences.

mogu da imaju ozbiljne posledice.

In Wisconsin, a defendant was sentenced to six years in prison

Okrivljeni iz Viskonsina je osuđen na šest godina zatvora

for evading the police.

zbog izbegavanja policije.

You may not know this,

Možda ne znate za ovo,

but algorithms are increasingly used in parole and sentencing decisions.

ali algoritme sve češće koriste u odlučivanju o uslovnoj ili kazni.

He wanted to know: How is this score calculated?

Želeo je da zna: kako su izračunali ovaj rezultat?

It's a commercial black box.

To je komercijalna crna kutija.

The company refused to have its algorithm be challenged in open court.

Firma je odbila da njen algoritam izazovu na javnom suđenju.

But ProPublica, an investigative nonprofit, audited that very algorithm

No, ProPublica, istraživačka neprofitna organizacija je proverila taj algoritam

with what public data they could find,

sa podacima koje su uspeli da nađu

and found that its outcomes were biased

i otkrili su da su njihovi rezultati pristrasni,

and its predictive power was dismal, barely better than chance,

a da je njihova moć predviđanja očajna, jedva bolja od nagađanja

and it was wrongly labeling black defendants as future criminals

i da su pogrešno označavali okrivljene crnce kao buduće kriminalce,

at twice the rate of white defendants.

dvostruko češće nego okrivljene belce.

So, consider this case:

Pa, razmotrite ovaj slučaj:

This woman was late picking up her godsister

ova žena je kasnila da pokupi svoje kumče

from a school in Broward County, Florida,

iz okruga Brauard u Floridi,

running down the street with a friend of hers.

trčala je niz ulicu sa svojom prijateljicom.

They spotted an unlocked kid's bike and a scooter on a porch

Spazile su nezaključan dečji bicikl i skuter na tremu

and foolishly jumped on it.

i nesmotreno su sele na bicikl.

As they were speeding off, a woman came out and said,

Dok su jurile dalje, žena je izašla i rekla:

"Hey! That's my kid's bike!"

"Hej! To je bicikl mog deteta!"

They dropped it, they walked away, but they were arrested.

Ostavile su ga, odšetale, ali su uhapšene.

She was wrong, she was foolish, but she was also just 18.

Pogrešila je, bila je nesmotrena, ali je takođe imala svega 18 godina.

She had a couple of juvenile misdemeanors.

Imala je nekoliko maloletničkih prekršaja.

Meanwhile, that man had been arrested for shoplifting in Home Depot --

U međuvremenu, ovaj čovek je uhapšen zbog krađe u supermarketu -

85 dollars' worth of stuff, a similar petty crime.

robe u vrednosti od 85 dolara, sličan manji zločin.

But he had two prior armed robbery convictions.

Ali je pre toga imao dve osude zbog oružane pljačke.

But the algorithm scored her as high risk, and not him.

Ali je algoritam nju ocenio kao visokorizičnu, a njega nije.

Two years later, ProPublica found that she had not reoffended.

Dve godine kasnije, ProPublica je otkrila da ona nije imala novih prekršaja.

It was just hard to get a job for her with her record.

Samo joj je sa dosijeom bilo teško da nađe posao.

He, on the other hand, did reoffend

On, s druge strane, ponovo je u zatvoru

and is now serving an eight-year prison term for a later crime.

i ovaj put služi kaznu od osam godina zbog kasnijeg zločina.

Clearly, we need to audit our black boxes

Očigledno, moramo da proveravamo naše crne kutije

and not have them have this kind of unchecked power.

kako ne bi imale sličnu nekontrolisanu moć.

(Applause)

(Aplauz)

Audits are great and important, but they don't solve all our problems.

Provere su sjajne i važne, ali ne rešavaju sve naše probleme.

Take Facebook's powerful news feed algorithm --

Uzmite Fejsbukov moćan algoritam za dostavu vesti -

you know, the one that ranks everything and decides what to show you

znate, onaj koji sve rangira i odlučuje šta da vam pokaže

from all the friends and pages you follow.

od svih prijatelja i stranica koje pratite.

Should you be shown another baby picture?

Da li da vam pokaže još jednu sliku bebe?

(Laughter)

(Smeh)

A sullen note from an acquaintance?

Sumornu poruku od poznanika?

An important but difficult news item?

Važnu, ali tešku vest?

There's no right answer.

Nema pravog odgovora.

Facebook optimizes for engagement on the site:

Fejsbuk najviše ima koristi od angažmana na sajtu:

likes, shares, comments.

sviđanja, deljenja, komentara.

In August of 2014,

Avgusta 2014,

protests broke out in Ferguson, Missouri,

izbili su protesti u Fergusonu, Misuri,

after the killing of an African-American teenager by a white police officer,

nakon ubistva afroameričkog tinejdžera od strane policajca belca,

under murky circumstances.

pod nejasnim okolnostima.

The news of the protests was all over

Vesti o protestima su bile svuda

my algorithmically unfiltered Twitter feed,

po mom algoritamski nefilterisanom Tviter nalogu,

but nowhere on my Facebook.

ali nigde na mom Fejsbuku.

Was it my Facebook friends?

Da li su krivi prijatelji na Fejsbuku?

I disabled Facebook's algorithm,

Onemogućila sam Fejsbukov algoritam,

which is hard because Facebook keeps wanting to make you

a to je teško jer Fejsbuk nastoji da vas natera

come under the algorithm's control,

da budete pod kontrolom algoritma,

and saw that my friends were talking about it.

i videla sam da moji prijatelji govore o tome.

It's just that the algorithm wasn't showing it to me.

Samo mi moj algoritam to nije pokazivao.

I researched this and found this was a widespread problem.

Istražila sam ovo i otkrila da je ovo raširen problem.

The story of Ferguson wasn't algorithm-friendly.

Priča o Fergusonu nije bila prihvatljiva za algoritam.

It's not "likable."

Nije nešto za "sviđanje".

Who's going to click on "like?"

Ko će da pritisne "sviđanje"?

It's not even easy to comment on.

Nije je čak lako ni komentarisati.

Without likes and comments,

Bez sviđanja i komentara,

the algorithm was likely showing it to even fewer people,

algoritam je težio da je prikaže čak i manjem broju ljudi,

so we didn't get to see this.

pa nismo mogli da je vidimo.

Instead, that week,

Umesto toga, te sedmice,

Facebook's algorithm highlighted this,

Fejsbukov algoritam je izdvojio ovo,

which is the ALS Ice Bucket Challenge.

a to je ALS ledeni izazov.

Worthy cause; dump ice water, donate to charity, fine.

Plemenit cilj; polij se ledenom vodom, doniraj u dobrotvorne svrhe, fino.

But it was super algorithm-friendly.

Ali bilo je veoma algoritamski prihvatljivo.

The machine made this decision for us.

Mašina je donela ovu odluku u naše ime.

A very important but difficult conversation

Veoma važan, ali težak razgovor

might have been smothered,

je mogao da bude ugušen,

had Facebook been the only channel.

da je Fejsbuk bio jedini kanal.

Now, finally, these systems can also be wrong

Sad, naposletku, ovi sistemi takođe mogu da greše

in ways that don't resemble human systems.

drugačije od ljudskih sistema.

Do you guys remember Watson, IBM's machine-intelligence system

Ljudi, sećate li se Votsona, IBM-ovog sistema mašinske inteligencije

that wiped the floor with human contestants on Jeopardy?

koji je obrisao pod ljudskim takmičarima na kvizu?

It was a great player.

Bio je sjajan takmičar.

But then, for Final Jeopardy, Watson was asked this question:

Ali su ga onda u finalnom izazovu upitali sledeće pitanje:

"Its largest airport is named for a World War II hero,

"Njegov najveći aerodrom je nazvan po heroju iz II svetskog rata,

its second-largest for a World War II battle."

a drugi po veličini po bici iz II svetskog rata."

(Hums Final Jeopardy music)

(Pevuši finalnu temu iz kviza)

Chicago.

Čikago.

The two humans got it right.

Oba ljudska bića su pogodila.

Watson, on the other hand, answered "Toronto" --

Votson, s druge strane, odgovorio je: "Toronto" -

for a US city category!

za kategoriju gradova SAD-a.

The impressive system also made an error

Impresivni sistem je napravio i grešku

that a human would never make, a second-grader wouldn't make.

koju ljudsko biće nikad ne bi, drugaš je nikad ne bi napravio.

Our machine intelligence can fail

Naša mašinska inteligencija može da omane

in ways that don't fit error patterns of humans,

na načine koji se ne uklapaju sa obrascima grešenja kod ljudi,

in ways we won't expect and be prepared for.

na načine koji su neočekivani i na koje nismo pripremljeni.

It'd be lousy not to get a job one is qualified for,

Bilo bi loše ne dobiti posao za koji ste kvalifikovani,

but it would triple suck if it was because of stack overflow

ali bi bilo tri puta gore ako bi to bilo zbog preopterećenja

in some subroutine.

u nekakvoj sistemskoj podrutini.

(Laughter)

(Smeh)

In May of 2010,

U maju 2010,

a flash crash on Wall Street fueled by a feedback loop

munjevit krah na Vol Stritu je pokrenut povratnom petljom

in Wall Street's "sell" algorithm

u Vol Stritovom algoritmu "prodaja",

wiped a trillion dollars of value in 36 minutes.

izbrisao je vrednost od trilion dolara za 36 minuta.

I don't even want to think what "error" means

Ne želim da razmišljam o tome šta znači "greška"

in the context of lethal autonomous weapons.

u kontekstu smrtonosnog autonomnog oružja.

So yes, humans have always made biases.

Dakle, da, ljudi su oduvek bili pristrasni.

Decision makers and gatekeepers,

Donosioci odluka i čuvari informacija,

in courts, in news, in war ...

na sudovima, vestima, u ratu...

they make mistakes; but that's exactly my point.

oni greše; ali u tome je poenta.

We cannot escape these difficult questions.

Ne možemo da izbegnemo ova teška pitanja.

We cannot outsource our responsibilities to machines.

Ne možemo da delegiramo naša zaduženja mašinama.

(Applause)

(Aplauz)

Artificial intelligence does not give us a "Get out of ethics free" card.

Veštačka inteligencija nam ne pruža besplatnu kartu za "beg od etike".

Data scientist Fred Benenson calls this math-washing.

Naučnik za podatke Fred Benson to naziva matematičkim ispiranjem.

We need the opposite.

Potrebno nam je suprotno.

We need to cultivate algorithm suspicion, scrutiny and investigation.

Moramo da negujemo sumnju u algoritme, nadzor i istraživanje.

We need to make sure we have algorithmic accountability,

Moramo da se postaramo da imamo algoritamsku odgovrnost,

auditing and meaningful transparency.

proveru i smislenu transparentnost.

We need to accept that bringing math and computation

Moramo da prihvatimo da uvođenje matematike i kompjutera

to messy, value-laden human affairs

u neuredne ljudske odnose vođene vrednostima

does not bring objectivity;

ne donosi objektivnost;

rather, the complexity of human affairs invades the algorithms.

već pre složenost ljudskih odnosa osvaja algoritme.

Yes, we can and we should use computation

Da, možemo i treba da koristimo kompjutere

to help us make better decisions.

kako bi donosili bolje odluke.

But we have to own up to our moral responsibility to judgment,

Ali moramo da ovladamo našom moralnom odgovornošću i rasuđivanjem

and use algorithms within that framework,

i da koristimo algoritme unutar tog okvira,

not as a means to abdicate and outsource our responsibilities

ne kao sredstva da se odreknemo i da delegiramo naše odgovornosti

to one another as human to human.

nekom drugom, kao čovek čoveku.

Machine intelligence is here.

Mašinska inteligencija je tu.

That means we must hold on ever tighter

To znači da se kao nikad pre moramo čvrsto držati

to human values and human ethics.

ljudskih vrednosti i ljudske etike.

Thank you.

Hvala vam.

(Applause)

(Aplauz)