How we can build AI to help humans, not hurt us | Margaret Mitchell

Prevodilac: Marko Mijatović Lektor: Ivana Krivokuća

I work on helping computers communicate about the world around us.

Ja pomažem da računari saopštavaju informacije o svetu oko nas.

There are a lot of ways to do this,

Postoji mnogo načina da se to uradi,

and I like to focus on helping computers

i ja volim da se fokusiram na pomaganje da računari

to talk about what they see and understand.

pričaju o tome šta vide i razumeju.

Given a scene like this,

Na slici kao što je ova,

a modern computer-vision algorithm

savremeni algoritmi za prepoznavanje oblika

can tell you that there's a woman and there's a dog.

će videti da su tu žena i pas.

It can tell you that the woman is smiling.

Prepoznaće da se žena smeje.

It might even be able to tell you that the dog is incredibly cute.

Može čak proceniti da je pas neverovatno sladak.

I work on this problem

Ja radim na ovom problemu

thinking about how humans understand and process the world.

razmišljajući kako ljudi razumeju svet i razmišljaju o njemu.

The thoughts, memories and stories

Misli, sećanja i priče

that a scene like this might evoke for humans.

koje slika kao što je ova mogu izazvati kod ljudi.

All the interconnections of related situations.

Sve međuzavisnosti povezanih situacija.

Maybe you've seen a dog like this one before,

Možda ste već videli psa kao što je ovaj,

or you've spent time running on a beach like this one,

ili ste provodili vreme u trčanju na plaži kao što je ova,

and that further evokes thoughts and memories of a past vacation,

a to dalje podstiče misli i sećanja na odmor iz prošlosti,

past times to the beach,

vreme provedeno na plaži

times spent running around with other dogs.

i u trčanju sa drugim psima.

One of my guiding principles is that by helping computers to understand

Jedan od mojih vodećih principa je da pomaganjem računarima da shvate

what it's like to have these experiences,

šta znači imati ovakva iskustva,

to understand what we share and believe and feel,

da shvate šta mi imamo zajedničko, u šta verujemo i šta osećamo,

then we're in a great position to start evolving computer technology

dolazimo do odlične pozicije za početak razvoja računarske tehnologije

in a way that's complementary with our own experiences.

tako da dopunjava naša sopstvena iskustva.

So, digging more deeply into this,

Ako uđemo malo dublje u temu,

a few years ago I began working on helping computers to generate human-like stories

pre par godina sam počela rad

na pomaganju računarima da stvaraju priče

from sequences of images.

na osnovu niza slika.

So, one day,

Jednog dana

I was working with my computer to ask it what it thought about a trip to Australia.

sam radila na računaru i pitala ga šta misli o putovanju u Australiju.

It took a look at the pictures, and it saw a koala.

Pogledao je slike i video je koalu.

It didn't know what the koala was,

Nije znao šta je koala,

but it said it thought it was an interesting-looking creature.

ali je rekao da misli da to biće izgleda zanimljivo.

Then I shared with it a sequence of images about a house burning down.

Zatim sam sa njim podelila niz slika kuća u plamenu.

It took a look at the images and it said,

Pogledao je slike i rekao:

"This is an amazing view! This is spectacular!"

„Ovo izgleda sjajno! Ovo je spektakularno!“

It sent chills down my spine.

Niz kičmu su mi prošli žmarci.

It saw a horrible, life-changing and life-destroying event

Video je užasan događaj koji menja i uništava živote ljudi

and thought it was something positive.

i mislio je da je to nešto pozitivno.

I realized that it recognized the contrast,

Shvatila sam da je prepoznao kontrast,

the reds, the yellows,

crvene i žute boje,

and thought it was something worth remarking on positively.

i da je mislio da je to nešto što treba označiti kao pozitivno.

And part of why it was doing this

Delimično je razlog tome

was because most of the images I had given it

to što je većina slika koje sam mu davala

were positive images.

bila pozitivna.

That's because people tend to share positive images

To je zbog toga što su ljudi skloni da dele pozitivne slike

when they talk about their experiences.

kada pričaju o svojim iskustvima.

When was the last time you saw a selfie at a funeral?

Kada ste poslednji put videli selfi sa sahrane?

I realized that, as I worked on improving AI

Shvatila sam da, dok sam radila na unapređenju veštačke inteligencije

task by task, dataset by dataset,

zadatak po zadatak, bazu po bazu,

that I was creating massive gaps,

da sam u stvari stvarala velike pukotine,

holes and blind spots in what it could understand.

rupe i „mrtve uglove“

u onome što može da razume.

And while doing so,

Time sam sam ubacivala različite pristrasnosti.

I was encoding all kinds of biases.

Biases that reflect a limited viewpoint,

Pristrasnosti koje odražavaju ograničeno gledište,

limited to a single dataset --

ograničeno na samo jedan skup podataka -

biases that can reflect human biases found in the data,

pristrasnosti koje odražavaju ljudske pristrasnosti pronađene u podacima

such as prejudice and stereotyping.

kao što su predrasude i stereotipi.

I thought back to the evolution of the technology

Razmislila sam o evoluciji tehnologije

that brought me to where I was that day --

koja me je dovela tu gde sam bila u tom trenutku -

how the first color images

kako su prve slike u boji

were calibrated against a white woman's skin,

bile podešene prema boji kože belih žena,

meaning that color photography was biased against black faces.

što znači da je fotografija u boji

bila naklonjena protiv crnih lica.

And that same bias, that same blind spot

Ta ista naklonjenost, isti „mrtav ugao“

continued well into the '90s.

se nastavio duboko u '90-im godinama.

And the same blind spot continues even today

Isti „mrtav ugao“ traje čak i danas

in how well we can recognize different people's faces

u tome koliko dobro možemo prepoznati lica različitih ljudi

in facial recognition technology.

u tehnologiji za prepoznavanje lica.

I though about the state of the art in research today,

Razmišljala sam o najnaprednijim istraživanjima danas,

where we tend to limit our thinking to one dataset and one problem.

gde pokušavamo da ograničimo svoje razmišljanje

na jedan skup podataka i jedan problem.

And that in doing so, we were creating more blind spots and biases

A čineći to, stvaramo još više „mrtvih uglova“ i predrasuda

that the AI could further amplify.

koje veštačka inteligencija može još više da pojača.

I realized then that we had to think deeply

Tada sam shvatila da moramo da dobro razmislimo

about how the technology we work on today looks in five years, in 10 years.

o tome kako će tehnologija na kojoj danas radimo

da izgleda za pet ili 10 godina.

Humans evolve slowly, with time to correct for issues

Ljudi evoluiraju polako, imaju vremena da isprave greške

in the interaction of humans and their environment.

u interakciji sa ljudima i okruženjem.

In contrast, artificial intelligence is evolving at an incredibly fast rate.

Za razliku od njih, veštačka inteligencija se razvija neverovatnom brzinom.

And that means that it really matters

To znači da je zaista bitno

that we think about this carefully right now --

da o ovome pažljivo razmislimo sada -

that we reflect on our own blind spots,

da razmislimo o svojim „mrtvim uglovima“,

our own biases,

o sopstvenim predrasudama,

and think about how that's informing the technology we're creating

i da razmislimo o tome kako to utiče na tehnologiju koju stvaramo

and discuss what the technology of today will mean for tomorrow.

i da razmotrimo šta će značiti tehnologija današnjice za sutrašnjicu.

CEOs and scientists have weighed in on what they think

Direktori i naučnici raspravljaju o tome kakva će biti

the artificial intelligence technology of the future will be.

veštačka inteligencija u budućnosti.

Stephen Hawking warns that

Stiven Hoking upozorava da

"Artificial intelligence could end mankind."

„Veštačka inteligencija može uništiti čovečanstvo.“

Elon Musk warns that it's an existential risk

Ilon Mask upozorava da je to rizik za preživljavanje

and one of the greatest risks that we face as a civilization.

i jedan od najvećih rizika sa kojima se susrećemo kao civilizacija.

Bill Gates has made the point,

Bil Gejts je izneo stav:

"I don't understand why people aren't more concerned."

„Ne razumem zašto ljudi nisu više zabrinuti.“

But these views --

Ali ova mišljenja

they're part of the story.

su deo priče.

The math, the models,

Matematika, modeli,

the basic building blocks of artificial intelligence

osnovni sastavni delovi veštačke inteligencije

are something that we call access and all work with.

su nešto čemu svi možemo da pristupimo i da radimo sa time.

We have open-source tools for machine learning and intelligence

Imamo alate otvorenog koda za mašinsko učenje i inteligenciju

that we can contribute to.

kojima možemo da doprinesemo.

And beyond that, we can share our experience.

Osim toga, možemo da delimo naše iskustvo.

We can share our experiences with technology and how it concerns us

Možemo da delimo naša iskustva sa tehnologijom, kako nas brine

and how it excites us.

i kako nas uzbuđuje.

We can discuss what we love.

Možemo da razmatramo šta volimo.

We can communicate with foresight

Možemo da govorimo

about the aspects of technology that could be more beneficial

o budućim aspektima tehnologije koji bi bili delotvorniji

or could be more problematic over time.

ili problematičniji tokom vremena.

If we all focus on opening up the discussion on AI

Ako se svi fokusiramo na otvaranje diskusije o veštačkoj inteligenciji

with foresight towards the future,

sa pogledom na budućnost,

this will help create a general conversation and awareness

to će pomoći kreiranju uopštenih razgovora i svesti

about what AI is now,

o tome šta je veštačka inteligencija sada,

what it can become

šta može da postane

and all the things that we need to do

i sve stvari koje treba da uradimo

in order to enable that outcome that best suits us.

da ostvarimo rezultat koji bi nam najbolje služio.

We already see and know this in the technology that we use today.

Već vidimo i znamo to iz tehnologije koju danas koristimo.

We use smart phones and digital assistants and Roombas.

Koristimo pametne telefone, digitalne pomoćnike i robotske usisivače.

Are they evil?

Da li su zli?

Maybe sometimes.

Možda ponekad.

Are they beneficial?

Da li su korisni?

Yes, they're that, too.

Jesu i to.

And they're not all the same.

I nisu svi isti.

And there you already see a light shining on what the future holds.

Već možemo da vidimo da je budućnost svetla.

The future continues on from what we build and create right now.

Budućnost nastaje iz onoga što kreiramo upravo sada.

We set into motion that domino effect

Mi pokrećemo taj domino efekat

that carves out AI's evolutionary path.

koji kreira evolucioni put veštačke inteligencije.

In our time right now, we shape the AI of tomorrow.

U svom vremenu sada, mi oblikujemo veštačku inteligenciju sutrašnjice.

Technology that immerses us in augmented realities

Tehnologiju koja nas uvodi u izmenjenu stvarnost

bringing to life past worlds.

oživljavajući prošlost.

Technology that helps people to share their experiences

Tehnologiju koja pomaže ljudima da dele svoja iskustva

when they have difficulty communicating.

kada imaju problem da komuniciraju.

Technology built on understanding the streaming visual worlds

Tehnologiju zasnovanu na prenosu sadržaja vizuelnog sveta,

used as technology for self-driving cars.

korišćenu kao tehnologiju za samovozeće automobile.

Technology built on understanding images and generating language,

Tehnologiju zasnovanu na razumevanju slika i generisanju jezika,

evolving into technology that helps people who are visually impaired

koja evoluira u tehnologiju koja pomaže ljudima sa oštećenim vidom

be better able to access the visual world.

da imaju bolji pristup vizuelnom svetu.

And we also see how technology can lead to problems.

Takođe vidimo kako tehnologija može da dovede do problema.

We have technology today

Danas imamo tehnologiju

that analyzes physical characteristics we're born with --

koja analizira fizičke karakteristike sa kojima smo rođeni -

such as the color of our skin or the look of our face --

kao što je boja kože ili izgled lica -

in order to determine whether or not we might be criminals or terrorists.

kako bi odredila da li smo možda kriminalci ili teroristi.

We have technology that crunches through our data,

Imamo tehnologiju koja pretražuje naše podatke,

even data relating to our gender or our race,

čak i podatke o našem polu ili rasi,

in order to determine whether or not we might get a loan.

da bi odredila da li možemo da dobijemo kredit.

All that we see now

Sve što sada vidimo

is a snapshot in the evolution of artificial intelligence.

je trenutna slika u evoluciji veštačke inteligencije.

Because where we are right now,

Zato što je ovo gde smo sada

is within a moment of that evolution.

samo trenutak u toj evoluciji.

That means that what we do now will affect what happens down the line

To znači da će ono što sada radimo uticati na to šta se dešava

and in the future.

u budućnosti.

Ukoliko želimo da pomognemo da se veštačka inteligencija razvije

If we want AI to evolve in a way that helps humans,

tako da pomogne ljudima,

then we need to define the goals and strategies

onda moramo da definišemo ciljeve i strategije

that enable that path now.

koje će omogućiti taj put sada.

What I'd like to see is something that fits well with humans,

Ono što bih želela da vidim je nešto što će se uklopiti sa ljudima,

with our culture and with the environment.

sa našom kulturom i okruženjem.

Technology that aids and assists those of us with neurological conditions

Tehnologiju koja pomaže onima sa neurološkim problemima

or other disabilities

ili drugim vidovima invaliditeta

in order to make life equally challenging for everyone.

kako bi život bio podjednako izazovan za svakoga.

Technology that works

Tehnologiju koja radi

regardless of your demographics or the color of your skin.

bez obzira na poreklo ili boju kože.

And so today, what I focus on is the technology for tomorrow

Tako da ono na šta se danas fokusiram je tehnologija sutrašnjice

and for 10 years from now.

i 10 godina od danas.

AI can turn out in many different ways.

Veštačka inteligencija može da krene različitim putevima.

But in this case,

Ali u ovom slučaju,

it isn't a self-driving car without any destination.

to nije samovozeći automobil bez cilja.

This is the car that we are driving.

To je automobil koji mi vozimo.

We choose when to speed up and when to slow down.

Mi biramo kada da ubrzamo i kada da usporimo.

We choose if we need to make a turn.

Mi biramo da li treba da skrenemo.

We choose what the AI of the future will be.

Mi biramo šta će veštačka inteligencija budućnosti zapravo biti.

There's a vast playing field

Veliki je broj mogućnosti

of all the things that artificial intelligence can become.

šta veštačka inteligencija može postati.

It will become many things.

I postaće dosta toga.

And it's up to us now,

Sada je na nama

in order to figure out what we need to put in place

da smislimo šta treba da uradimo

to make sure the outcomes of artificial intelligence

da bismo bili sigurni da rezultati veštačke inteligencije

are the ones that will be better for all of us.

budu oni koji su dobri za sve nas.

Thank you.

Hvala.

(Applause)

(Aplauz)