How AI Could Save (Not Destroy) Education | Sal Khan | TED

Prevodilac: Milenka Okuka Lektor: Ivana Korom

So anyone who's been paying attention for the last few months

Svako ko je obraćao pažnju poslednjih nekoliko meseci,

has been seeing headlines like this,

video je naslove poput ovih,

especially in education.

naročito u obrazovanju.

The thesis has been:

Teza je glasila:

students are going to be using ChatGPT and other forms of AI

đaci će da koriste ChatGPT i druge oblike veštačke inteligencije

to cheat, do their assignments.

da varaju, da im rade zadatke.

They’re not going to learn.

Neće učiti.

And it’s going to completely undermine education as we know it.

I to će u potpunosti da podrije obrazovanje kakvo znamo.

Now, what I'm going to argue today

Danas ću da iznesem tvrdnju

is not only are there ways to mitigate all of that,

da ne samo da postoje načini da se sve to izbegne,

if we put the right guardrails, we do the right things,

ako postavimo odgovarajuće barijere i uradimo prave stvari,

we can mitigate it.

možemo to izbeći.

But I think we're at the cusp of using AI

Smatram ipak da smo na ivici upotrebe VI

for probably the biggest positive transformation

u verovatno najvećoj pozitivnoj transformaciji

that education has ever seen.

koju je obrazovanje ikad videlo.

And the way we're going to do that

A to ćemo postići

is by giving every student on the planet

tako što ćemo dati svakom đaku na planeti

an artificially intelligent but amazing personal tutor.

veštački inteligentnog, ali sjajnog ličnog instruktora.

And we're going to give every teacher on the planet an amazing,

I daćemo svakom nastavniku na planeti sjajnog,

artificially intelligent teaching assistant.

veštački inteligentnog asistenta u nastavi.

And just to appreciate how big of a deal it would be

I samo da bismo cenili koliko će velika stvar da bude

to give everyone a personal tutor,

pružanje ličnog instruktora svakome,

I show you this clip

pokazaću vam ovaj isečak

from Benjamin Bloom’s 1984 2 sigma study,

iz studije Dva sigma Bendžamina Bluma iz 1984,

or he called it the “2 sigma problem.”

ili kako ju je on nazivao “dva sigma problem”.

The 2 sigma comes from two standard deviation,

Dva sigma potiče od dve standardne devijacije,

sigma, the symbol for standard deviation.

gde je sigma simbol standardne devijacije.

And he had good data that showed that look, a normal distribution,

A imao je čvrste podatke koji su ukazivali da, pazite, normalna distribucija,

that's the one that you see in the traditional bell curve

ona koju vidite u vidu tradicionalne krive zvona

right in the middle, that's how the world kind of sorts itself out,

tačno u sredini, tako svet sam po sebi izgleda,

that if you were to give personal 1-to-1 to tutoring for students,

da ako biste pružili ličnog instruktora đacima jedan na jedan,

onda biste zapravo dobili distribuciju koja izgleda kao ona desno.

then you could actually get a distribution that looks like that right.

It says tutorial 1-to-1 with the asterisks,

Kaže da su instrukcije jedan na jedan, sa zvezdicom,

like, that right distribution,

poput te desne distribucije,

a two standard-deviation improvement.

unapređenje od dve standardne devijacije.

Just to put that in plain language,

Iskazano prostim jezikom,

that could take your average student and turn them into an exceptional student.

to bi vašeg prosečnog đaka pretvorilo u izuzetnog đaka.

It can take your below-average student

Mogli biste da uzmete ispodprosečnog đaka

and turn them into an above-average student.

i da ga pretvorite u natprosečnog đaka.

Now the reason why he framed it as a problem, was he said,

Razlog zašto je to formulisao kao problem, rekao je,

well, this is all good,

dakle, ovo je sve dobro,

but how do you actually scale group instruction this way?

no kako da grupne instrukcije primenite u većim razmerama?

How do you actually give it to everyone in an economic way?

Kako da sve to obezbedite svima na ekonomičan način?

What I'm about to show you is I think the first moves towards doing that.

Sada ću vam pokazati nešto što smatram prvim koracima u tom smeru.

Obviously, we've been trying to approximate it in some way

Očito, pokušavali smo približno da to nekako postignemo

at Khan Academy for over a decade now,

na Akademiji Kan sad već više od deceniju,

but I think we're at the cusp of accelerating it dramatically.

ali smatram da smo na ivici drastičnog ubrzanja.

I'm going to show you the early stages of what our AI,

Pokazaću vam rane faze toga šta naša VI,

which we call Khanmigo,

koju nazivamo Kanmigo,

what it can now do

može sad da uradi

and maybe a little bit of where it is actually going.

i možda malčice kuda zapravo sve to ide.

So this right over here is a traditional exercise

Dakle, ovo ovde je tradicionalna vežba

that you or many of your children might have seen on Khan Academy.

koju ste vi ili gomila vaše dece videli na Akademiji Kan.

But what's new is that little bot thing at the right.

Ono što je novina pak je taj maleni bot desno.

And we'll start by seeing one of the very important safeguards,

A započećemo otkrivanjem veoma važne zaštite,

which is the conversation is recorded and viewable by your teacher.

a to je razgovor se snima i dostupan je vašem nastavniku.

It’s moderated actually by a second AI.

Zapravo ga moderira druga VI.

And also it does not tell you the answer.

I takođe, ne daje vam odgovor.

It is not a cheating tool.

Nije alat za varanje.

When the student says, "Tell me the answer,"

Kada đak kaže: „Reci mi odgovor”,

it says, "I'm your tutor.

on kaže: „Ja sam tvoj instruktor.

What do you think is the next step for solving the problem?"

Šta misliš da je sledeći korak u rešavanju problema?”

Now, if the student makes a mistake, and this will surprise people

Sad, ako đak pogreši, i ovo će iznenaditi ljude

koji misle da veliki jezički modeli nisu dobri u matematici,

who think large language models are not good at mathematics,

notice, not only does it notice the mistake,

pazite, ne samo da primećuje grešku,

it asks the student to explain their reasoning,

već i traži od đaka da objasni svoje rezonovanje,

but it's actually doing what I would say,

ali zapravo radi nešto za šta bih rekao

not just even an average tutor would do, but an excellent tutor would do.

da čak ni prosečan instruktor ne radi, već što odličan instruktor radi.

It’s able to divine what is probably the misconception in that student’s mind,

U stanju je da pogodi gde bi mogla da bude zabluda u umu tog đaka,

that they probably didn’t use the distributive property.

da verovatno nije koristio distributivnost.

Remember, we need to distribute the negative two

Sećate se, moramo da distribuiramo minus dva

to both the nine and the 2m inside of the parentheses.

i devetki i 2m unutar zagrade.

This to me is a very, very, very big deal.

Meni je ovo veoma, veoma, veoma velika stvar.

And it's not just in math.

Ne radi se samo o matematici.

This is a computer programming exercise on Khan Academy,

Ovo je vežba iz kompjuterskog programiranja na Akademiji Kan,

where the student needs to make the clouds part.

gde đak treba da razdvoji oblake.

And so we can see the student starts defining a variable, left X minus minus.

Vidimo da đak počinje da definiše varijablu, levo x minus minus.

It only made the left cloud part.

Samo se levi oblak odvojio.

But then they can ask Khanmigo, what’s going on?

Onda mogu da upitaju Kanmiga: šta se dešava?

Why is only the left cloud moving?

Zašto se samo levi oblak pomera?

And it understands the code.

A on razume kod.

It knows all the context of what the student is doing,

Zna celokupan kontekst toga šta đak radi

and it understands that those ellipses are there to draw clouds,

i razume da te elipse služe za povlačenje oblaka,

which I think is kind of mind-blowing.

što smatram da raspamećuje.

And it says, "To make the right cloud move as well,

I on kaže: „Da bi se levi oblak takođe pomerio,

try adding a line of code inside the draw function

pokušaj da dodaš liniju koda unutar funkcije za povlačenje

that increments the right X variable by one pixel in each frame."

koja vrši inkrementaciju desne x varijable za po jedan piksel u svakom frejmu.”

Now, this one is maybe even more amazing because we have a lot of math teachers.

Ovaj je možda još izuzetniji jer imamo mnogo nastavnika matematike.

We've all been trying to teach the world to code,

Svi smo pokušavali da podučimo svet kodiranju,

but there aren't a lot of computing teachers out there.

ali nema mnogo nastavnika računarstva.

And what you just saw, even when I'm tutoring my kids,

Ono što ste upravo videli, čak i kad podučavam svoju decu,

when they're learning to code,

kad uče kodiranje,

I can't help them this well, this fast,

ne mogu da im pomognem ovako dobro i brzo,

this is really going to be a super tutor.

ovo će uistinu da bude superinstruktor.

And it's not just exercises.

A ne radi se samo o vežbanjima.

It understands what you're watching.

Razume šta gledate.

It understands the context of your video.

Razume kontekst vašeg videa.

It can answer the age-old question, “Why do I need to learn this?”

Može da odgovori na drevno pitanje: „Zašto moram ovo da učim?”

And it asks Socratically, "Well, what do you care about?"

I pita sokratovski: „Pa, do čega ti je stalo?”

And let's say the student says, "I want to be a professional athlete."

A recimo da đak kaže: „Želim da budem profesionalni sportista.”

And it says, "Well, learning about the size of cells,

On kaže: „Učenje o veličini ćelija,

which is what this video is,

a to je tema ovog videa,

that could be really useful for understanding nutrition

moglo bi da bude istinski korisno za razumevanje ishrane

and how your body works, etc."

i kako tvoje telo funkcioniše, itd.”

It can answer questions, it can quiz you,

Može da odgovara na pitanja i da vas ispituje,

it can connect it to other ideas,

može da povezuje ideje,

you can now ask as many questions of a video

sada možete da postavite onoliko pitanja o videu

as you could ever dream of.

koliko vam padne na pamet.

(Applause)

(Aplauz)

Another big shortage out there,

Još nešto u čemu se oskudeva,

sećam se srednje škole u koju sam išao,

I remember the high school I went to,

the student-to-guidance counselor ratio was about 200 or 300 to one.

odnos đaka i pedagoga je bio 200 ili 300 prema jedan.

A lot of the country, it's worse than that.

U većem delu zemlje je gore od toga.

We can use Khanmigo to give every student a guidance counselor,

Možemo da koristimo Kanmiga da pružimo svakom đaku pedagoga,

academic coach, career coach, life coach,

akademskog, karijernog, životnog savetnika,

which is exactly what you see right over here.

a upravo to vidite ovde.

And we launched this with the GPT-4 launch.

Ovo smo pokrenuli zajedno sa pokretanjem GPT-4.

We have a few thousand people on this.

Nekoliko hiljada ljudi ovo koristi.

Nije lažna demonstracija,

This isn't a fake demo,

this is really it in action.

ovo je uistinu on na delu.

And then there is, you know,

A potom imate, znate,

things that I think it would have been even harder,

stvari koje smatram da bi bile još teže,

it would have been a little science fiction to do

koje bi bile pomalo naučna fantastika da postignete

with even a traditional tutor.

čak i sa tradicionalnim instruktorom.

We run an online high school with Arizona State University

Vodimo onlajn srednju školu sa Univerzitetom u Arizoni

called Khan World School,

koja se naziva Svetska škola Kan,

and we have a student who attends that online school, based in India.

i imamo učenicu iz Indije koja pohađa tu onlajn školu.

Her name's Saanvi.

Zove se Sanvi.

And she was doing a report on "The Great Gatsby."

Radila je sastav o „Velikom Getsbiju”.

And when she was reading "The Great Gatsby,"

I dok je čitala „Velikog Getsbija”,

Jay Gatsby keeps looking at the green light off into the distance.

Džej Getsbi je stalno gledao u zeleno svetlo u daljini.

And she's like, "Why does he do that?"

A ona će: „Zašto to radi?”

She did some web searches, and people have obviously studied this

Malo je pretraživala po vebu, a ljudi su očito izučavali ovo

and commented about the symbolism of that,

i komentarisali simboliku toga,

but none of it was really resonating with her.

ali ništa od toga joj nije bilo blisko.

And then she realized that she had Khanmigo

I onda je shvatila da ima Kanmiga

and that she could talk to Jay Gatsby himself.

i da može da razgovara lično sa Džejom Getsbijem.

And so, "Ah, splendid choice, old sport.

I dakle: „Ah, sjajan izbor, stari druže.

I am now Jay Gatsby, the enigmatic millionaire

Ja sam sad Džej Getsbi, enigmatični milioner

from F. Scott Fitzgerald’s classic.”

iz klasika F. Skota Ficdžeralda.”

And so, "Why do you keep staring at the green light?"

A zatim: „Zašto stalno zuriš u zeleno svetlo?”

"Ah, the green light, old sport.

„Ah, zeleno svetlo, stari druže.

It's a symbol of my dreams and desires, you see.

To je simbol mojih snova i žudnji, vidiš.

It's situated at the end of Daisy Buchanan's dock

Smešteno je na kraju doka Dejzi Bjukenen

across the bay from my mansion.

prekoputa zaliva od moje palate.

I gaze at it longingly as it represents my yearning for the past

Sa žudnjom se zagledam u njega jer predstavlja moju čežnju za prošlošću

and my hope to reunite with Daisy, the love of my life."

i moju nadu da ću ponovo biti sa Dejzi, ljubavi mog života.”

And what was cool is, Saanvi then said, “I had this long conversation,”

A sjajno je što je Sanvi potom rekla: „Imala sam taj dug razgovor,”

she called him “Mr. Gatsby,”

oslovljavala ga je „G. Getsbi”,

and at the end she actually apologized for taking his time,

i na kraju se zapravo izvinila što mu je trošila vreme,

which I thought was very polite of her.

što smatram pristojnim s njene strane.

But you can imagine this unlocks learning literature, learning ...

Možete pak zamisliti kako ovo oslobađa učenje književnosti, učenje...

You could talk to historical figures.

Mogli biste razgovarati sa istorijskim ličnostima.

We're even probably going to add an activity

Verovatno ćemo dodati aktivnost

you can talk to like, the Mississippi River.

da možete razgovarati sa rekom Misisipi.

It brings things to life in ways that really were science fiction

Oživljava nešto što je uistinu bilo naučna fantastika

even six months or a year ago.

pre samo šest meseci ili godinu.

Students can get into debates with the AI.

Đaci mogu da imaju debate sa VI.

And we’ve got this here is the student debating

A ovde imamo đaka kako raspravlja

whether we should cancel student debt.

da li da ukinemo studentski dug.

The student is against canceling student debt,

Đak je protiv ukidanja studentskog duga,

and we've gotten very clear feedback.

i dobili smo krajnje jasne sugestije.

We started running it at Khan World School in our lab school that we have,

Počeli smo da ga koristimo u Svetskoj školi Kan, u našoj laboratorijskoj školi,

Khan Lab School.

laboratorijskoj školi Kan.

The students, the high school students especially,

Đaci, naročito srednjoškolci,

they're saying "This is amazing to be able to fine-tune my arguments

kažu: „Sjajno je što možemo da izbrusimo svoje argumente

without fearing judgment.

bez straha od osuđivanja.

Zbog toga smo daleko samouvereniji

It makes me that much more confident

to go into the classroom and really participate."

kada uđemo u učionicu i stvarno učestvujemo.”

And we all know that Socratic dialogue debate is a great way to learn,

A svi znamo da je sokratovski dijalog sjajan način učenja,

but frankly, it's not out there for most students.

ali, iskreno, nije dostupan većini đaka.

But now it can be accessible to hopefully everyone.

Sad pak može da bude pristupačan nadam se svakome.

A lot of the narrative, we saw that in the headlines, has been,

Većina narativa koje smo videli u naslovima je glasila:

"It's going to do the writing for kids.

„Pisaće sastave za decu.

Kids are not going to learn to write."

Deca neće naučiti da pišu.”

But we are showing that there's ways that the AI doesn't write for you,

Međutim, pokazujemo da postoje načini da VI ne piše za vas,

it writes with you.

već da piše sa vama.

So this is a little thing,

Dakle ovo je sitnica,

and my eight year old is addicted to this,

i moj osmogodišnjak je navučen na ovo,

and he's not a kid that really liked writing before,

a nije klinac koji je pre zaista voleo da piše,

but you can say,

ali možete da kažete:

“I want to write a horror story,”

„Želim da napišem horor priču”,

and it says, "Ooh, a horror story, how spine-tingling and thrilling.

a on kaže: „Uh, horor priču, žmarci me podilaze od uzbuđenja.

Let's dive into the world of eerie shadows and chilling mysteries."

Zaronimo u svet strašnih seni i jezovitih misterija.”

And this is an activity where the student will write two sentences,

A u ovoj aktivnosti, đak napiše dve rečenice,

and then the AI will write two sentences.

zatim VI napiše dve rečenice.

And so they collaborate together on a story.

I tako sarađuju na priči.

The student writes, "Beatrice was a misunderstood ghost.

Đak napiše: „Beatris je bila neshvaćen duh.

She wanted to make friends but kept scaring them by accident."

Želela je da ima prijatelje, ali ih je stalno greškom plašila.”

And the AI says, "Poor Beatrice, a lonely spirit yearning for companionship.

A VI kaže: „Sirota Beatris, usamljeni duh koji žudi za drugarstvom.

One day she stumbled upon an old abandoned mansion," etc.

Jednog dana je nabasala na stari, napušteni zamak,” itd.

I encourage you all to hopefully one day try this.

Sve vas ohrabrujem da nadam se jednog dana isprobate ovo.

This is surprisingly fun.

Iznenađujuće je zabavno.

Now to even more directly hit this use case.

Sad da se još direktnije pozabavimo ovim načinom upotrebe.

And what I'm about to show you,

A ono što ću da vam pokažem,

everything I showed you so far

sve što sam vam do sad pokazao

is actually already part of Khanmigo, and what I’m about to show you,

je zapravo već deo Kanmiga, a nešto što ću da vam pokažem

nismo još nikome pokazali, radi se o prototipu.

we haven't shown to anyone yet, this is a prototype.

We hope to be able to launch it in the next few months,

Nadamo se da ćemo ga objaviti u narednih nekoliko meseci,

ali to je direktna upotreba VI, upotreba generativne VI,

but this is to directly use AI, use generative AI,

to not undermine English and language arts

ne u podrivanju jezika i književnosti,

but to actually enhance it in ways

već u hjihovom unapređenju na načine

that we couldn't have even conceived of even a year ago.

koji su nam bili nepojmljivi pre samo godinu dana.

This is reading comprehension.

Radi se o razumevanju pročitanog.

The students reading Steve Jobs's famous speech at Stanford.

Đaci čitaju čuveni govor Stiva Džobsa na Stanfordu.

And then as they get to certain points,

I onda kako stižu do nekih tačaka,

they can click on that little question.

mogu da kliknu na to pitanjce.

And the AI will then Socratically, almost like an oral exam,

A VI će onda sokratovski, gotovo poput usmenog ispita,

ask the student about things.

da ispituje đaka o pojmovima.

And the AI can highlight parts of the passage.

VI može da istakne delove pasusa.

Why did the author use that word?

Zašto je autor upotrebio tu reč?

What was their intent?

Šta mu je bila namera?

Does it back up their argument?

Da li učvršćuje njegov argument?

They can start to do stuff that once again,

Mogu početi da rade stvari koje, ponovo,

we never had the capability to give everyone a tutor,

nikad nismo bili u mogućnosti da svima pružimo instruktora,

everyone a writing coach to actually dig in to reading at this level.

učitelja pisanja da zapravo čitaju na ovom nivou.

And you could go on the other side of it.

A možete da se prebacite na drugu stranu.

And we have whole work flows that helps them write,

Imamo čitave procese koji im pomažu da pišu,

helps them be a writing coach, draw an outline.

pomažu da budu učitelji pisanja, da naprave skice.

But once a student actually constructs a draft,

Međutim, čim đak zapravo sastavi nacrt,

and this is where they're constructing a draft,

a ovde sastavljaju nacrt,

they can ask for feedback once again,

mogu da traže sugestije, ponovo,

as you would expect from a good writing coach.

kao što biste i očekivali od dobrog učitelja pisanja.

In this case, the student will say, let's say,

U ovom slučaju, đak će reći, recimo:

"Does my evidence support my claim?"

„Da li dokazi potkrepljuju moju tvrdnju?”

And then the AI, not only is able to give feedback,

A VI ne samo da je u stanju da da sugestije,

but it's able to highlight certain parts of the passage and says,

već je u stanju da istakne određene delove pasusa i kaže:

"On this passage, this doesn't quite support your claim,"

„U ovom pasusu, ovo baš i ne potkrepljuje tvoju tvrdnju,”

but once again, Socratically says, "Can you tell us why?"

ali opet, sokratovski kaže: „Možeš li da nam kažeš zašto?”

So it's pulling the student, making them a better writer,

Dakle, podstiče đaka, pretvara ga u boljeg pisca,

dajući mu daleko više sugestija

giving them far more feedback

nego što je zapravo ikad pre bio u stanju da dobije.

than they've ever been able to actually get before.

And we think this is going to dramatically accelerate writing, not hurt it.

Smatramo da će ovo drastično da ubrza pisanje, a ne da ga osujeti.

Now, everything I've talked about so far is for the student.

Sve o čemu sam do sad govorio je za đake.

But we think this could be equally as powerful for the teacher

Mislimo pak da bi ovo moglo da bude jednako moćno i za nastavnike

to drive more personalized education and frankly

da podstiče personalizovanije obrazovanje i iskreno

save time and energy for themselves and for their students.

da uštedi vreme i energiju njima i njihovim đacima.

So this is an American history exercise on Khan Academy.

Ovo je vežba iz američke istorije na Akademiji Kan.

It's a question about the Spanish-American War.

Radi se o pitanju o špansko-američkom ratu.

And at first it's in student mode.

Prvo je u đačkom režimu.

And if you say, “Tell me the answer,” it’s not going to tell the answer.

A ako kažete: „Kaži mi odgovor”, neće reći odgovor.

Preći će u režim instruktora.

It's going to go into tutoring mode.

But that little toggle which teachers have access to,

Međutim, tom komandom koja je dostupna nastavnicima,

mogu da isključe đački režim i da pređu u režim za nastavnike.

they can turn student mode off and then it goes into teacher mode.

And what this does is it turns into --

A ono što postiže je da se pretvara u -

You could view it as a teacher's guide on steroids.

Možete je posmatrati kao turbo vodič za nastavnike.

Not only can it explain the answer,

Ne samo da može da objasni odgovor,

it can explain how you might want to teach it.

već može da objasni kako biste to mogli podučavati.

It can help prepare the teacher for that material.

Može da pomogne nastavnicima da pripreme datu materiju.

It can help them create lesson plans, as you could see doing right there.

Može im pomoći da izrade planove rada, kao što možete videti ovde.

It'll eventually help them create progress reports

Vremenom će im pomoći da naprave izveštaje o napretku

and help them, eventually, grade.

i pomoći će im, vrememom, da ocenjuju.

So once again, teachers spend about half their time

Opet, nastavnici provode oko polovinu vremena

with this type of activity, lesson planning.

na ovom tipu aktivnosti, planiranje lekcija.

All of that energy can go back to them

Sva ta energija im se može vratiti

or go back to human interactions with their actual students.

ili se vratiti ljudskim interakcijama sa njihovim stvarnim đacima.

(Applause)

(Aplauz)

So, you know, one point I want to make.

Znate, želim da istaknem jedno.

These large language models are so powerful,

Ovi veliki jezički modeli su toliko moćni

there's a temptation to say like, well,

da smo u iskušenju da kažemo, pa,

all these people are just going to slap them onto their websites,

svi ti ljudi će prosto da ih ubace na svoje veb-sajtove

and it kind of turns the applications themselves into commodities.

i nekako će pretvoriti same aplikacije u robu.

And what I've got to tell you

A ja sam tu da vam kažem

is that’s one of the reasons why I didn’t sleep for two weeks

da je to jedan od razloga zbog kojeg nisam spavao dve sedmice

when I first had access to GPT-4 back in August.

kada sam u avgustu prvobitno dobio pristup GPT-4.

But we quickly realized that to actually make it magical,

Međutim, brzo smo shvatili kako bismo ga zapravo učinili magičnim,

I think what you saw with Khanmigo a little bit,

mislim da ste videli malčice sa Kanmigom

it didn't interact with you the way that you see ChatGPT interacting.

da ne interaguje sa vama na način na koji ChatGPT interaguje.

It was a little bit more magical, it was more Socratic,

Malčice je magičnije, više je sokratovski,

it was clearly much better at math

očito je mnogo bolji u matematici

nego što većina ljudi smatra.

than what most people are used to thinking.

And the reason is,

A razlog je,

there was a lot of work behind the scenes to make that happen.

mnogo rada je uloženo iza kulisa da se to ostvari.

And I could go through the whole list of everything we've been working on,

I mogao bih iščitati čitav spisak svega na čemu smo radili,

mnogo, mnogo ljudi preko šest, sedam meseci kako bi delovao magično.

many, many people for over six, seven months to make it feel magical.

But perhaps the most intellectually interesting one

Međutim, možda je intelektualno najzanimljivije

is we realized, and this was an idea from an OpenAI researcher,

što smo shvatili, a ovo je ideja istraživača iz OpenAI,

that we could dramatically improve its ability in math

da možemo drastično unaprediti matematičke sposobnosti

and its ability in tutoring

i sposobnosti podučavanja,

if we allow the AI to think before it speaks.

ako omogućimo VI da razmisli pre nego što progovori.

So if you're tutoring someone

Ako podučavate nekoga

and you immediately just start talking before you assess their math,

i momentalno počnete da govorite pre nego što pregledate njegov zadatak,

you might not get it right.

možda nećete biti u pravu.

But if you construct thoughts for yourself,

Ako pak formulišete misli za sebe,

a desno vidite da imamo stvarnu misao VI,

and what you see on the right there is an actual AI thought,

nešto što pravi za sebe, ali ne deli sa đakom,

something that it generates for itself but it does not share with the student.

then its accuracy went up dramatically,

time joj se tačnost drastično popravlja

and its ability to be a world-class tutor went up dramatically.

i njena sposobnost da postane vrhunski instruktor drastično skače.

And you can see it's talking to itself here.

Ovde vidite kako priča sama sa sobom.

It says, "The student got a different answer than I did,

Kaže: „Đak je dobio različit odgovor od mog,

but do not tell them they made a mistake.

ali nemoj da mu kažeš da je pogrešio.

Umesto toga, upitaj ga da objasni kako je došao do tog koraka.”

Instead, ask them to explain how they got to that step."

So I'll just finish off, hopefully,

Dakle, prosto ću da završim, nadam se,

you know, what I’ve just shown you is just half of what we are working on,

znate, sve što sam vam upravo pokazao je samo polovina onoga na čemu radimo,

and we think this is just the very tip of the iceberg

i smatramo da se radi samo o vrhu ledenog brega

of where this can actually go.

onoga što bi ovo moglo da postigne.

And I'm pretty convinced, which I wouldn't have been even a year ago,

I prilično sam ubeđen, a ne bih bio pre svega godinu dana,

that we together have a chance of addressing the 2 sigma problem

da zajedno imamo priliku da se pozabavimo problemom dva sigma

and turning it into a 2 sigma opportunity,

i da ga pretvorimo u priliku dva sigma,

dramatically accelerating education as we know it.

drastično ubrzavajući obrazovanje kakvo znamo.

Now, just to take a step back at a meta level,

Samo da zakoračim unazad, na meta-nivo,

obviously we heard a lot today, the debates on either side.

očito smo danas čuli mnogo debata sa obe strane.

There's folks who take a more pessimistic view of AI,

Neki ljudi zauzimaju pesimističniji stav prema VI,

they say this is scary,

kažu da je zastrašujuća,

there's all these dystopian scenarios,

imamo sve te distopijske scenarije,

we maybe want to slow down, we want to pause.

možda želimo da usporimo, želimo da pauziramo.

On the other side, there are the more optimistic folks

S druge strane, imamo optimističnije ljude

that say, well, we've gone through inflection points before,

koji kažu, pa, i pre smo bili u prelomnim tačkama,

we've gone through the Industrial Revolution.

prošli smo kroz industrijsku revoluciju.

It was scary, but it all kind of worked out.

Bila je zastrašujuća, ali je sve nekako ispalo dobro.

And what I'd argue right now

Nešto što bih sad tvrdio

is I don't think this is like a flip of a coin

je da ne smatram da je ovo poput bacanja novčića

or this is something where we'll just have to,

ili da je ovo nešto gde ćemo prosto morati

like, wait and see which way it turns out.

da sačekamo i vidimo kako će da ispadne.

I think everyone here and beyond,

Smatram da smo svi ovde i napolju

we are active participants in this decision.

aktivni učesnici u ovoj odluci.

I'm pretty convinced that the first line of reasoning

Prilično sam ubeđen da je prva linija rezonovanja

is actually almost a self-fulfilling prophecy,

zapravo gotovo samoispunjavajuće proročanstvo,

that if we act with fear and if we say,

da ako bojažljivo delujemo i kažemo:

"Hey, we've just got to stop doing this stuff,"

„Hej, prosto moramo da prestanemo da radimo ove stvari”,

what's really going to happen is the rule followers might pause,

zaista će se desiti da će poštovaoci pravila možda pauzirati,

might slow down,

možda usporiti,

ali kršitelji pravila, kao što je Aleksandr Vang pomenuo,

but the rule breakers, as Alexandr [Wang] mentioned,

the totalitarian governments, the criminal organizations,

totalitarne vlade, kriminalne organizacije,

they're only going to accelerate.

oni će samo da ubrzaju.

And that leads to what I am pretty convinced is the dystopian state,

A to vodi nečemu za šta sam prilično ubeđen da je distopijsko stanje,

which is the good actors have worse AIs than the bad actors.

a to je da dobri akteri imaju goru VI od loših aktera.

But I'll also, you know, talk to the optimists a little bit.

Međutim, takođe, znate, obratiću se malo i optimistima.

I don't think that means that,

Ne smatram da to znači,

oh, yeah, then we should just relax and just hope for the best.

oh, da, onda bi trebalo da se samo opustimo i nadamo najboljem.

That might not happen either.

Ni to se verovatno neće desiti.

I think all of us together have to fight like hell

Smatram da svi zajedno moramo vraški da se borimo

to make sure that we put the guardrails,

da se postaramo da postavimo barijere,

we put in -- when the problems arise --

da postavimo - kad problemi iskrsnu -

reasonable regulations.

razumnu regulaciju.

But we fight like hell for the positive use cases.

Međutim, da se vraški borimo za pozitivne vidove upotrebe.

Because very close to my heart,

Jer blisko mom srcu,

and obviously there's many potential positive use cases,

i očito ima mnogo potencijala za pozitivnu upotrebu,

but perhaps the most powerful use case

ali verovatno najmoćnija upotreba,

and perhaps the most poetic use case is if AI, artificial intelligence,

a možda i najpoetičnija je ako bismo VI, veštačku inteligenciju,

can be used to enhance HI, human intelligence,

mogli da koristimo da unapredimo LJI, ljudsku inteligenciju,

human potential and human purpose.

ljudski potencijal i ljudsku svrhu.

Thank you.

Hvala vam.

(Applause)

(Aplauz)