A headset that reads your brainwaves | Tan Le

Prevodilac: Žarko Milićević Lektor: Sandra Gojic

Up until now, our communication with machines

Sve do nedavno, komunikacija sa mašinama

has always been limited

je bila ograničena

to conscious and direct forms.

na namerne i direktne postupke.

Whether it's something simple

Svejedno da li je u pitanju nešto jednostavno,

like turning on the lights with a switch,

recimo, paljenje svetla prekidačem,

or even as complex as programming robotics,

ili složeno, poput programiranja robota,

we have always had to give a command to a machine,

mašini smo uvek morali zadati naredbu,

or even a series of commands,

ili, čak, niz naredbi,

in order for it to do something for us.

kako bi ona tad nešto izvršila.

Communication between people, on the other hand,

Komunikacija između ljudi

is far more complex and a lot more interesting

je daleko složenija i zanimljivija

because we take into account

jer obuhvata mnogo više

so much more than what is explicitly expressed.

od onoga što se iskazuje rečima.

We observe facial expressions, body language,

Bitan je izraz lica, govor tela,

and we can intuit feelings and emotions

jer neposredno opažamo osećanja sagovornika

from our dialogue with one another.

iz samog dijaloga ali i ponašanja.

This actually forms a large part

To je jako veliki deo

of our decision-making process.

procesa donošenja odluka.

Our vision is to introduce

Naš cilj je da uvedemo

this whole new realm of human interaction

tu novu dimenziju komunikacije

into human-computer interaction

u interakciju čoveka sa kompjuterom,

so that computers can understand

da omogućimo kompjuteru da razume

not only what you direct it to do,

ne samo šta treba da uradi,

but it can also respond

već da reaguje

to your facial expressions

i na izraze lica

and emotional experiences.

i ispoljavanje emocija.

And what better way to do this

A nema boljeg načina da postignemo cilj

than by interpreting the signals

od tumačenja signala

naturally produced by our brain,

koje i inače emituje čovekov mozak,

our center for control and experience.

centar za kontrolu i opažanje.

Well, it sounds like a pretty good idea,

Jeste, zvuči kao jako lepa ideja,

but this task, as Bruno mentioned,

ali, kao što Bruno reče,

isn't an easy one for two main reasons:

nije je lako ostvariti, iz dva razloga:

First, the detection algorithms.

Prvi su algoritmi za detekciju.

Our brain is made up of

Naš mozak sačinjavaju

billions of active neurons,

milijarde aktivnih neurona,

around 170,000 km

ukupne dužine aksona

of combined axon length.

preko 170.000 km.

When these neurons interact,

Pri radu neurona,

the chemical reaction emits an electrical impulse,

hemijske reakcije u njima prave električne impulse

which can be measured.

koji su merljivi.

The majority of our functional brain

Deo mozga koji vrši nama potrebne funkcije

is distributed over

je raspoređen po

the outer surface layer of the brain,

spoljnoj površini mozga.

and to increase the area that's available for mental capacity,

A radi uvećanja površine te aktivne oblasti

the brain surface is highly folded.

kora mozga je veoma naborana.

Now this cortical folding

Zbog svog tog savijanja i nabiranja

presents a significant challenge

veliki je problem

for interpreting surface electrical impulses.

tačno razabrati električne impulse.

Each individual's cortex

Kora mozga svakog čoveka

is folded differently,

se nabira drugačije, jedinstveno,

very much like a fingerprint.

kao otisci prstiju.

So even though a signal

Mada signali dolaze

may come from the same functional part of the brain,

iz istog moždanog centra,

by the time the structure has been folded,

zbog nabiranja površine mozga

its physical location

lokacije samih centara

is very different between individuals,

vrlo variraju među ljudima,

even identical twins.

čak i kod identičnih blizanaca.

There is no longer any consistency

Tako se gubi poreklo

in the surface signals.

površinskih električnih signala.

Our breakthrough was to create an algorithm

Proboj smo napravili izradom algoritma

that unfolds the cortex,

koji "izravna" površinu velikog mozga,

so that we can map the signals

pa signale sad tačnije mapiramo

closer to its source,

na njihova izvorišta,

and therefore making it capable of working across a mass population.

što nam omogućava masovnu primenu.

The second challenge

Drugi problem je

is the actual device for observing brainwaves.

sam uređaj za hvatanje moždanih talasa.

EEG measurements typically involve

EEG snimci se obično rade

a hairnet with an array of sensors,

sa mrežicom za glavu načičkanom senzorima,

like the one that you can see here in the photo.

kao ova na slici.

A technician will put the electrodes

Tehničar stavlja elektrode

onto the scalp

direktno na površinu glave

using a conductive gel or paste

namazanu provodnim gelom ili namazom,

and usually after a procedure of preparing the scalp

a sve to nakon pripremanja

by light abrasion.

i plitke abrazije kože glave.

Now this is quite time consuming

To sve zahteva dosta vremena

and isn't the most comfortable process.

a nije baš ni prijatno.

And on top of that, these systems

Povrh svega, ovakvi sistemi

actually cost in the tens of thousands of dollars.

koštaju desetihe hiljada dolara.

So with that, I'd like to invite onstage

Sa svim tim na umu, molim vas da pozdravite

Evan Grant, who is one of last year's speakers,

Evana Granta, prošlogodišnjeg govornika,

who's kindly agreed

koji je ljubazno pristao

to help me to demonstrate

da mi pomogne da pokažem

what we've been able to develop.

šta smo napravili.

(Applause)

(Aplauz)

So the device that you see

Uređaj koji vidite je

is a 14-channel, high-fidelity

14-kanalni, vrlo precizni

EEG acquisition system.

EEG snimač.

It doesn't require any scalp preparation,

Ne zahteva prethodnu pripremu glave,

no conductive gel or paste.

nikakav provodni gel.

It only takes a few minutes to put on

Namesti se i kalibriše

and for the signals to settle.

za dva-tri minuta.

It's also wireless,

Takođe je bežičan,

so it gives you the freedom to move around.

i omogućava punu slobodu kretanja.

And compared to the tens of thousands of dollars

Za razliku od dosadašnjih EEG sistema,

for a traditional EEG system,

koji koštaju desetine hiljada dolara,

this headset only costs

ovaj uređaj košta

a few hundred dollars.

nekoliko stotina dolara.

Now on to the detection algorithms.

Nazad na algoritme detekcije.

So facial expressions --

Prepoznavanje izraza lica,

as I mentioned before in emotional experiences --

-- bitno za prepoznavanje emotivnog stanja --

are actually designed to work out of the box

funkcioniše bez prethodnog podešavanja,

with some sensitivity adjustments

a osetljivost se može naknadno

available for personalization.

fino podešavati prema korisniku.

But with the limited time we have available,

Ali, pošto nam je vreme ograničeno,

I'd like to show you the cognitive suite,

prikazaću vam kognitivnu aplikaciju

which is the ability for you

koja daje mogućnost

to basically move virtual objects with your mind.

da umom pomerate virtuelne objekte.

Now, Evan is new to this system,

Evan prvi put koristi sistem,

so what we have to do first

pa prvo moramo

is create a new profile for him.

da mu napravimo lični profil.

He's obviously not Joanne -- so we'll "add user."

Očigledno, on nije "Džoana", dodaću korisnika...

Evan. Okay.

"Evan". OK...

So the first thing we need to do with the cognitive suite

Prvo je nephodno da

is to start with training

kalibrišemo aplikaciju

a neutral signal.

na neutralan signal.

With neutral, there's nothing in particular

Na neutralnom, Evan ne treba da

that Evan needs to do.

radi ništa konkretno.

He just hangs out. He's relaxed.

Treba da je miran i opušten.

And the idea is to establish a baseline

Cilj je da se snimi

or normal state for his brain,

"normalno" stanje njegovog mozga,

because every brain is different.

jer svaki je mozak drugačiji.

It takes eight seconds to do this,

Ovo traje osam sekundi.

and now that that's done,

Sad smo to završili,

we can choose a movement-based action.

možemo da radimo neki pokret.

So Evan, choose something

Evane, izaberi nešto

that you can visualize clearly in your mind.

što možeš jasno da zamisliš.

Evan Grant: Let's do "pull."

Evan Grant: Hajde da "povučemo".

Tan Le: Okay, so let's choose "pull."

Tan Le: Ok, izabrali smo "povlačenje".

So the idea here now

E sad,

is that Evan needs to

Evan bi sada trebalo da

imagine the object coming forward

zamišlja kako povlači objekat

into the screen,

ka ekranu, ka sebi.

and there's a progress bar that will scroll across the screen

Tu je i pokazivač koji će

while he's doing that.

da prati kalibraciju.

The first time, nothing will happen,

Prvi put se neće ništa desiti,

because the system has no idea how he thinks about "pull."

jer sistem ne zna kako Evan misli o "povlačenju".

But maintain that thought

Ali ako zadrži tu misao

for the entire duration of the eight seconds.

svih osam sekundi...

So: one, two, three, go.

Ajmo: jedan, dva, tri, kreni!

Okay.

OK.

So once we accept this,

Nakon što smo uneli komandu,

the cube is live.

kocka može odmah da reaguje.

So let's see if Evan

Hajde da vidimo da li Evan

can actually try and imagine pulling.

može da je povuče.

Ah, good job!

Aaa, odlično!

(Applause)

(aplauz)

That's really amazing.

To je bilo neverovatno.

(Applause)

(tapšu)

So we have a little bit of time available,

Imamo još nešto vremena,

so I'm going to ask Evan

pa ću zamoliti Evana

to do a really difficult task.

da probamo teži zadatak.

And this one is difficult

Ovaj zadatak je teži

because it's all about being able to visualize something

jer treba zamisliti nešto

that doesn't exist in our physical world.

što ne postoji u svakodnevici.

This is "disappear."

To je komanda "nestani".

So what you want to do -- at least with movement-based actions,

Komande vezane za kretanje su lake,

we do that all the time, so you can visualize it.

to radimo i zamišljamo i inače,

But with "disappear," there's really no analogies --

ali za "nestani" nema dobrih analogija.

so Evan, what you want to do here

Dakle, Evane,

is to imagine the cube slowly fading out, okay.

zamisli kocku kako polako nestaje.

Same sort of drill. So: one, two, three, go.

Isti je postupak. Jedan, dva, tri.

Okay. Let's try that.

Ok, sad da probamo.

Oh, my goodness. He's just too good.

O moj bože. Mnogo je dobar.

Let's try that again.

Probaj opet.

EG: Losing concentration.

Evan: Gubim koncentraciju.

(Laughter)

(smeju se)

TL: But we can see that it actually works,

Tan Le: Vidimo da proces počinje,

even though you can only hold it

mada ne možeš dugo da

for a little bit of time.

održiš misao.

As I said, it's a very difficult process

Kažem opet, jako je teško

to imagine this.

zamišljati "nestajanje".

And the great thing about it is that

Ono što je bitno je

we've only given the software one instance

da smo programu samo jednom dali

of how he thinks about "disappear."

kako Evan misli "nestani".

As there is a machine learning algorithm in this --

Program je u stanju da uči --

(Applause)

(aplauz)

Thank you.

Hvala.

Good job. Good job.

Bravo, bravo.

(Applause)

(tapšu)

Thank you, Evan, you're a wonderful, wonderful

Hvala ti, Evane, divno si,

example of the technology.

divno prikazao ovu tehnologiju.

So, as you can see, before,

Kao što vidite,

there is a leveling system built into this software

postoji više nivoa i komandi u progamu,

so that as Evan, or any user,

tako da Evan, i bilo koji drugi korisnik,

becomes more familiar with the system,

kad se privikne na sistem,

they can continue to add more and more detections,

može da dodaje sve više i više obrazaca,

so that the system begins to differentiate

te će sistem biti u stanju da razlikuje

between different distinct thoughts.

više različitih pojedinačnih misli.

And once you've trained up the detections,

A jednom kad se potpuno izvežbate,

these thoughts can be assigned or mapped

te komande se mogu dodeliti ili prebaciti

to any computing platform,

drugim kompjuterskim platformama,

application or device.

programima ili uređajima.

So I'd like to show you a few examples,

Prikazaću vam samo nekoliko primera,

because there are many possible applications

jer ima puno mogućih primena

for this new interface.

za ovaj novi interfejs.

In games and virtual worlds, for example,

U igrama i virtuelnim svetovima,

your facial expressions

vaš izraz lica

can naturally and intuitively be used

se može prirodno i interaktivno koristiti

to control an avatar or virtual character.

da upravljate avatarom ili virtuelnim likom.

Obviously, you can experience the fantasy of magic

Možete iskusiti fantastiku magije

and control the world with your mind.

i kontrolisati svet svojim mislima.

And also, colors, lighting,

Boje, osvetljenje,

sound and effects

zvuk i efekti,

can dynamically respond to your emotional state

sve može dinamički da reaguje na vaše raspoloženje

to heighten the experience that you're having, in real time.

da dodatno pojača doživljaj.

And moving on to some applications

Evo nekih primena

developed by developers and researchers around the world,

razvijanih širom sveta

with robots and simple machines, for example --

sa robotima i jednostavnim uređajima.

in this case, flying a toy helicopter

Ovo je mali helikopter,

simply by thinking "lift" with your mind.

koji poleće na misaonu komandu "poleti".

The technology can also be applied

Ista tehnologija se može primeniti

to real world applications --

u svakodnevnom životu,

in this example, a smart home.

recimo u "pametnoj kući".

You know, from the user interface of the control system

Iz interfejsa kontrolog sistema možete

to opening curtains

razvući zavese,

or closing curtains.

navući zavese.

And of course, also to the lighting --

Naravno, osvetljenje.

turning them on

Možete paliti

or off.

ili gasiti svetla.

And finally,

I najzad,

to real life-changing applications,

primene koje menjaju živote ljudima,

such as being able to control an electric wheelchair.

poput mogućnosti da se upravlja električnim kolicima.

In this example,

U ovom primeru,

facial expressions are mapped to the movement commands.

izrazi lica se povezuju na kontrole kretanja.

Man: Now blink right to go right.

"Sad namigni desnim da skreneš desno."

Now blink left to turn back left.

"Sad levim, da skreneš levo."

Now smile to go straight.

"Sad se smeši da ideš pravo."

TL: We really -- Thank you.

TL: Hvala...

(Applause)

(aplauz)

We are really only scratching the surface of what is possible today,

Stvarno smo tek na početku shvatanja šta je sve sa ovim moguće.

and with the community's input,

A sa uplivom zajednice,

and also with the involvement of developers

sa tehničkom saradnjom,

and researchers from around the world,

sa istraživačima širom sveta,

we hope that you can help us to shape

nadamo se da ćete nam pomoći da

where the technology goes from here. Thank you so much.

oblikujemo budućnost ove tehnologije. Hvala vam puno.