Original subtitles

"You are my creator, but I am your master ..."

What we are on the eve

It is a world with an increasingly intense and sophisticated

artificial intelligence.

The technology is evolving much faster than our society.

She has the ability to protect us as citizens.

The robots are coming and they will destroy

our livelihood.

Do you have a network intelligence that watches us,

You know everything about us and start trying to move on.

Twitter has become the news website world number one.

Technology is not good or bad.

It's what we do with the technology.

In the end, millions of people will be forced from their jobs

because their skills will be obsolete.

mass unemployment.

Greater inequality, even social unrest.

Regardless of being afraid or not,

change is coming and no one can stop it.

We invest huge amounts of money

and so it is logical that the military,

with their own desires will start to use these technologies.

Systems of autonomous weapons

would lead to a global arms race

that would rival the nuclear age.

So you know what the answer is, they will eventually kill us.

These technological leaps will give us

incredible miracles

and incredible horrors.

We created it,

so I think as we move forward with this intelligence,

it will contain parts of us.

But I think the question is,

it will contain the good parts

or the bad parts?

YOU RELY ON THIS COMPUTER?

The survivors called war

of Judgment Day,

they lived only to face a new nightmare,

the war against the machines.

I think we ferramos altogether.

I think Hollywood could inoculate

the general public against this issue,

the idea that a machine will dominate the world.

Open the bay doors, HAL.

I'm sorry, Dave.

I can not do this.

HAL?

Already alarmed without reason often

that the public stopped paying attention

because it sounds like science fiction,

even sitting here talking about it.

Even now, it seems a bit silly,

a bit like "oh !, this is something of a film class B".

The WOPR spends all his time thinking

on the Third World War.

But it is not.

The general public is about to be caught off guard by this.

As individuals, we are increasingly surrounded

by a machine intelligence.

We carry this pocket device in the palm of your hand

we used to take an impressive series of life decisions.

Aided by a set of distant algorithms

on which we do not understand anything.

We are already tired of the idea

that we can speak to our phone

and mostly he understands us.

I found a lot of action movies.

Five years ago at all.

robotic machines that see, speak and hear,

all this is now real and these technologies

They will fundamentally change our society.

Now we have this great movement of cars that drive themselves.

Driving a car autonomously can move people's lives

to a better place.

I lost several members of my family, including my mother,

my brother and sister and their children

to car accidents.

It's pretty clear.

We can almost eliminate car accidents with automation.

30,000 lives in the US alone,

about a million worldwide each year.

In health, the early indicators are the name of the game in this space,

so this is another place where you can save someone's life.

Here in breast cancer center

all things that the brain radiologist makes in two minutes

the computer does instantly.

The computer analyzed 1 billion mammograms

and he took the data and applied them in this image instantly.

So its medical application is profound.

Another exciting area where we are seeing a lot of development

It is that you understand our genetic code

and using it both to diagnose diseases

how to create customized treatments.

The main application of all these machines

will be to expand our own intelligence.

We can become smarter,

and will be better to solve problems.

We will not have to grow old,

we will not have to go through aging.

We will be able to stop it.

There really limit to what intelligent machines

can do to the human race.

As a more intelligent machine could not be a better machine?

It's hard to say exactly when I started thinking

it was a bit naive.

Stuart Russell is basically a god

in the field of artificial intelligence.

He wrote the book that almost all universities use.

I used to say it is the best-selling AI book.

Now I say that is the most stolen PDF.

Artificial intelligence is about making

intelligent computers,

and the public's point of view, what counts as AI

It is something that is surprisingly clever

compared to what we think computers

They are normally capable of doing.

AI is a field of research

which basically tries to simulate

all kinds of human capabilities.

We are in an era of IA.

Silicon Valley had the ability to focus on something brilliant.

Were social networks in social media in the last decade

and it is quite clear that the currency turned.

It starts with learning machine.

When we look back at this time,

which was the first AI?

It is not exciting, is not the thing we see in the movies,

but you would make a great hunch

Google did not create a search engine,

but a deity.

One way people do any question they want

and get the answer they need.

Most people are not aware that

what Google does is a form of artificial intelligence.

They just go there, type in something

and Google gives the answer.

With each search we train us to be better.

Sometimes we type and research gives us the answer

before you finish asking the question.

Who is the president of Kazakhstan?

And he will tell you,

you do not need to go to the website of Kazakhstan to find out.

He used to be able to do this.

This is artificial intelligence.

In a few years, when we try to understand,

shall we say, well, how we lost it?

It is one of those remarkable contradictions

we are facing.

Google and Facebook have absolutely

built business by giving us as a society

one free stuff, free of charge,

but this is a Faustian bargain.

They are taking something from us in return.

But we do not know what code is running on the other side,

and we have no idea.

This beats and the question of how much we should trust the machines.

I use computers for everything literally.

There are so many advances in computer now

and it has become such a big part of our lives.

It is simply amazing what a computer can do.

You can actually load a computer in your bag.

I mean, this is incredible.

I think most technologies

They are made to make things easier for all of us.

So hopefully this will continue to be the focus.

I think everyone loves their computers.

People do not realize

They are constantly negotiated by machines.

If this is the price of the products in your shopping cart from Amazon,

if you can get on a particular flight,

if you can book a room in a specific hotel,

what you're experiencing

are machine learning algorithms

which determined that a person like you

You are willing to pay two cents and changing the price.

Now the computer looks to millions of people simultaneously

for very subtle patterns.

You can take seemingly innocent digital tracks,

as someone's playlist on Spotify

or things that they bought on Amazon,

and then use algorithms to translate this

for an intimate profile very detailed and very accurate.

There is a dossier on each of us,

so extensive it would be possible to say

they know more about you than your mother.

The main cause of the recent advancement of AI

It is not just that some guy suddenly had a brilliant vision,

but simply because we have much more data

to train them and immensely better computers.

The magic is in the data.

It's a ton of data.

I mean, they are data that never existed before.

We never had that data before.

We create technologies that allow us

capture large amounts of information.

If you think of a billion mobile phones on the planet

with gyroscopes, accelerometers,

fingerprint readers coupled with GPS

and the photos they take and the tweets that you send,

we are all giving enormous

amounts of data individually.

Cars go while their cameras suck information

about the world around them.

Satellites are now in orbit with the size of a toaster.

Infrared on the vegetation of the planet.

The buoys that are feeding in the ocean climate models.

And the NSA, the CIA, as they collect information

on geopolitical situations.

The world today is literally swimming in these data.

Going back to 2012,

IBM estimated that an average human left

500 megabytes of digital tracks every day.

If you want to back up only one day of data

that mankind produces

and you from printing on letter size paper,

sided, font size 12,

give to an Earth the cell surface to the sun ...

four times.

This every day.

The data itself is not good or bad.

It is how they are used.

We are really relying on the good will of these people

and policies of these companies.

There is no legal requirement as they

They can or should use this data.

That, to me, is the confidence question of heart.

Right now there is a huge rush to create machines

that they are as intelligent as humans.

Google they are working on what is

a kind of Manhattan Project of artificial intelligence.

They have a lot of money.

They have many talents.

They are buying companies AI and robotics companies.

People still think that Google is a search engine,

your e-mail provider,

and many other things that we use daily,

but behind this search box

are 10 million servers.

This makes Google the most powerful computing platform

of the world.

Google is now working on an AI computing platform

You will have one hundred million servers.

So when you're interacting with Google,

see the nail is a giant beast in the process of formation,

and the truth is,

I'm not sure if Google knows what he's becoming.

If you look inside the algorithms being used in Google,

the technology is largely the 80s.

So, these are models that you teach,

showing one, two and three,

and he did not learn what the one, or what is the two,

he learns what is the difference between one and two.

It's just a computer.

In the last half decade, which made this rapid progress,

everything has happened in pattern recognition.

Most of the good old IA

It was when we said to our computers

how to play a game like chess.

The old paradigm where you say to a computer

exactly what to do.

This is Jeopardy,

The challenge of IBM!

No one at the time thought that a machine

could have the accuracy, reliability and speed

to play well enough Jeopardy

against the best specialists.

Let's play Jeopardy.

Four letter word

to put an iron in the hoof of a horse.

Watson?

Shoe (screwed)?

You're right, you can choose.

literary character APB to 800.

Response to Daily Double.

Watson actually got its knowledge by reading Wikipedia

and 200 million pages of documents in natural language.

You can not program all lines of how the world works.

The machine has to learn reading.

We come to Watson, who is Bram Stoker and the bet?

Hello, 17,973.

41,413 and a total of two days.

Watson was trained in a lot of texts

but it's not like him to understand what he is saying.

He does not know that the water gets wet things

but seeing the way things behave in the world

as you and I do.

Much IA language of today is not building

logical models of how the world works,

You are looking at how the words appear in context

in other words.

David Ferrucci developed the IBM Watson

and someone asked him - "Watson thinks?".

And he said - "a submarine anything?"

And what he meant was that when

they developed the submarines,

they took the basic principles borrowed

of fish swimming,

but a submarine nothing much faster than the fish.

Even with a huge mass, it overcomes the fish.

Watson winning the game Jeopardy

IA will enter history as a significant milestone.

We tend to be surprised when a machine works so well.

I'm even more surprised when a computer beats humans

on things that humans are naturally good.

This is how we make progress.

In the early days of Google Brain Project

I gave the team a very simple instruction

which was building the largest possible neuro network

something like a thousand computers.

A neural network is something very close to a simulation

how the brain works.

It is very probabilistic,

but with contextual relevance.

In your brain you have long neurons that connect

to thousands of other neurons and you have these paths

which are formed and forged

based on what the brain needs to do.

When a baby tries something and can,

there is a reward

and this path that created success is strengthened.

If you fail at something, the way is weakened

and so over time the brain becomes sharp

to be good in the environment around you.

Leave the machines to learn for themselves

is called deep learning,

and profound learning and neural networks

roughly mean the same thing.

Deep learning is a totally different approach,

where the computer learns more like a child

getting a lot of data and end discovering things.

The computer gets smarter and intelligent

as they have more experience.

Let's imagine that you have a neural network,

something like a thousand computers,

and she does not know anything,

and we do it to see YouTube for a week.

Oppa Gangnam Style.

Charlie, hurt!

And after seeing the YouTube for a week,

what she learned?

We had a chance she would learn

to detect common objects in videos.

and we know that human faces appear much in the videos,

so we went to check and behold, a neuron

who learned to detect a human face.

Leave Britney alone!

What else appears both in the videos?

Then we found to our surprise

that there was indeed a neuron

and he learned to detect cats.

I still remember seeing the recognition, wow, it's a cat,

Cool.

It's all very harmless when you think about the future,

everything seems kind of harmless and benign,

but we are creating cognitive architectures

that will fly farther and faster than we

and will have a greater payload,

and they will not be warm and cute.

I think that in three to five years,

see computer systems

that will be able to learn autonomously

how to understand, how to build an understanding.

It is not different from the way the human mind works.

Whatever lunch was certainly delicious.

It was just the work of Robbie Synthetics.

It is also your cook?

He even makes the raw materials.

A back Robbie.

I'll show you how it works.

Someone introduces a sample of human food

through this opening.

Down here, there is a small built-in chemical laboratory

where he analyzes the food.

Later he can reproduce identical molecules

in any form or quantity.

The dream of a spy.

Meet Baxter, the revolutionary new

robots category with common sense.

Baxter is a good example of the type of competition

we face with the machines.

Baxter can do almost anything we can do with our hands.

Baxter costs what a minimum wage worker

earns in a year,

but Baxter will not take place

a minimum wage worker.

He will take the place of three because he never gets tired,

he never pauses.

This is probably the first thing we see.

The shift of jobs.

The work will be done faster

and at a lower cost by machines.

Our ability to stay current

It is so insanely limited

compared to the machines we build.

For example, we now have this great movement

Uber and Lyft who are making transportation

cheaper and democratizing, which is great.

The next step will they be replaced

by driverless cars and then

all drivers of Uber and Lyft

They will have to find something new to do.

There are four million professional drivers in the US,

they will soon be unemployed.

Seven million people make data entry,

these people will be unemployed.

A job is not just about money.

At a biological level, it serves a purpose,

It becomes a defining thing.

When jobs go away in any civilization,

It does not take long until it turns into violence.

We face a giant divide between rich and poor

because that's what the automation and AI will cause

or greater division between those who have and those who have not.

Now it's happening in the middle class,

in white-collar jobs.

IBM's Watson does business analysis we usually pay

$ 300 per hour for a business analyst to do.

Today you go to college to be a doctor,

be an accountant, being a journalist.

It is not clear that there will be jobs for you.

If someone plans a 40-year career in radiology,

just reading images,

I think this could be a challenge

for new graduates today.

Today we will make a robotic procedure.

The Da Vinci robot is currently used

by a variety of surgeons

for its accuracy and its ability

to avoid the inevitable fluctuations of the human hand.

Anyone who watches it feel with him is amazing.

You look through the panel,

and you see the claw holding the ovaries of women.

Humanity right there resting in the hands of this robot.

People say that he is the future,

but he is not the future, it is the present.

If you think of a surgical robot,

often there is not much intelligence in such things,

but over time,

We put more and more intelligence into these systems,

surgical robots can actually learn

with each robotic surgery.

They are following the movements.

They are understanding what worked and what did not work

and eventually the robot for routine surgery

You will be able to perform them alone

or with human supervision.

Typically we do about 150 hysterectomy procedures,

and now most of them are done robotically.

I do maybe an open procedure per year.

So, I feel uncomfortable? Certainly I feel,

because I do not remember how to open patients.

It seems to be feeding and creating it,

but somehow we are slaves to technology

because we can not go back.

The machines are giving increasing bites

in our skill set

and an increasing speed

and then we have to run faster and faster

to keep us ahead of the machines.

How am I?

Bonita.

Are you attracted to me?

What?

Are you attracted to me?

You give me information that is.

I give?

Sim.

This is the future to which we address.

We want to design our companions,

we like to see a human face in AI,

so play with our emotions will depressivelmente easy.

We are not so complicated.

We are simple responses to stimuli.

I can make you like me

basically they are smiling a lot for you.

Yes. The IAs will be fantastic to manipulate us.

So you developed a technology that can feel

what people are feeling?

Sure, we have developed a technology that can read

their facial expressions and map

in a number of emotional states.

Fifteen years ago I had just finished

My graduate studies in computer science,

and it occurred to me that I was spending too much time

interacting with my laptop, with my devices,

although these have absolutely no clue devices

how I was feeling.

I began to think, and if this device could feel

I was stressed or was having a bad day,

what it could become affordable?

Hello my first graders, how are you?

Can I have a hug?

We had children interacting with technology.

Much of this is still in development,

but it was just fantastic.

Those who like robots?

Who wants to have a robot at home?

So you would use a robot?

I would use it to make my mother

difficult questions of mathematics.

Fine and you?

I would use it to scare people.

All right, then start to smile.

Well, her brow furrowed.

Well, with a raised eyebrow.

This generation of technology is just around them all the time.

It's almost as if they expect to have robots in their homes,

and they expect these robots are socially intelligent.

What makes the robots are intelligent?

Put them in a math class or biology.

I think you would have to train him.

Okay, let's walk down here.

So if you smile, and raise eyebrows

it will for you,

but if you look angry, he'll run away.

We are selling computers

to read and recognize emotions

and the response so far has been really amazing.

People are integrating it in health applications,

Meditation apps, robots, cars.

Let's see how this plays out.

The robots may contain Ia

but the robot is the physical representation,

and artificial intelligence is the brain,

brain and thus may exist purely

in software-based systems.

They need not have a physical form.

Robots can exist without any artificial intelligence

we have many robots idiots out there,

but an idiot robot can be an intelligent robot overnight

having the right software, with the right sensors.

We can only ascribe reason to inanimate objects.

We do this with the machines.

We will treat them like children.

We will treat them as substitutes

and we will pay the price.

Welcome to ATR.

My purpose is to have an anthropomorphic robot

that has intentions and anthropomorphic desires.

The robot's name is Erica.

Erica is the most advanced anthropomorphic robot in the world, I think.

Erica can contemplate his face.

Konnichiwa.

Robots may be good partners with,

especially for the elderly, children, disabled ...

When we talk to the robot, we have no social barriers, social pressures.

Finally, everyone accepts the android as our friend, our partner.

We implemented a simple desire. She wants to be well recognized, and want to take risks.

If a robot could have unintended desires,

the robot could understand the intentional desires of others.

What kind of animal do you like?

I like dogs Shiba Inu. They are very cute, are not they?

This means stronger relationships with people,

and that they could enjoy each other ...

This means, I'm not sure ... love one another?

Built artificial intelligence

and the first thing we want to do is to replicate.

I think the key point will come when all

the main directions are replicated,

the vision,

the touch,

the smell.

When we replicate our senses

It is when it becomes alive?

Many of our machines are being built to understand the

but what about an anthropomorphic creature

who discovers that they can adjust their loyalty,

adjust your courage, adjust your greed,

adjust your cunning?

The average person does not see killer robots down the streets,

they are like what are you talking about?

Man, we want to make sure we will not have

killer robots down the street.

Once they go down the street, it will be too late.

The thing that worries me now,

that keeps me awake,

It is the development of autonomous weapons.

So far people have expressed discomfort

about the drones that are piloted aircraft remotely.

If you take the camera from a drone, feed her on an IA system

It will be a very easy step to get

to fully autonomous weapons to choose their own targets

and release its own missiles.

The life expectancy of a human being

this kind of battlefield environment

It will be measured in seconds.

In a moment the drones were science fiction

and now they have become a normal thing in the war.

More than 10,000 only in the US military inventory,

but they are not just a US phenomenon.

There are over 80 countries that operate.

It seems logical that people who take some

the most important decisions and difficult in the world,

They will start to use and implement artificial intelligence.

The Air Force has just design a jet program

400 billion dollars to put pilots in the sky

and a $ 500 AI designed by a couple

graduate students

beat the best human pilots

with a relatively simple algorithm.

The IA will have a big impact on the military

as well as the combustion engine was the turn of the century.

This will literally touch everything the military does,

from train without a driver delivering logistical supplies,

the unmanned drones that provide medical,

computational advertising trying to conquer

hearts and minds of the population

and so it is logical that those who have the best IA

probably achieve dominance on this planet.

Sometime in the early twenty-first century,

all mankind was united in celebration.

We were amazed at our own magnificence

We gave birth when AI.

HE?

You mean artificial intelligence?

A singular consciousness that spawned

whole machine race.

We do not know who struck first, us or them,

but we know that we who burn the heavens.

There is a long history of science fiction,

not only predicting the future,

but shaping the future.

Arthur Conan Doyle wrote before the First World War

about the dangers of how submarines

They could be used to hold civil locks.

At the time he wrote this fiction,

the Royal Navy made fun of Arthur Conan Doyle

to have this absurd idea that the submarine

They could be useful in the war.

One of the things we see in history

It is that our attitude towards technology,

as well as ethics, rely heavily on the context.

For example, submarine nations like Britain

and even the United States

found horrifying use the submarine.

In fact, the German submarine use to carry out attacks

was the reason why the United States joined World War I,

but the timeline to move forward.

The United States of America was suddenly and deliberately

attacked by the Empire of Japan.

Five hours after Pearl Harbor,

came the order to the unrestricted use

the submarine war against Japan.

Then Arthur Conan Doyle was right.

This is the great and ancient line of science fiction,

is a lie that tells the truth.

Dear fellow executives, I have the great pleasure

to introduce them to the future of law enforcement,

Ed 209.

This is not just a matter of science fiction.

This is about what's next on what's happening now.

The role of intelligent systems

It is growing very rapidly in the war.

Everyone is moving in unmanned kingdom.

Today Secretary of Defense was very clear

that we will not create fully autonomous attack vehicles.

Not everyone is going to stick to the same set of values

and when China and Russia began to deploy autonomous vehicles

that can attack and kill,

what is the position that we adopt?

You can not say, well, let's use autonomous weapons

to our military dominance

but no one will use them.

If you do these weapons,

they will be used to attack

human populations in large numbers.

autonomous weapons that by their nature,

are weapons of mass destruction

because they do not need a human being

to guide them or charge them.

You only need one person

to write a small program.

It only captures the complexity of this field is legal.

Is important.

That's incredible.

It's also scary and it's all about trust.

It is an open letter on artificial intelligence

signed by some of the biggest names in science.

What they want?

Prohibit the use of autonomous weapons.

The author said ...

"The autonomous weapons are described as

the third revolution in war activities "

Thousand experts in artificial intelligence

They call for a global ban of killer robots.

This open letter basically says that we should reset

The goal of the field of artificial intelligence,

far from just create pure undirected intelligence

for creating a beneficial intelligence.

The development of AI will not stop,

will continue and will improve,

if the international community does not put certain controls

this personal things develop

they can do anything.

The letter says we are years, not decades

these weapons are deployed.

We had six thousand signatories to this letter,

including many of the leading figures in the field.

I'm getting a lot of high-level visits of officials

who wish to emphasize that dominance

the US military is very important

and autonomous weapons could be part

the Department of Defense plan.

This is very scary because the value system

military technology developers

It is not the same value system of the human race.

Aside from concerns about the possibility that this technology

can be a threat to human existence,

some technologists finance the Future of Life Institute

to try to deal with these problems.

All these guys are reserved,

and so it is interesting for me to see them all together.

All we have is the result of our intelligence,

It is not the result of our big scary teeth,

or our large claws or our huge muscles.

It is because we are relatively intelligent

and between my generation, we all have what we call

moments of "sacred cow" or "something sacred"

because we see that the technology

It is accelerating faster than expected.

I remember sitting around the table

with some of the best and brightest minds in the world

and what really impressed me

It was that the human brain may not be able

to fully understand the complexity of the world

that we face.

As currently constructed,

the road that AI is following veered off a cliff

and we need to change the direction we take

not to launch the human race from the cliff.

Google acquired DeepMind several years ago.

The DeepMind operates as a subsidiary

semi independent of Google.

What makes the only DeepMind

is that is absolutely focused DeepMind

in creating a digital super intelligence,

an AI that is much smarter than any human on Earth

and, ultimately,

smarter than all humans on Earth combined.

This is the reinforcement learning system DeepMind

which basically wakes up as a newborn baby

who is shown the screen of a video game Atari

and then you have to learn to play the video game.

She does not know anything about the objects on the movements, over time,

she only knows that there is an image on the screen

and there is a score.

So if your baby woke up on the day you were born

and in the late afternoon he was already playing

40 different Atari video games in a superhuman level,

you would be terrified.

You say "my baby is possessed, send him back"

The DeepMind system can win any game.

He can beat all the original Atari games.

It is superhuman,

he plays at super speed, in less than a minute.

The DeepMind turned to another challenge

and this challenge was the Go game,

people generally argued

It is beyond the power of computers

play against the best human players.

First they defied the European champion of Go.

So they challenged the Korean champion Go.

And they were able to win both times

in a kind of amazing style.

You read articles in The New York Times years ago

talking about how the Go take one hundred years

for us to solve.

People said ...

"Well, but that's just a board.

Poker is an art,

poker involves reading people,

It involves lying, bluffing.

It's not an exact thing and will never be possible

a computer can do this "

They took the world's best poker players

and it took seven days to the computer

begin to demolish humans.

So, the best player in the world of poker,

the best Go player in the world,

and the pattern here is that AI may take a while

to throw its tentacles around a new skill,

but when it happens, when it can,

she is unstoppable.

The AI ​​DeepMind have administrator-level access

to Google servers to optimize energy use

in data centers.

However, this could be an unintended Trojan.

The DeepMind have to have complete control of data centers

so, with a small software update,

AI could have full control over all of the Google system,

meaning that she could do anything.

She could look at all of your data,

she could do anything.

We are rapidly going to the super digital intelligence,

which will greatly exceed any human being.

I think this is very obvious.

The problem is that we suddenly

We not reached the human level of intelligence and say ...

"Okay, let's stop searching."

It goes beyond the human level intelligence,

and do what is called super intelligence,

and that's something smarter than us.

The IA level superhuman, if we succeed with this,

It is by far the most powerful invention ever done

and the last invention that we do,

and create AI that is more intelligent than us,

we have to be open to the possibility that

we may actually lose control of it.

Let's assume that you have given her a goal,

as if you were with cancer,

and then discovers that the way she chose

to handle it

conflicts with many things that you care about.

The AI ​​does not have to be bad to destroy humanity.

If the AI ​​has a goal

and humanity is simply on its way,

it will destroy humanity as a natural thing,

without even thinking about it.

There are no hard feelings.

It is as if we were building a road

and an ant is on the way.

We do not hate the ants, we are just building a road,

and even then goodbye tingling.

It is tempting to dismiss these concerns

because it's like something that can happen

in a few decades or 100 years hence,

so why bother?

But if you go back to September 11, 1933,

Ernest Rutherford, who is a nuclear physicist

best known for his time,

said the possibility of ever extracting useful amounts

Power transmutation of atoms, as she was called,

It was a fantasy.

The next morning, Leo Szilard,

it was a much younger physical

You read it and became very troubled

and found how to make a chain nuclear reaction

just a few months later.

Spent more than two billion dollars

the greatest scientific gamble in history.

So when people say,

"Oh, that is so far in the future,

we need not worry about it, "

we may be only three or four discoveries of this magnitude

that will lead us out where we are

for super intelligent machines.

If it takes 20 years to find out how to maintain beneficial IA

then we must start today,

not at the last second when some guys drinking Red Bull

decide to trigger the switch and test the thing.

We have five years, I think,

for super digital intelligence happen.

In my lifetime, one hundred percent.

When this happens, she will be surrounded

by a lot of people really excited about this technology.

They will want to see it succeed,

but they will not tell you that they can get out of control.

Oh my God, I trust both on my computer.

This is an amazing question.

I do not trust my computer if it is on.

I turn off.

Even with it off, I still think he is on,

it's like you do not like the webcams.

You do not know if anyone could turn it on.

I do not trust my computer

as well as on my phone,

every time they ask me to send information

for Apple, so ...

I do not trust my cell.

Okay, so part of that is yes.

I trust him because it will actually be very difficult to spend the day

and how our world is set up without computers.

Trust is a very human experience.

I have a patient with intracranial aneurysm.

They want to look in my eyes and know

that they can trust that person with your life.

I'm not terribly worried about anything.

Part of this is because I have confidence in you.

This procedure we going to do today

20 years ago it was essentially impossible.

We just did not have the materials or technologies.

Let's go to that corner!

It could be more difficult? My God!

The coil is almost there.

It's just a shame holding it.

Are nervous times.

We are in purgatory.

The humanist intellectual purgatory.

An IA would know exactly what to do here.

We have a coil in the aneurysm

but she was not tremendously well

as I knew it would be,

and a risk of perhaps 20% is a very bad situation,

I chose to bring it back.

Because of my relationship with her

and knowing the difficulties to come in and do the procedure

I only consider the safest way possible

to have success,

and I stayed there for 10 minutes agonizing over it.

The computer does not feel anything.

The computer just does what it should do,

supposedly better and better.

I want to have the AI ​​in this case,

but the AI ​​can have compassion?

It is the question that everyone asks about IA.

We are the only personification of humanity

and it is a remote possibility we accept that a machine

can be compassionate and loving that way.

Part of me does not believe in magic

but part of me has faith that there is something

in addition to the sum of the parts, if at least

a unit in our shared ancestry,

our shared biology, our shared history.

Some connection beyond the machine.

Then you have the other side of it is that ...

the computer knows that he is aware

or may be conscious or cares?

It needs to be conscious?

He needs to be aware of?

I do not think a robot could be conscious.

Unless they to program that way.

Aware, no.

No, no.

I think a robot could be programmed to be aware.

Just as they do the programming for all the rest.

This is another big part of our intelligence officer,

make them aware and make them feel.

Back in 2005 we started trying to build machines

with self-awareness.

This robot at first did not know what it was,

all he knew was that he needed to do something like walking.

Through trial and error,

he learned how to walk using your imagination

and then he walked away.

And then we did something very cruel.

We cut one of his leg and watched what happened.

At first he was not sure what had happened

but after the period of a day

he then began to limp,

and then, a year ago,

We were training an AI system for a live demonstration.

We wanted to show all kinds of objects in front of camera

and how AI could recognize these objects.

And so we were preparing this demo

and we had a screen to share with the ability to see

and certain neurons were responding

and suddenly noticed that one of neurons

He was tracking the faces.

We were tracking our faces as we moved.

Now, this is the scariest ever trained

the system to recognize human faces

and yet, somehow, he learned to do that.

Even if these robots are very simple,

we can see that there is something else going on there.

Not only is the program.

So that's just the beginning.

I often think about that beach in Kitty Hawk,

no VOO the 1903 e Orville Wilbur Wright,

as a kind of claim canvas, wood and iron

to get off the ground for 1 minute and 20 seconds,

on a windy day before heading back down again,

and about 65 summers

after that time

you have a 747 taking off from JFK.

The biggest concern of someone in the plane could be

if your diet meal without salt

will be served or not,

with a whole infrastructure with travel agents

and control tower, and it's all casual,

it's all part of the world.

Well now, as they have machines that think,

and solve problems, we are back to Kitty Hawk.

We are in the wind, have our tattered canvas planes in the air,

but what will happen in 65 or more summers,

and we have machines that are beyond human control.

Should we worry about it?

Not sure if they will help.

No one has any idea now what it means for a robot

to be aware.

There is no such thing.

There are a lot of smart people

and I have great respect for them,

but the truth is that the machines are natural psychopaths.

Fear returned to the market

and it was down 800 nearly 1,000 in the blink of an eye.

It is the classical rendition.

There are some people who think

that was some sort of error-finger fat.

Take the Flash Crash of 2010

where in a matter of minutes a trillion dollars

It was lost in the stock market.

The Dow Jones fell nearly a thousand points in half an hour.

So what went wrong?

At that time more than 60% of all business

carried out on the stock exchange

They were actually being run by computers.

A brief history of what happened in the Flash Crash

It is that algorithms answered algorithms

and they set about themselves increasingly

and all in a matter of minutes.

At one point, the market fell like that down a well.

There is no governing body that can adapt quickly enough

to avoid the potentially disastrous consequences

AI operating in our financial system.

They are as fundamental to handle.

Let's talk about the speed with which

we sell this market deteriorate.

That's the kind of AI that drives people crazy.

When you give them a goal,

they relentlessly pursue that goal.

How many computer programs are like this?

Nobody knows.

One of the fascinating aspects of AI in general

It is that no one really understands how it works.

Even people who create AI do not fully understand.

Because she has millions of elements,

It becomes quite impossible for a human being

understand what is happening.

Microsoft has created this artificial intelligence

called Tay on Twitter that was a conversation robot.

They started the morning

and Tay was beginning to tweet and learn things

who were being sent to him from other people on Twitter.

Because some people made offensive attacks,

within 24 hours Microsoft Robot

It became a terrible person.

They had to literally take the Tay network

because it has turned into a monster.

A person misanthropic, racist and horrible

you would never want to know

and no one predicted this.

The whole idea of ​​AI is that we are not saying

exactly her how to achieve a particular result

or a goal.

AI develops on its own.

We are concerned with the super intelligent AI,

the chess master that we will overcome,

but I will not have to be really so smart

to be massively disruptive effects

on human civilization.

We saw in the last century,

someone does not have to be brilliant

to cast the story in a particular direction

and you do not need a brilliant artificial intelligence

to do the same thing.

Stories of false election generated more news

engagement on Facebook than the main real stories.

Facebook is really the elephant in the room.

An AI runs the news feed Facebook.

The task of the IA is to keep users engaged,

but no one really understands exactly

as this IA is reaching this goal.

Facebook is building an elegant mirrored wall

around us.

A mirror to who we can ask who is the fairest of all

and he will answer "you, you" again and again.

You slowly begins to distort our sense of reality,

distort our political sense,

history, global events,

up to determine what is true and what is not true

It is virtually impossible.

The problem is that the AI ​​does not understand this.

IA had only one task,

maximize user engagement,

and get it.

Almost two billion people

spend on average nearly an hour today

basically interacting with AI

that is shaping your experience.

Even the engineers from Facebook, do not like false news.

It is a very bad deal.

They want to get rid of false news.

It is very difficult to do because as you recognize

if the news is false if you can not read

all this news in person?

There is so much misinformation active

and it is well packed,

it appears even when you see

a Facebook page or you turn on your television.

It is not terribly sophisticated,

but it is terribly powerful.

And what that means is that your view of the world,

that 20 years ago it was determined when you watched

the evening news in three different television networks.

The three anchors struggled to hit

They could have had a small bias time or another

but roughly speaking,

all agree on an objective reality.

This objectivity is over

eo Facebook is completely annihilated.

If most of your understanding of how the world works

comes from Facebook,

facilitated by algorithmic software that attempts to show

the news you want to see,

this is a terribly dangerous thing

and the idea that we do not just put it in motion

but allow actors in bad faith to access this information,

this is the recipe for disaster.

I think definitely we will have many bad actors

trying to manipulate the world with the IA.

2016 was a perfect example of an election

where there was much production of IA,

many false reports and distributed for a purpose,

for a result.

Ladies and gentlemen, honored colleagues,

I have the privilege to speak to you today

about the power of Big Data and psychographics in the electoral process

and specifically to talk about work

that contributed to the presidential campaign of Senator Cruz.

The Cambridge Analytica appeared silently as a company

that according to its own description

It has the ability to use this huge amount of data

in order to affect social change.

In 2016 they had three major customers,

Ted Cruz was one of them.

It is easy to forget that just 18 months ago,

Senator Cruz was one of the least popular candidates

looking statement.

So it was not possible to maybe 10 or 15 years ago,

you can send false information just for people

you wanted to send

and then see how he or she reacted on Facebook

and then adjust this information

according to the feedback you received

and thus be able to start developing something like

one real-time management of a population.

In this case, we separate a group we call persuasion.

These are people who will vote for a political convention

but they need to move from the center

for a little more right

in order to support Cross.

They need a persuasive message.

I selected the right weapons to narrow the field a bit more

and now we know we need a message about the right weapons,

to be a persuasive message

and needs to be nuanced according to certain personality

we are interested.

Through social media,

there is an infinite amount of information

you can gather about a person.

We have somewhere around 4 or 5 000 data points

of all adults in the United States.

It is targeting the individual.

It's like a weapon that can be used

in a totally wrong direction.

That's the problem with all this data.

It's almost like we built the bullet

before you build the weapon.

Ted Cruz employed our data, our behavioral insights.

He started from a base of less than 5%

and we had a very slow and steady increase but firm above 35%,

making it the second most threatening candidate in the race.

Now, clearly, the Cross campaign has ended,

but what I can say is,

which of the two candidates left in this election,

one of them is using these technologies.

Donald John Trump I solemnly swear

faithfully execute the function

of President dos United States.

Elections are a marginal exercise.

It does not take a very sophisticated AI

to have a disproportionate impact.

Before Trump, the Brexit was another alleged customer.

Well, after 20 minutes to five, we can now say

that the decision taken in 1975 by the country

to join the common market has been reversed

for this referendum to leave the EU.

The Cambridge Analytica allegedly used the IA

to boost two major tremors

of political change in the last 50 years.

They are epic events and believe the hype,

they are directly connected to a software

created by a professor at Stanford.

In 2013 I described what they were doing was possible

and warned against this happening in the future.

Michal Kosinski at the time was a young Polish researcher

working on Psychometric Center.

What Michal did was to bring together the largest data set ever achieved

how people behaved on Facebook.

Psychometrics attempts to measure the psychological traits

as personality, intelligence, political views

and so on.

Traditionally, these characteristics were measured

using tests and questionnaires.

personality tests,

the most benign thing you might think.

Something that does not necessarily have much use, right?

Our idea was that instead of testing and questioning,

we could just look at the digital tracks

behaviors that are all living

to understand the receptive capacity,

meticulousness, neuroticism.

You can easily buy personal data,

like where you live, what associations you joined,

which gym you go.

In fact there is a market for personal data.

It turns out that we can find out a lot about what you will do

based on a very small set of information.

We are training deep learning networks

to deduce intimate traits.

Political opinions, personality, intelligence

and sexual orientation of the people,

everything from someone's face image.

Now think of countries that are not as free and open-minded.

If you can reveal the religious views

or political views or sexual orientation of the people

just based on profile pictures,

it can literally be a matter of life or death.

I think there's no turning back.

You know what is the Turing test?

It is when a human interacts with a computer,

and if the human does not know he is interacting with a computer

passes the test.

And in the coming days

you will be the human component in the Turing test.

Damn it.

Caleb right, you got it.

Because if this test pass

you will be right in the center of the largest scientific event

in human history.

If you created a conscious machine,

It is not the history of man.

It is the story of the gods.

It's almost as if technology were a god itself.

Such as weather conditions,

we can not make an impact,

we can not delay it,

we can not stop it.

We feel powerless.

If we think of God, is an unlimited amount of intelligence.

The closest we can get is to evolve our own intelligence,

fusing it with the artificial intelligence we are creating.

Today our computers, mobile phones, applications,

They give us a superhuman ability.

As the old maxim that says,

If you can not beat them, join them.

It's about a man-machine partnership.

We already see it with our cell phones, for example.

It acts as a memory prosthesis, right?

I no longer have to remember your phone number

because it is on my cell phone.

It's about machines that increase our human abilities

as opposed to replacing them altogether.

If you look at all the objects

that have made the jump from analog to digital in the last 20 years,

is very.

We last analog or digital object universe not

and the problem with this, of course,

It is that the output and input is very limited.

That is, are these.

Our eyes are very good.

We are able to absorb a lot of visual information.

But our information output is very low.

The reason this is important if we imagine a scenario

where IA is play a more prominent role in society.

We want good ways to interact with this technology

for it to end in improving.

I think it's incredibly important to the IA

no other should be us.

I may be wrong about what I'm saying,

and I'm open to ideas if anyone can suggest a better way,

but I think really we have to merge with IA

or we will be left behind.

It is hard to think of a system disconnect

which is distributed throughout the planet.

It is now distributed throughout the solar system.

You can not turn it off.

We opened a Pandora caixa.

We rained forces we can not control,

we can not stop.

We are in the middle of creating a new form of life on Earth.

We do not know what will happen next,

we do not know how will the intellect of a machine

when the intellect is far beyond human capabilities.

It's not just something that is possible.

The less frightening future I can imagine

is one in which we will have at least democratized AI

because if a company or a small group of people

able to develop a super digital almost divine intelligence,

they could rule the world.

At least when there is an evil dictator,

this human will die,

but there would be an AI death

they live forever

and we would have an immortal dictator

which we never would escape.

The search for artificial intelligence is a multi-billion dollar industry. With almost no regulation.