Humans & AI: A Healthy Relationship? | Marcel Blattner | TEDxZurich

two years ago a program crawling the web

two years ago a program crawling the web looking for public accessible social

looking for public accessible social

looking for public accessible social media platform data found the Instagram

media platform data found the Instagram

media platform data found the Instagram profile of my son

profile of my son

profile of my son despite the program fetch the data and

despite the program fetch the data and

despite the program fetch the data and injected it into a database somewhere

injected it into a database somewhere

injected it into a database somewhere located abroad just a few days later my

located abroad just a few days later my

located abroad just a few days later my son resentfully said to me houston we

son resentfully said to me houston we

son resentfully said to me houston we have a problem I see all my Instagram

have a problem I see all my Instagram

have a problem I see all my Instagram data on another website I didn’t upload

data on another website I didn’t upload

data on another website I didn’t upload any data to this website nor did I

any data to this website nor did I

any data to this website nor did I subscribe to their services how is that

subscribe to their services how is that

subscribe to their services how is that possible what’s going on here together

possible what’s going on here together

possible what’s going on here together we double-checked and we figured out

we double-checked and we figured out

we double-checked and we figured out that he changed something in the

that he changed something in the

that he changed something in the security settings of his Instagram app

security settings of his Instagram app

security settings of his Instagram app making his profile publicly accessible

making his profile publicly accessible

making his profile publicly accessible for everybody I didn’t like that at all

for everybody I didn’t like that at all

for everybody I didn’t like that at all because you could easily reverse

because you could easily reverse

because you could easily reverse engineer a typical day in the life of my

engineer a typical day in the life of my

engineer a typical day in the life of my son now the whole story triggered a very

son now the whole story triggered a very

son now the whole story triggered a very intense and deep discussion about the

intense and deep discussion about the

intense and deep discussion about the internet ecosystem technical details on

internet ecosystem technical details on

internet ecosystem technical details on how messages are transmitted the history

how messages are transmitted the history

how messages are transmitted the history of the Internet why do we have an

of the Internet why do we have an

of the Internet why do we have an Internet at all data privacy and search

Internet at all data privacy and search

Internet at all data privacy and search engines and how they influence our daily

engines and how they influence our daily

engines and how they influence our daily decisions I really like this discussion

decisions I really like this discussion

decisions I really like this discussion and I learned two things first my son

and I learned two things first my son

and I learned two things first my son knows all the apps from

knows all the apps from

knows all the apps from our universe and in case there are any

our universe and in case there are any

our universe and in case there are any parallel universes he knows all the apps

parallel universes he knows all the apps

parallel universes he knows all the apps from there as well he has quite a good

from there as well he has quite a good

from there as well he has quite a good overview of all the functionalities and

overview of all the functionalities and

overview of all the functionalities and he uses all these apps as responsibly as

he uses all these apps as responsibly as

he uses all these apps as responsibly as he can second he has rather limited

he can second he has rather limited

he can second he has rather limited knowledge on the forces governing the

knowledge on the forces governing the

knowledge on the forces governing the internet ecosystem he is not really

internet ecosystem he is not really

internet ecosystem he is not really aware how this ecosystem influences the

aware how this ecosystem influences the

aware how this ecosystem influences the decisions he makes and the way he sees

decisions he makes and the way he sees

decisions he makes and the way he sees and understands our world

and understands our world

and understands our world of course that worries me a lot

of course that worries me a lot

of course that worries me a lot now I can safely say that my son was not

now I can safely say that my son was not

now I can safely say that my son was not well enough educated in those aspects

well enough educated in those aspects

well enough educated in those aspects now you might say Oh Marcel come on

now you might say Oh Marcel come on

now you might say Oh Marcel come on chill out I mean it’s important to know

chill out I mean it’s important to know

chill out I mean it’s important to know how to use all these apps and

how to use all these apps and

how to use all these apps and technologies in a ficient way but it

technologies in a ficient way but it

technologies in a ficient way but it isn’t as important to know how they are

isn’t as important to know how they are

isn’t as important to know how they are engineered and by the way Marcel isn’t

engineered and by the way Marcel isn’t

engineered and by the way Marcel isn’t it your job your challenge to teach your

it your job your challenge to teach your

it your job your challenge to teach your kids on all those aspects I say yes you

kids on all those aspects I say yes you

kids on all those aspects I say yes you are right but that’s just one part of

are right but that’s just one part of

are right but that’s just one part of the story and the other part goes as

the story and the other part goes as

the story and the other part goes as follows the government our educational

follows the government our educational

follows the government our educational system and we as a society are in charge

system and we as a society are in charge

system and we as a society are in charge as well to educate future generations on

as well to educate future generations on

as well to educate future generations on the ongoing digital revolution and I

the ongoing digital revolution and I

the ongoing digital revolution and I want to support this statement by

want to support this statement by

want to support this statement by talking about artificial intelligence

talking about artificial intelligence

talking about artificial intelligence whoo artificial intelligence I’m sure

whoo artificial intelligence I’m sure

whoo artificial intelligence I’m sure everybody here heard about it in one or

everybody here heard about it in one or

everybody here heard about it in one or the other way lately

the other way lately

the other way lately you might have rats that Google has a

you might have rats that Google has a

you might have rats that Google has a built a system based on artificial

built a system based on artificial

built a system based on artificial intelligence which beats a top player at

intelligence which beats a top player at

intelligence which beats a top player at the game of goal now that’s really

the game of goal now that’s really

the game of goal now that’s really impressive because the game of go has a

impressive because the game of go has a

impressive because the game of go has a much higher complexity compared to the

much higher complexity compared to the

much higher complexity compared to the game of chess systems based on

game of chess systems based on

game of chess systems based on artificial intelligence are at least as

artificial intelligence are at least as

artificial intelligence are at least as good as humans at brain cancer detection

good as humans at brain cancer detection

good as humans at brain cancer detection self-driving cars cell flying drones

self-driving cars cell flying drones

self-driving cars cell flying drones face detection

face detection

face detection voice recognition algorithmic trading

voice recognition algorithmic trading

voice recognition algorithmic trading news book and music recommendations

news book and music recommendations

news book and music recommendations robots chat BOTS and and and artificial

robots chat BOTS and and and artificial

robots chat BOTS and and and artificial intelligence is here and now and you

intelligence is here and now and you

intelligence is here and now and you carry it in your pocket now one type of

carry it in your pocket now one type of

carry it in your pocket now one type of system or one type of technology making

system or one type of technology making

system or one type of technology making possible all this break truth is called

possible all this break truth is called

possible all this break truth is called deep learning or deep neural network

deep learning or deep neural network

deep learning or deep neural network these systems are based on layers each

these systems are based on layers each

these systems are based on layers each layers is composed of small

layers is composed of small

layers is composed of small computational units called neurons

computational units called neurons

computational units called neurons neurons within one layer are connected

neurons within one layer are connected

neurons within one layer are connected to neurons within the next preceding

to neurons within the next preceding

to neurons within the next preceding layer now assume you have a bunch of

layer now assume you have a bunch of

layer now assume you have a bunch of images which are unlabeled and the image

images which are unlabeled and the image

images which are unlabeled and the image show different animals and you want the

show different animals and you want the

show different animals and you want the system to classify those images how does

system to classify those images how does

system to classify those images how does this work well the system reads the

this work well the system reads the

this work well the system reads the image and passes the data to the first

image and passes the data to the first

image and passes the data to the first layer neurons in the first layer

layer neurons in the first layer

layer neurons in the first layer responds to very simple patterns and

responds to very simple patterns and

responds to very simple patterns and shapes like edges the activated neurons

shapes like edges the activated neurons

shapes like edges the activated neurons then pass the data to the next layer the

then pass the data to the next layer the

then pass the data to the next layer the second one neurons in the second layer

second one neurons in the second layer

second one neurons in the second layer respond to more complex patterns like

respond to more complex patterns like

respond to more complex patterns like arms or legs or years

arms or legs or years

arms or legs or years again the activated neurons pass the

again the activated neurons pass the

again the activated neurons pass the data to the next layer and this goes on

data to the next layer and this goes on

data to the next layer and this goes on and on and on till the last layer the

and on and on till the last layer the

and on and on till the last layer the last layer knows about the complex

last layer knows about the complex

last layer knows about the complex concept of how an animal looks like like

concept of how an animal looks like like

concept of how an animal looks like like a fish cat or dog finally the output of

a fish cat or dog finally the output of

a fish cat or dog finally the output of the system is a bunch of numbers telling

the system is a bunch of numbers telling

the system is a bunch of numbers telling you how confident the system is at

you how confident the system is at

you how confident the system is at detecting a cat or an elephant or any

detecting a cat or an elephant or any

detecting a cat or an elephant or any other animal now the whole thing works

other animal now the whole thing works

other animal now the whole thing works somehow similar to what is going on in

somehow similar to what is going on in

somehow similar to what is going on in our brains when we humans classify

our brains when we humans classify

our brains when we humans classify images different regions in our brain

images different regions in our brain

images different regions in our brain respond to different patterns on

respond to different patterns on

respond to different patterns on different complexities deep learning

different complexities deep learning

different complexities deep learning systems are of course not limited to

systems are of course not limited to

systems are of course not limited to process images you could input let’s say

process images you could input let’s say

process images you could input let’s say text from an email and the system tells

text from an email and the system tells

text from an email and the system tells you what this is BAM or not or you could

you what this is BAM or not or you could

you what this is BAM or not or you could input voice records and the output of

input voice records and the output of

input voice records and the output of the system might be transcriptions or

the system might be transcriptions or

the system might be transcriptions or you could input data related to the US

you could input data related to the US

you could input data related to the US presidential election and the output

presidential election and the output

presidential election and the output might be well let’s talk about something

might be well let’s talk about something

might be well let’s talk about something else please

these systems are even capable kind of

these systems are even capable kind of dreaming so they show signs of

dreaming so they show signs of

dreaming so they show signs of creativity take the image on top the

creativity take the image on top the

creativity take the image on top the famous image and input it to the system

famous image and input it to the system

famous image and input it to the system but you don’t ask the system to classify

but you don’t ask the system to classify

but you don’t ask the system to classify the image but you ask the system to

the image but you ask the system to

the image but you ask the system to amplify patterns

amplify patterns

amplify patterns it recognizes on the image patterns that

it recognizes on the image patterns that

it recognizes on the image patterns that it has seen and learned before and you

it has seen and learned before and you

it has seen and learned before and you can end up with something like the image

can end up with something like the image

can end up with something like the image on the bottom this is Ammar very high

on the bottom this is Ammar very high

on the bottom this is Ammar very high accurate in classifying and show signs

accurate in classifying and show signs

accurate in classifying and show signs of creativity now more and more system

of creativity now more and more system

of creativity now more and more system coming up learning by direct

coming up learning by direct

coming up learning by direct interactions with humans Microsoft did a

interactions with humans Microsoft did a

interactions with humans Microsoft did a very interesting experiment lately they

very interesting experiment lately they

very interesting experiment lately they release the chat bot on Twitter and the

release the chat bot on Twitter and the

release the chat bot on Twitter and the bot was supposed to engage Millennials

bot was supposed to engage Millennials

bot was supposed to engage Millennials with artificial intelligence shortly

with artificial intelligence shortly

with artificial intelligence shortly after they released about they had to

after they released about they had to

after they released about they had to shut down the machine what happens the

shut down the machine what happens the

shut down the machine what happens the boat turned racist and started to

boat turned racist and started to

boat turned racist and started to insulting other people now what is the

insulting other people now what is the

insulting other people now what is the lesson learned here what we see here is

lesson learned here what we see here is

lesson learned here what we see here is that machines can be biased but this

that machines can be biased but this

that machines can be biased but this should not come as a surprise because

should not come as a surprise because

should not come as a surprise because machines learn from data and we humans

machines learn from data and we humans

machines learn from data and we humans we generate those data but we humans are

we generate those data but we humans are

we generate those data but we humans are biased

biased

biased that’s how something like discrimination

that’s how something like discrimination

that’s how something like discrimination can find a path into those machines now

can find a path into those machines now

can find a path into those machines now here it’s obvious that something went

here it’s obvious that something went

here it’s obvious that something went completely wrong but take a more subtle

completely wrong but take a more subtle

completely wrong but take a more subtle case like an assistant for churches in

case like an assistant for churches in

case like an assistant for churches in risk assessment on cream

risk assessment on cream

risk assessment on cream there are well-documented cases were

there are well-documented cases were

there are well-documented cases were non-white persons were discriminated by

non-white persons were discriminated by

non-white persons were discriminated by such machines now that means that we

such machines now that means that we

such machines now that means that we have to think about somehow how we

have to think about somehow how we

have to think about somehow how we incorporate societal values in those

incorporate societal values in those

incorporate societal values in those machines but who is going to do that I

machines but who is going to do that I

machines but who is going to do that I mean who will decide what values will go

mean who will decide what values will go

mean who will decide what values will go in is it the government or IT cracks or

in is it the government or IT cracks or

in is it the government or IT cracks or researchers or you and me the point is

researchers or you and me the point is

researchers or you and me the point is that at the moment almost nobody is able

that at the moment almost nobody is able

that at the moment almost nobody is able to participate in that discussion and

to participate in that discussion and

to participate in that discussion and the reason why is well we are not

the reason why is well we are not

the reason why is well we are not educated but think about that I mean we

educated but think about that I mean we

educated but think about that I mean we teach our kids on the importance to know

teach our kids on the importance to know

teach our kids on the importance to know that we had graffiti in forms of cave

that we had graffiti in forms of cave

that we had graffiti in forms of cave painting ten thousand years ago but we

painting ten thousand years ago but we

painting ten thousand years ago but we don’t educate our kids on the drivers of

don’t educate our kids on the drivers of

don’t educate our kids on the drivers of the ongoing digital revolution a digital

the ongoing digital revolution a digital

the ongoing digital revolution a digital revolution which is about to change

revolution which is about to change

revolution which is about to change society now and a revolution which has

society now and a revolution which has

society now and a revolution which has the potential to change humanity in the

the potential to change humanity in the

the potential to change humanity in the future Nick Bostrom a recognized expert

future Nick Bostrom a recognized expert

future Nick Bostrom a recognized expert in the field of artificial intelligence

in the field of artificial intelligence

in the field of artificial intelligence hypothesized on the question what

hypothesized on the question what

hypothesized on the question what happens when a machine reaches human

happens when a machine reaches human

happens when a machine reaches human level of intelligence and he says that

level of intelligence and he says that

level of intelligence and he says that such a machine will go piont

such a machine will go piont

such a machine will go piont any human cognitive capabilities very

any human cognitive capabilities very

any human cognitive capabilities very quickly now what does this mean for us

quickly now what does this mean for us

quickly now what does this mean for us there are several scenarios of course

there are several scenarios of course

there are several scenarios of course one scenario is this one we do nothing

one scenario is this one we do nothing

one scenario is this one we do nothing and hang around all day along

and hang around all day along

and hang around all day along and the machines will do our work not my

and the machines will do our work not my

and the machines will do our work not my preferred scenario because work is too

preferred scenario because work is too

preferred scenario because work is too much fun right another scenario probably

much fun right another scenario probably

much fun right another scenario probably more realistic is that such a machine

more realistic is that such a machine

more realistic is that such a machine would even not recognize anymore that

would even not recognize anymore that

would even not recognize anymore that humans are intelligent what would that

humans are intelligent what would that

humans are intelligent what would that mean for us what would that mean for a

mean for us what would that mean for a

mean for us what would that mean for a relationship between humans and machines

relationship between humans and machines

relationship between humans and machines of course nobody knows you know I’m

of course nobody knows you know I’m

of course nobody knows you know I’m really a big fan of all these

really a big fan of all these

really a big fan of all these technologies and in particular of deep

technologies and in particular of deep

technologies and in particular of deep learning and I think that these machines

learning and I think that these machines

learning and I think that these machines can help us solving complex and

can help us solving complex and

can help us solving complex and important problems in our world I think

important problems in our world I think

important problems in our world I think we can use these machines for the good

we can use these machines for the good

we can use these machines for the good but to release the potential in the

but to release the potential in the

but to release the potential in the right direction that we really can

right direction that we really can

right direction that we really can benefit from it we have to think about

benefit from it we have to think about

benefit from it we have to think about some really hard questions questions

some really hard questions questions

some really hard questions questions such as who is responsible for machine

such as who is responsible for machine

such as who is responsible for machine generated decisions should intelligent

generated decisions should intelligent

generated decisions should intelligent machines have their own rights and where

machines have their own rights and where

machines have their own rights and where and how do we want to engage artificial

and how do we want to engage artificial

and how do we want to engage artificial intelligence and what societal values

intelligence and what societal values

intelligence and what societal values should go in do we need regulation I

should go in do we need regulation I

should go in do we need regulation I hope you come along with me now that

hope you come along with me now that

hope you come along with me now that these questions are by far too important

these questions are by far too important

these questions are by far too important to be discussed just among experts this

to be discussed just among experts this

to be discussed just among experts this discussion has to be brought to the

discussion has to be brought to the

discussion has to be brought to the public domain and one key element to do

public domain and one key element to do

public domain and one key element to do so is education education on all the

so is education education on all the

so is education education on all the fundamental topics touching our digital

fundamental topics touching our digital

fundamental topics touching our digital lives education starting at primary

lives education starting at primary

lives education starting at primary school so the healthy relationship

school so the healthy relationship

school so the healthy relationship between humans and machines is triggered

between humans and machines is triggered

between humans and machines is triggered by education

by education

by education by the way my son and me we wrote to the

by the way my son and me we wrote to the

by the way my son and me we wrote to the website owner and told him to delete all

website owner and told him to delete all

website owner and told him to delete all the data he fetched from my son’s

the data he fetched from my son’s

the data he fetched from my son’s Instagram profile and you know what at

Instagram profile and you know what at

Instagram profile and you know what at that time my son knew that the website

that time my son knew that the website

that time my son knew that the website owner has to do that there was no option

owner has to do that there was no option

owner has to do that there was no option to not to do it so the key was education

to not to do it so the key was education

to not to do it so the key was education thank you

thank you

thank you [Applause]

[Applause]

[Applause] [Music]