FOR EDUCATIONAL AND KNOWLEDGE SHARING PURPOSES ONLY. NOT-FOR-PROFIT. SEE COPYRIGHT DISCLAIMER.

17 Aug 2023. Jaan Tallinn is no stranger to disruptive tech: 25 years ago he co-engineered Kazaa, which allowed for the free download of films and music. He also co-engineered Skype, which disrupted traditional voice and video communication. But when he looks at the way Big Tech and governments are pushing the boundaries of artificial intelligence, he worries about our future. Could we be fast approaching the point when machines don’t need human input anymore? Host Steve Clemons asks Tallinn, who founded the Centre for the Study of Existential Risk at Cambridge University, about risks and opportunities posed by AI.

hi I’m Steve Clements and I have a

question is the wild quest for advanced

artificial intelligence more like a

suicide Race For Humanity let’s get to

the bottom line

[Music]

science fiction has been obsessed with

this idea for decades robots and super

intelligent computers take over our

decision-making powers and then they

enslave us and we only realize what’s

going on when it’s way too late like

when Keanu Reeves character in The

Matrix film figured out that he and all

of humanity were basically just

batteries for a huge AI machine well is

fiction becoming reality governments and

big corporations are locked in a frantic

race to come up with more advanced

applications for artificial intelligence

to manage everything cheaper and more

competently Health Care education even

battlefields in a way artificial

intelligence is the new arms race

Russian President Vladimir Putin once

said that the nation that leads in AI

will be the ruler of the world so is

there an existential risk to humans and

is it time to pause and take a step back

to make sure these efforts are regulated

and more guard rails are put in place to

minimize those risks my guest today says

it’s definitely that time he is Yan

talin one of the engineers behind hit

programs like Skype and Kaza and now the

founder of the center for the study of

existential risk at Cambridge University

he’s also co-founder of the future of

Life Institute that took the lead in

calling for a six-month moratorium on AI

research John it’s a real pleasure to

have you with us today and I and I just

want to start 23 years ago I read an

article that was the cover of Wired

Magazine by another technologist I knew

named Bill Joy and the title of that is

saying is is was why the future doesn’t

need us and I would love to hear where

we are in Bill Joy’s predictions and

what you think we need to be wary of as

we as we move into this new era of AI

indeed protections like Bill Joyce and

even they can go in further for example

Alan Turing in 1951 said that once AI

becomes smarter than humans we will

likely lose control to it

and I think the correct

position to take here is that as soon as

we cannot rule out that we will remain

controlled firmly in controlled for a

long time we should kind of

take necessary cautions precautions to

make sure that either we remain in

control or if we lose control the things

the future will still be good for us now

you have written about your concerns uh

in this area not being something that

evolves tomorrow but down the road as

super intelligence really evolves and

takes hold that mankind may be less

relevant to the equation can you explain

to our audience our lay audience what

your concerns are about computer super

intelligence

uh so turns out like there was just

recently

um a survey result I think by

organization called ugal and that kind

of a normal Ordinary People on the

streets they actually have like pretty

good intuitions uh what could go wrong

and in some ways uh sort of people in

Academia intellectuals they kind of have

a habit of downplaying the problems than

both the experts like Joshua bengio or

Jeff Hinton the inventors of of deep

learning or that people on the street

they understand

because like if you look at it the

reason why humans are like firmly in

control of this planet are not

chimpanzees is that because we are more

capable than they are we are more

intelligent we are not we are not

stronger but we know how to do long-term

planning Etc and now we are we as

species we are in a race to yield that

advantage to machines which indicate

is not a good idea

now you were involved with the founding

of a number of institutes that are

fascinating one is the future you know

looking at the study of existential risk

and others the future of Life Institute

which I find fascinating and led with

the letter signed by many of the world’s

most important technologists today

asking for a moratorium and research

this Elon Musk uh who’s been I guess on

the board of your Center but also was a

cosigner of that letter I guess my

question is can can Pandora be put back

in the Box can can can a letter like

that can a moratorium a call for a pause

actually have an impact on the global

development of AI today or things just

proceeded too far at this point

it definitely can there are multiple

reasons to be hopeful for example we

have done it with other technology

before we did uh ban human cloning even

though technologically this is very

possible and

also things like bio weapons Etc like

for we have been like less successful

about them but still like the success

has not been known has not been zero and

the other thing is that again uh they’re

really experts uh top experts in the

field they are concerned uh therefore

like there isn’t this uh sort of Luddite

uh voice that advocates for for pause no

they’re like this there’s a very kind of

mixed and wide uh developing consensus

about the need to take things lower and

one reason to take things lower is to

just make sure make sure that things go

well uh if we are just at the mercy of

Market forces we might not have the

ability to steer and things sufficiently

yeah and you’re a wealthy guy and you’ve

invested in a lot of these companies

over the years and you’ve been investing

in AI companies so it’s sort of

interesting on one hand to see you

warning about these dangers and yet

you’re a significant investor you run

with the crowd that’s out there that is

bringing this new technology apology

forward how do you square that with

yourself

yeah that is a good question and uh

my default way of looking at things is

that like what is the counterfactual

that I’m displacing that in other words

like if I didn’t invest like who else

would take my place then and uh so

that’s like one uh one approach I’ve

taken with companies like anthropic for

example uh the other kind of approach I

take is that I just tried not to be kind

of decisive investor uh like for example

in deepmind where I invested in 2010 11

if I remember correctly uh I was just

like very small shareholder uh so I kind

of viewed my investment as a ticket to

be you know present at the company and

talk about the concerns that I have that

I already had back then

now let’s talk about this letter and

President Biden’s meeting with seven

leading companies I have them listed as

Amazon anthropic Google inflection meta

Microsoft and open Ai and many people

believe that this meeting in the White

House of President Biden with these

Executives talking about voluntary

measures that they might take to Think

Through the impact and dangers and risks

of AI was precipitated by the future of

Life institute’s letter now love to hear

your thoughts on that but I think more

importantly is do you believe that these

seven Executives these these companies

and President Biden are sincere in what

they’re doing or is this fake

performance acting like they’re doing

something while behind the scenes

they’re just chugging along as they were

before

so I don’t think this is a fake

performance I mean I do have some

concerns that they like many people too

many people this might be like very new

topic so they’re just uh kind of uh

don’t exactly know what to think at this

point but like you kind of like everyone

has to now learn about this new

situation and when it comes to

sort of

taking credit for these things I think

like the most biggest credit really goes

for good or for bad to open AI about

releasing the chat chipity because that

kind of like caused a coast planet to

gonna pay attention uh to AI in a way

that it didn’t really pay before but yes

like that so the future of Life

institutes letter open letter gonna fell

into this like fertile ground prepared

by open Ai and then like in turn the

extinction statement by Center for AI

safety

there’s like one uh one sentence

uh declaration that they are Extinction

risk should be there at the same level

than risks from nuclear and bio you know

it’s it’s an interesting conversation I

actually try to think try not to

sensationalize when we have these

conversations but sometimes you know to

overstate for effect is part of learning

and also thinking around the corner uh

and trying to anticipate the

unanticipatable if you will things we

haven’t thought about as part of the

conversation and your future of Life

Institute actually has been doing this

and they they’ve developed something I I

best thing I call them is what they call

them are Slaughter Bots can you tell us

about Slaughter Bots and and what you

know just thinking in contemporary terms

and capacity in drones what could

possibly happen uh more recently than

people think with the slaughter Bots

yes so there are like two really big

problems one is that uh fully automated

you know putting AI in the military

causes makes it very hard for Humanity

Control Data as trajectory because I got

this point you are in a very

you know literal arms race and like when

you are in arms race your

you don’t have much maneuvering room

when it comes to thinking about what how

do you approach this new technology you

just have to go what the capabilities

are and what are kind of strategic

advantages so that’s like one big worry

uh about putting AI military the other

problem and as we’ve seen with cyber uh

cyber warfare is that as Things become

autonomous

and the attribution becomes very hard so

like telegram natural

Evolution for for automated fully

automated

Warfare as future Life Institute

slaughterbots videos show is that you

will get this like swarms of

miniaturized drones uh that like anyone

with money can produce and release uh

without attribution so you might we

might be just creating a world where

it’s no longer safe to be outside

because you might be chased down by a

swarms of of StoryBots one of the

interesting things I reaction I had to

the future of Life Institute letter when

you brought all of these major

technologists to call for a moratorium

is I wondered how does that compete that

Consortium compete if China is not

involved or or is China involved is

there an opportunity for a truly Global

deal that we’re not thinking you know

enough about how do you do that if

you’ve got one side that’s going to

follow guard rails and guidelines and

the other side that might not

so I’m not an expert on China but with

that caveat I just want to point out

that China already has AI regulation in

a way that U.S does not so so in some

ways us is kind of falling behind in

that like regularized regulative uh area

uh as a CEO even though EU is also

farther ahead than than us so uh and

also

like this is a global problem so like

just like we’ve seen with global warming

it’s not

enough

correct to assume that Chinese are just

the bad guys in the room like if there’s

a global problem like you have to like

you know

get people together and discuss it in

you know with with open cars rather than

uh trying to kind of shift the blame uh

to to the other guys it occurred to me

reading a Goldman Sachs report that says

that extrapolating our estimates

globally suggests that generative AI

could expose the equivalent of 300

million full-time jobs uh to Automation

in Europe and the United States and

along this line it just it just made me

think are we going to look as as the

world has changed when we went from

farming and large families of kids that

were farmers and helping their parents

to a much smaller population in the

world when there are far more humans

actually being employed

so uh yeah like if you’re if you pick

the

knowledge that they are just another

technology you can like firmly make the

case that well you shouldn’t be worried

about because like technology has in the

end always have been has been good uh

for humans including employment however

if you if you take this analogy of like

species then it’s clear that the

introduction of a smarter species quite

often does not go well for the for the

less Advanced species well it’s a

fascinating problem and another

dimension here I read you know uh met

recently with Barry Diller and Barry

Diller has started Consortium uh to

basically Sue AI operations that scrape

um material intellectual property the

property of others that are particularly

in the news business in news industry

and threaten these major lawsuits and

there’s now this very interesting thing

in the media and Publishing World about

property rights uh and whether or not

guard rails that can limit how AI learns

are a smart thing to do have you thought

about that at all and what do you think

the chances are that Publishers or

artists or people that create can

somehow get carve outs where they’re not

part of the AI world I find it perhaps

naive but I’d love to hear from you

so perhaps as someone who was 20 years

ago working on a peer-to-peer flight

sharing program or probably shouldn’t

like take like strong stances on this

issue however like one

point that I think about is that or I

suspect is that the issue with copyright

is just going to be very temporary in

some ways

it is more

like a feature of the current generation

of AI and that’s like AIS will get

smarter they might just actually need

less and less data and less and less

copyrighted data uh to kind of exhibit

the same uh or perhaps in a stronger

capabilities so I think in some ways the

copyright people

perhaps are fighting a good fight but

eventually I think it’s going to be a

losing fight I mean you did this I mean

as you just mentioned I mean you were

the the king of this you did this with

Kazan you you know brought new techno

technology and Skype I mean I think that

that that is a very interesting tension

out there if if you would kind of put

yourself 20 years ago when you were

doing dude were there ways that the

system could have slowed you down and

and I asked this as someone that when

you were 25 years old you were changing

the world and not worried about these

boundaries now you’re 50 years old and

you’re saying hey we need guidelines and

boundaries are you worried about the 25

year old version of you that’s ignoring

uh those guardrails and concerns today

that is a fascinating question

hmm

I kind of feel that I could if I would

face uh with like 20 year old myself I

could kind of like talk sense to him but

I think it’s uh kind of perhaps like

deceptive uh thought because indeed uh

like

one thing that I have for example

inherited from 20 year old 20 year old

myself is that I really hate software

patents I think they’re just attacks on

programmers uh that is just not a great

uh thing to have but uh

but like definitely My Views about open

source have been become like more

moderate as I see that the potentially

the open source AIS could actually be

the source of catastrophic events

so

is resistance futile I mean I mean I I

guess you know one hand you know I look

at technology advancing so much and

you’ve been such a big driver of this

and I do believe that we do make choices

I’m glad you mentioned human cloning

what I’m interested in is how do we take

the work that you’re doing in future of

life in the future of uh the concerns

for existential threats and give it

scale so that it becomes more the norm

uh and less of a boutique topic

I think that it’s now time to uh to

really put forward some uh early

regulations

at the very least to do something to

just exercise a muscle of Technology

regulation I mean EU has done some of

that I think more than the US has done

but but like specifically I’m thinking

about things like uh making sure that

data centers are certified certified so

like if you want to train an AI and they

are done like the big AI experiments are

done in data centers Big Data Centers

those data centers have to be certified

uh I I think that’s uh one of the steps

perhaps even easier step where there

seems to be got a lot of consensus on is

that AI outputs should be labeled like

nobody should be faced uh with a phone

call or or video or text

and fooled into thinking that this came

from Human there should be clear

indications that this is AI output and

there are things like liability so like

if for example Facebook puts out open

source Ai and that falls into bad hands

and something really catastrophic

happens as a result that responsibility

should go back to Facebook

we uh have a survey that was done in

2019 and I guess 4 000 people were asked

they got hundreds and hundreds of

responses asking experts who are

actually working on these technology

issues on uh whether machines would be

vastly better than humans at all

professions and at that time uh it said

within 10 years or two years it would be

ten percent within 30 years it would be

60 percent where do you fall in this

spectrum

I’m just like very uncertain I think

there is like significant

bump in probability in the next few

years just because uh like people have

sort of like discovered the gold mine

and like just throwing more and more

compute more and more uh people more and

more money I mean

uh I just watched the uh senat uh

testimony uh by uh Stuart Russell and he

said that that like there’s currently

about 10 10 billion dollars per month uh

being invested in AI startups uh and

that’s like more than the uh entire

science funding in the us for the rest

of the science therefore like granted

there is this Rush uh sort of Cold Rush

happening in AI in a very distinct kind

of manner when you compare it to the

rest of the kind of the funding crisis

in startup land and technology in

general and so perhaps like this

might actually going to precipitate uh

some certain capability gains that I’m

very concerned about if that doesn’t

happen then like sort of like bets are

off all of sort of off the table again

and and kind of remains to be seen like

how much more time we will have I’m

going to tell our audience uh beyond

that you are the real John talian you’re

not a deep fake we haven’t conjured you

to do all of this but you know maybe

someday we would be able to do that

maybe illegally I’ve seen deep fakes of

Tom Cruise on Tick Tock uh at an event

that semaphore where I work put together

we did a deep deep fake of Barry Diller

uh this big you know media Titan uh he

was not happy at all about it but it was

one of these things where as you kind of

look at the convergence of a lot of

different dimensions of of how it

changed what do you think truth is

actually in jeopardy

uh

in some sense sure but in some sense not

really so there are

ultimately I think

like we have lived with the ability to

produce kind of fake texts uh for a long

time and we have developed things like

uh you know security digital signatures

uh like website uh traffic encryption

things like that to dealing with this uh

with video content we have been you know

used to uh trusting it uh so there will

be like a period uh during which we

kind of

like many people will be fooled uh by by

fake videos but uh of course if nothing

nothing worse would happen because of AI

then I wouldn’t pay that word because

people just adjust it and start

demanding uh kind of authentication of

the sources and and indeed start

demanding laws that if you’re gonna fool

someone with a generated videos you

should go to jail

let me just ask you finally we’ve had a

discussion you and I about this before

and that is about the fragility of

democracy and whether technology is

worsening the problem or enhancing

Democratic options down the road we

we’re you know I’m in a country right

now we’re a former American president

just had his fourth indictment I’m not

sure how well we’re exhibiting democracy

today but when you think about this part

of the question is might AI get

democracy better

I mean hey I could get everything better

sure uh I I do think that there is this

just like with deep fakes there is kind

of like this turbulent Waters ahead uh

and if uh if the data doesn’t turn

existential I think we could

potentially just develop counter

measures uh to everything and kind of

adjusted a new situation like we did

with with previous uh powerful

Technologies such as like internet or

smartphones or cameras everywhere Etc

but yeah my own kind of worry is that uh

like uh every new generation of AI will

just present bigger and even larger

problems

singularitarian hacker investor

physicist Jan talin founder of the

center of the study of existential risk

and co-founder of the future of Life

Institute thank you so much for being

with us today

thank you very much

so what’s the bottom line we all would

love AI to help doctors diagnose our

ailments better or to protect us let’s

say against fraud and identity theft but

those are just the toes in the door

generative AI will eventually affect

everything and I mean everything and

there are significant human-less

Dimensions to it where data and machines

actually talk to each other they learn

from each other and they evolve without

us some of us would like to buy a car

that could drive itself that’s great but

are you willing to live in a country

that has an autonomously run government

actually when we look around that

doesn’t look like a bad idea but add to

that lethal autonomous weapon systems or

robotic Killers or cities that run

themselves without any workers and

things start to look a bit more scary we

should be worried about the power of a

handful of people who are making the big

decisions on artificial intelligence

today and then we should be even more

worried when the handful of people are

gone and AI is making all those

decisions by itself and that’s the

bottom line

foreign

[Music]

hi I’m Steve Clements and I have a question is the wild quest for advanced artificial intelligence more like a suicide Race For Humanity let’s get to the bottom line [Music] science fiction has been obsessed with this idea for decades robots and super intelligent computers take over our decision-making powers and then they enslave us and we only realize what’s going on when it’s way too late like when Keanu Reeves character in The Matrix film figured out that he and all of humanity were basically just batteries for a huge AI machine well is fiction becoming reality governments and big corporations are locked in a frantic race to come up with more advanced applications for artificial intelligence to manage everything cheaper and more competently Health Care education even battlefields in a way artificial intelligence is the new arms race Russian President Vladimir Putin once said that the nation that leads in AI will be the ruler of the world so is there an existential risk to humans and is it time to pause and take a step back to make sure these efforts are regulated and more guard rails are put in place to minimize those risks my guest today says it’s definitely that time he is Yan talin one of the engineers behind hit programs like Skype and Kaza and now the founder of the center for the study of existential risk at Cambridge University he’s also co-founder of the future of Life Institute that took the lead in calling for a six-month moratorium on AI research John it’s a real pleasure to have you with us today and I and I just want to start 23 years ago I read an article that was the cover of Wired Magazine by another technologist I knew named Bill Joy and the title of that is saying is is was why the future doesn’t need us and I would love to hear where we are in Bill Joy’s predictions and what you think we need to be wary of as we as we move into this new era of AI indeed protections like Bill Joyce and even they can go in further for example Alan Turing in 1951 said that once AI becomes smarter than humans we will likely lose control to it and I think the correct position to take here is that as soon as we cannot rule out that we will remain controlled firmly in controlled for a long time we should kind of take necessary cautions precautions to make sure that either we remain in control or if we lose control the things the future will still be good for us now you have written about your concerns uh in this area not being something that evolves tomorrow but down the road as super intelligence really evolves and takes hold that mankind may be less relevant to the equation can you explain to our audience our lay audience what your concerns are about computer super intelligence uh so turns out like there was just recently um a survey result I think by organization called ugal and that kind of a normal Ordinary People on the streets they actually have like pretty good intuitions uh what could go wrong and in some ways uh sort of people in Academia intellectuals they kind of have a habit of downplaying the problems than both the experts like Joshua bengio or Jeff Hinton the inventors of of deep learning or that people on the street they understand because like if you look at it the reason why humans are like firmly in control of this planet are not chimpanzees is that because we are more capable than they are we are more intelligent we are not we are not stronger but we know how to do long-term planning Etc and now we are we as species we are in a race to yield that advantage to machines which indicate is not a good idea now you were involved with the founding of a number of institutes that are fascinating one is the future you know looking at the study of existential risk and others the future of Life Institute which I find fascinating and led with the letter signed by many of the world’s most important technologists today asking for a moratorium and research this Elon Musk uh who’s been I guess on the board of your Center but also was a cosigner of that letter I guess my question is can can Pandora be put back in the Box can can can a letter like that can a moratorium a call for a pause actually have an impact on the global development of AI today or things just proceeded too far at this point it definitely can there are multiple reasons to be hopeful for example we have done it with other technology before we did uh ban human cloning even though technologically this is very possible and also things like bio weapons Etc like for we have been like less successful about them but still like the success has not been known has not been zero and the other thing is that again uh they’re really experts uh top experts in the field they are concerned uh therefore like there isn’t this uh sort of Luddite uh voice that advocates for for pause no they’re like this there’s a very kind of mixed and wide uh developing consensus about the need to take things lower and one reason to take things lower is to just make sure make sure that things go well uh if we are just at the mercy of Market forces we might not have the ability to steer and things sufficiently yeah and you’re a wealthy guy and you’ve invested in a lot of these companies over the years and you’ve been investing in AI companies so it’s sort of interesting on one hand to see you warning about these dangers and yet you’re a significant investor you run with the crowd that’s out there that is bringing this new technology apology forward how do you square that with yourself yeah that is a good question and uh my default way of looking at things is that like what is the counterfactual that I’m displacing that in other words like if I didn’t invest like who else would take my place then and uh so that’s like one uh one approach I’ve taken with companies like anthropic for example uh the other kind of approach I take is that I just tried not to be kind of decisive investor uh like for example in deepmind where I invested in 2010 11 if I remember correctly uh I was just like very small shareholder uh so I kind of viewed my investment as a ticket to be you know present at the company and talk about the concerns that I have that I already had back then now let’s talk about this letter and President Biden’s meeting with seven leading companies I have them listed as Amazon anthropic Google inflection meta Microsoft and open Ai and many people believe that this meeting in the White House of President Biden with these Executives talking about voluntary measures that they might take to Think Through the impact and dangers and risks of AI was precipitated by the future of Life institute’s letter now love to hear your thoughts on that but I think more importantly is do you believe that these seven Executives these these companies and President Biden are sincere in what they’re doing or is this fake performance acting like they’re doing something while behind the scenes they’re just chugging along as they were before so I don’t think this is a fake performance I mean I do have some concerns that they like many people too many people this might be like very new topic so they’re just uh kind of uh don’t exactly know what to think at this point but like you kind of like everyone has to now learn about this new situation and when it comes to sort of taking credit for these things I think like the most biggest credit really goes for good or for bad to open AI about releasing the chat chipity because that kind of like caused a coast planet to gonna pay attention uh to AI in a way that it didn’t really pay before but yes like that so the future of Life institutes letter open letter gonna fell into this like fertile ground prepared by open Ai and then like in turn the extinction statement by Center for AI safety there’s like one uh one sentence uh declaration that they are Extinction risk should be there at the same level than risks from nuclear and bio you know it’s it’s an interesting conversation I actually try to think try not to sensationalize when we have these conversations but sometimes you know to overstate for effect is part of learning and also thinking around the corner uh and trying to anticipate the unanticipatable if you will things we haven’t thought about as part of the conversation and your future of Life Institute actually has been doing this and they they’ve developed something I I best thing I call them is what they call them are Slaughter Bots can you tell us about Slaughter Bots and and what you know just thinking in contemporary terms and capacity in drones what could possibly happen uh more recently than people think with the slaughter Bots yes so there are like two really big problems one is that uh fully automated you know putting AI in the military causes makes it very hard for Humanity Control Data as trajectory because I got this point you are in a very you know literal arms race and like when you are in arms race your you don’t have much maneuvering room when it comes to thinking about what how do you approach this new technology you just have to go what the capabilities are and what are kind of strategic advantages so that’s like one big worry uh about putting AI military the other problem and as we’ve seen with cyber uh cyber warfare is that as Things become autonomous and the attribution becomes very hard so like telegram natural Evolution for for automated fully automated Warfare as future Life Institute slaughterbots videos show is that you will get this like swarms of miniaturized drones uh that like anyone with money can produce and release uh without attribution so you might we might be just creating a world where it’s no longer safe to be outside because you might be chased down by a swarms of of StoryBots one of the interesting things I reaction I had to the future of Life Institute letter when you brought all of these major technologists to call for a moratorium is I wondered how does that compete that Consortium compete if China is not involved or or is China involved is there an opportunity for a truly Global deal that we’re not thinking you know enough about how do you do that if you’ve got one side that’s going to follow guard rails and guidelines and the other side that might not so I’m not an expert on China but with that caveat I just want to point out that China already has AI regulation in a way that U.S does not so so in some ways us is kind of falling behind in that like regularized regulative uh area uh as a CEO even though EU is also farther ahead than than us so uh and also like this is a global problem so like just like we’ve seen with global warming it’s not enough correct to assume that Chinese are just the bad guys in the room like if there’s a global problem like you have to like you know get people together and discuss it in you know with with open cars rather than uh trying to kind of shift the blame uh to to the other guys it occurred to me reading a Goldman Sachs report that says that extrapolating our estimates globally suggests that generative AI could expose the equivalent of 300 million full-time jobs uh to Automation in Europe and the United States and along this line it just it just made me think are we going to look as as the world has changed when we went from farming and large families of kids that were farmers and helping their parents to a much smaller population in the world when there are far more humans actually being employed so uh yeah like if you’re if you pick the knowledge that they are just another technology you can like firmly make the case that well you shouldn’t be worried about because like technology has in the end always have been has been good uh for humans including employment however if you if you take this analogy of like species then it’s clear that the introduction of a smarter species quite often does not go well for the for the less Advanced species well it’s a fascinating problem and another dimension here I read you know uh met recently with Barry Diller and Barry Diller has started Consortium uh to basically Sue AI operations that scrape um material intellectual property the property of others that are particularly in the news business in news industry and threaten these major lawsuits and there’s now this very interesting thing in the media and Publishing World about property rights uh and whether or not guard rails that can limit how AI learns are a smart thing to do have you thought about that at all and what do you think the chances are that Publishers or artists or people that create can somehow get carve outs where they’re not part of the AI world I find it perhaps naive but I’d love to hear from you so perhaps as someone who was 20 years ago working on a peer-to-peer flight sharing program or probably shouldn’t like take like strong stances on this issue however like one point that I think about is that or I suspect is that the issue with copyright is just going to be very temporary in some ways it is more like a feature of the current generation of AI and that’s like AIS will get smarter they might just actually need less and less data and less and less copyrighted data uh to kind of exhibit the same uh or perhaps in a stronger capabilities so I think in some ways the copyright people perhaps are fighting a good fight but eventually I think it’s going to be a losing fight I mean you did this I mean as you just mentioned I mean you were the the king of this you did this with Kazan you you know brought new techno technology and Skype I mean I think that that that is a very interesting tension out there if if you would kind of put yourself 20 years ago when you were doing dude were there ways that the system could have slowed you down and and I asked this as someone that when you were 25 years old you were changing the world and not worried about these boundaries now you’re 50 years old and you’re saying hey we need guidelines and boundaries are you worried about the 25 year old version of you that’s ignoring uh those guardrails and concerns today that is a fascinating question hmm I kind of feel that I could if I would face uh with like 20 year old myself I could kind of like talk sense to him but I think it’s uh kind of perhaps like deceptive uh thought because indeed uh like one thing that I have for example inherited from 20 year old 20 year old myself is that I really hate software patents I think they’re just attacks on programmers uh that is just not a great uh thing to have but uh but like definitely My Views about open source have been become like more moderate as I see that the potentially the open source AIS could actually be the source of catastrophic events so is resistance futile I mean I mean I I guess you know one hand you know I look at technology advancing so much and you’ve been such a big driver of this and I do believe that we do make choices I’m glad you mentioned human cloning what I’m interested in is how do we take the work that you’re doing in future of life in the future of uh the concerns for existential threats and give it scale so that it becomes more the norm uh and less of a boutique topic I think that it’s now time to uh to really put forward some uh early regulations at the very least to do something to just exercise a muscle of Technology regulation I mean EU has done some of that I think more than the US has done but but like specifically I’m thinking about things like uh making sure that data centers are certified certified so like if you want to train an AI and they are done like the big AI experiments are done in data centers Big Data Centers those data centers have to be certified uh I I think that’s uh one of the steps perhaps even easier step where there seems to be got a lot of consensus on is that AI outputs should be labeled like nobody should be faced uh with a phone call or or video or text and fooled into thinking that this came from Human there should be clear indications that this is AI output and there are things like liability so like if for example Facebook puts out open source Ai and that falls into bad hands and something really catastrophic happens as a result that responsibility should go back to Facebook we uh have a survey that was done in 2019 and I guess 4 000 people were asked they got hundreds and hundreds of responses asking experts who are actually working on these technology issues on uh whether machines would be vastly better than humans at all professions and at that time uh it said within 10 years or two years it would be ten percent within 30 years it would be 60 percent where do you fall in this spectrum I’m just like very uncertain I think there is like significant bump in probability in the next few years just because uh like people have sort of like discovered the gold mine and like just throwing more and more compute more and more uh people more and more money I mean uh I just watched the uh senat uh testimony uh by uh Stuart Russell and he said that that like there’s currently about 10 10 billion dollars per month uh being invested in AI startups uh and that’s like more than the uh entire science funding in the us for the rest of the science therefore like granted there is this Rush uh sort of Cold Rush happening in AI in a very distinct kind of manner when you compare it to the rest of the kind of the funding crisis in startup land and technology in general and so perhaps like this might actually going to precipitate uh some certain capability gains that I’m very concerned about if that doesn’t happen then like sort of like bets are off all of sort of off the table again and and kind of remains to be seen like how much more time we will have I’m going to tell our audience uh beyond that you are the real John talian you’re not a deep fake we haven’t conjured you to do all of this but you know maybe someday we would be able to do that maybe illegally I’ve seen deep fakes of Tom Cruise on Tick Tock uh at an event that semaphore where I work put together we did a deep deep fake of Barry Diller uh this big you know media Titan uh he was not happy at all about it but it was one of these things where as you kind of look at the convergence of a lot of different dimensions of of how it changed what do you think truth is actually in jeopardy uh in some sense sure but in some sense not really so there are ultimately I think like we have lived with the ability to produce kind of fake texts uh for a long time and we have developed things like uh you know security digital signatures uh like website uh traffic encryption things like that to dealing with this uh with video content we have been you know used to uh trusting it uh so there will be like a period uh during which we kind of like many people will be fooled uh by by fake videos but uh of course if nothing nothing worse would happen because of AI then I wouldn’t pay that word because people just adjust it and start demanding uh kind of authentication of the sources and and indeed start demanding laws that if you’re gonna fool someone with a generated videos you should go to jail let me just ask you finally we’ve had a discussion you and I about this before and that is about the fragility of democracy and whether technology is worsening the problem or enhancing Democratic options down the road we we’re you know I’m in a country right now we’re a former American president just had his fourth indictment I’m not sure how well we’re exhibiting democracy today but when you think about this part of the question is might AI get democracy better I mean hey I could get everything better sure uh I I do think that there is this just like with deep fakes there is kind of like this turbulent Waters ahead uh and if uh if the data doesn’t turn existential I think we could potentially just develop counter measures uh to everything and kind of adjusted a new situation like we did with with previous uh powerful Technologies such as like internet or smartphones or cameras everywhere Etc but yeah my own kind of worry is that uh like uh every new generation of AI will just present bigger and even larger problems singularitarian hacker investor physicist Jan talin founder of the center of the study of existential risk and co-founder of the future of Life Institute thank you so much for being with us today thank you very much so what’s the bottom line we all would love AI to help doctors diagnose our ailments better or to protect us let’s say against fraud and identity theft but those are just the toes in the door generative AI will eventually affect everything and I mean everything and there are significant human-less Dimensions to it where data and machines actually talk to each other they learn from each other and they evolve without us some of us would like to buy a car that could drive itself that’s great but are you willing to live in a country that has an autonomously run government actually when we look around that doesn’t look like a bad idea but add to that lethal autonomous weapon systems or robotic Killers or cities that run themselves without any workers and things start to look a bit more scary we should be worried about the power of a handful of people who are making the big decisions on artificial intelligence today and then we should be even more worried when the handful of people are gone and AI is making all those decisions by itself and that’s the bottom line foreign [Music] 25:26 NOW PLAYING Rob

FOR EDUCATIONAL AND KNOWLEDGE SHARING PURPOSES ONLY. NOT-FOR-PROFIT. SEE COPYRIGHT DISCLAIMER.