Download "Why this top AI guru thinks we might be in extinction level trouble | The InnerView"

Please wait. We're preparing links for easy ad-free video watching and downloading.

Previous video: How CIA spies disguise their identity | Andrew Bustamante and Lex Fridman Next video: PAJAMA PARTY | Pajama Lookbook [AI Art] (Model Olivia) [AI Girls] (FASHION)

0:00

AI guru

2:11

What is AGI, and what are the risks it could bring?

3:51

Nobody knows why AI does what it does

5:41

From an AI enthusiast to advocating for a more cautious approach to AI development

7:58

What does Connor expect to happen when we lose control to AGI?

11:40

People like Sam Altman are rockstars

14:38

Connor's vision of a safe AI future

15:24

Imran: "One year left?"

17:26

Normal people do have the right intuition about AI

20:58

ChatGPT, limitations of AI, and a story about a frog

24:53

Control AI

PAJAMA PARTY | Pajama Lookbook [AI Art] (Model Olivia) [AI Girls] (FASHION)

PAJAMA PARTY | Pajama Lookbook [AI Art] (Model Olivia) [AI Girls] (FASHION)

Channel: AIGreatArt

How CIA spies disguise their identity | Andrew Bustamante and Lex Fridman

How CIA spies disguise their identity | Andrew Bustamante and Lex Fridman

Channel: Lex Clips

AI doom

AI extinction

AI manipulation

Agi

Ai Development

Ai Laws

Ai Regulation

Artificial General Intelligence

Chatgpt

Conjecture

Connor Leahy

Connor Leahy Imran Garda

Connor Leahy TRT World

Connor Leahy interview

Geoffrey Hinton

Openai

Sam Altman

Tech people

ai

artificial intelligence

control ai

robots

tech

technology

00:00:00

[Music]

00:00:05

coni is one of the world's leading Minds

00:00:08

in artificial

00:00:10

intelligence he's a hacker who sees the

00:00:12

rise of AI as an existential threat to

00:00:16

humanity he dedicates his life to make

00:00:19

sure its success doesn't spell our

00:00:23

Doom there will be intelligent creatures

00:00:28

on this planet that are not

00:00:31

human this is not

00:00:33

normal and there will be no going

00:00:38

back and if we don't control them then

00:00:41

the future will belong to them not to

00:00:45

us Ley is the CEO of conjecture AI a

00:00:49

startup that tries to understand how AI

00:00:52

systems think with the aim of aligning

00:00:55

them to human values he speaks to the

00:00:58

interview about why he believes the end

00:01:00

is near and explains how he's trying to

00:01:03

stop

00:01:05

[Music]

00:01:19

it and Connor Ley joins us now on the

00:01:23

interview he's the CEO of conjecture

00:01:27

he's in our London Studio good to see

00:01:29

you there good to have you on the

00:01:31

program Connor You're Something of an AI

00:01:34

Guru and you're also one of those voices

00:01:37

saying we need to be very very careful

00:01:39

right now and a lot of people don't

00:01:42

quite have the knowledge or the they

00:01:45

don't quite have the vocabulary or the

00:01:47

deeper understanding as to why they

00:01:48

should be worried they just feel some

00:01:50

sort of sense of Doom but they can't

00:01:54

quite map it out so maybe you can help

00:01:57

us along that path why should should we

00:02:00

be worried about

00:02:02

AGI and tell me the difference between

00:02:04

AGI and what is widely perceived as AI

00:02:08

right now so I'll answer the second

00:02:11

question first just to get some

00:02:12

definitions out of the way sure the

00:02:14

truth is is that there's really no true

00:02:16

definition of the word AGI and people

00:02:18

use it to mean all kinds of different

00:02:20

things when I talk about the word AGI

00:02:22

usually what I mean by this is AI

00:02:25

systems or computer systems that are

00:02:27

more capable than humans at all tasks

00:02:31

that they could do so this involves you

00:02:33

know any scientific task programming

00:02:36

remote work uh science business politics

00:02:40

anything and these are systems that do

00:02:42

not currently exist but are actively

00:02:45

attempting to be built there are many

00:02:47

people working on building of systems

00:02:48

and many experts believe these systems

00:02:50

are close and as for why these systems

00:02:53

could are going to be a problem well I

00:02:56

actually think that a lot of people have

00:02:57

the right intuition here the intuition

00:02:59

question here is just well if you build

00:03:02

something that is more competent than

00:03:04

you it's smarter than you and all the

00:03:06

people you know and all the people in

00:03:07

the world it is better at business

00:03:09

politics manipulation

00:03:12

deception science weapons development

00:03:14

everything and you don't control those

00:03:16

things which we currently do not know

00:03:19

how to do well why would you expect that

00:03:21

to go well yeah it reminds me a little

00:03:23

bit about the debate about whether we

00:03:25

should be looking for life in in the

00:03:28

universe beyond our solar system Stephen

00:03:30

Hawking said be careful look at the

00:03:33

history of the world anytime you sort of

00:03:34

invite us a stronger power more more

00:03:37

competent power they might come and

00:03:39

destroy you but then the counter to that

00:03:41

is that you're mapping human behavior

00:03:44

human

00:03:46

desires passions needs wants onto this

00:03:49

thing is this natural to do and fair to

00:03:53

do because humans created it humans

00:03:55

humans created the parameters for

00:03:57

it so it's actually worse than that in

00:04:01

that it's really important to understand

00:04:03

that when we talk about AI it's easy to

00:04:05

imagine it to be software and the way

00:04:08

software generally works it is written

00:04:11

by a pro by a programmer they write code

00:04:14

which tells the computer what to do step

00:04:16

by step this is not how AI Works AI is

00:04:21

more like organic it's more like it is

00:04:24

grown you use these big

00:04:27

supercomputers to take a bunch of data

00:04:30

and grow a program that can solve the

00:04:34

problems in the data now this program

00:04:37

does not look like something written by

00:04:38

humans it's not code it's not lines of

00:04:41

instructions it's more like a huge pile

00:04:44

of billions and billions of numbers and

00:04:47

we know if we can run all these numbers

00:04:49

re execute these numbers they can do

00:04:51

really amazing things but no one knows

00:04:54

why so it's way more like dealing with a

00:04:56

biological thing like if you look at

00:04:58

like a bacterium or something and the

00:05:00

bacteria can do some crazy things and we

00:05:01

don't really know why and this is kind

00:05:03

of how our AIS are so the question is

00:05:06

less you know will humans impart

00:05:09

emotions into these systems we don't

00:05:11

know how to do that it's more if you

00:05:13

build systems if you grow systems if you

00:05:15

grow bacteria who are designed to solve

00:05:19

problems to you know solve games to make

00:05:23

money or whatever what kind of things

00:05:26

will you grow and by default you're

00:05:29

going to grow things that are good at

00:05:31

solving problems at gaining power at

00:05:33

tricking people at you know building

00:05:36

things and so on because this is what we

00:05:38

want you reverse engineered gpt2 at the

00:05:44

age of 24 which was a few years

00:05:47

ago that's well part of the legend I

00:05:50

mean that's part of the the the

00:05:52

credentialing of you before they say

00:05:53

well this guy is saying we're in big

00:05:56

trouble they say well by the way you

00:05:58

know he knows what he's talking about

00:05:59

because because technically he knows

00:06:00

what he's

00:06:01

doing tell me tell me about the pivot

00:06:04

point between being a Believer and

00:06:06

enthusiastic about this to becoming a

00:06:08

Warner what

00:06:10

happened so uh the story goes back even

00:06:13

further than that um reverse engineering

00:06:15

is um a bit generous it's more like I

00:06:19

built a system I found out that no one

00:06:21

can reverse engineer it oh and this is a

00:06:23

big problem um but it was even before

00:06:25

then so I've been very into AI since I

00:06:28

was a teenager because I want to make

00:06:30

the world a better place and I think

00:06:31

that a lot of people who believe in AI a

00:06:33

lot of the tech people who are doing the

00:06:34

things which are think are dangerous I

00:06:35

think most of them maybe not most but

00:06:38

most of them probably are good people

00:06:40

they're trying to build technology to

00:06:41

make the world a better place you know

00:06:43

when I grew up uh technology was great

00:06:45

you know the internet was making people

00:06:47

more connected we were getting access to

00:06:49

better medicines and there was you know

00:06:51

solar power was improving there's all

00:06:53

these great things that science was

00:06:54

doing so I was very excited about more

00:06:56

science and about more technology and

00:06:58

well what is the what is the best

00:07:00

technology than intelligence if we just

00:07:02

had intelligence well wow we could solve

00:07:05

all the problems we could do all the

00:07:06

science we could you know invent all the

00:07:09

cancer medicines we could you know

00:07:11

develop all the cool stuff so I was

00:07:13

thinking when I was a teenager and this

00:07:16

is I think a common trajectory is that

00:07:17

people when they're kind of like first

00:07:19

exposed to some of these like techno

00:07:21

utopian AGI dreams it sounds great you

00:07:24

know it sounds like such a great great

00:07:26

solution but then as you think about

00:07:28

this problem more you kind of realize

00:07:30

that like the problem with AGI is not

00:07:33

really how to build it it's how to

00:07:35

control it that's much harder just

00:07:39

because you can make something which is

00:07:41

smart or that solves a problem does not

00:07:43

mean you can make something that will

00:07:44

listen to you that will do what you

00:07:46

truly want this is much much harder and

00:07:49

this is and as I started looking into

00:07:51

this problem more in my early 20s I

00:07:53

start realizing like wow we are really

00:07:55

really not making progress on this

00:07:57

problem so in that worst case scenario

00:07:59

whether we have an apocalyptic ending

00:08:02

for all of us we get destroyed

00:08:04

existentially or we become enslaved in

00:08:06

The Matrix or whatever it might

00:08:08

be tell me how it actually happens in

00:08:12

your mind how does this

00:08:15

AGI um assume control I mean there these

00:08:19

famous moments in Terminator and

00:08:21

Elsewhere One of the Terminators that

00:08:23

final scene where the nuclear bombs are

00:08:25

going off all over I mean there lots of

00:08:26

different ways people have imagined this

00:08:29

the way you see it tell me how it

00:08:31

happens and how if things continue to go

00:08:35

in in the direction that you fear how

00:08:38

long will it take to get

00:08:40

there well of course I don't personally

00:08:43

know how exactly things will play out I

00:08:46

can't see the future and I can give you

00:08:48

a feeling though of how I expect it to

00:08:50

feel how I expect it to feel like when

00:08:53

it happens the way I expect it to feel

00:08:55

is kind of like if you play chess

00:08:58

against a grandmas now I'm really bad at

00:09:00

chess I'm I'm not good at chess at all

00:09:03

but I you know I can play you know a

00:09:04

little bit of an amateur game and then

00:09:07

but when you play against a Grandmaster

00:09:08

or someone who's much much much better

00:09:10

than you the way it feels it's not like

00:09:12

you're having a heroic battle against

00:09:15

the Terminator you're having this

00:09:16

incredible back and forth and then you

00:09:18

lose no it feels more like you think

00:09:21

you're playing well you think everything

00:09:23

is okay and then suddenly you lose in

00:09:25

one move and you don't know why this is

00:09:27

what it feels like to play chess like

00:09:29

against a grandm and this is what it's

00:09:30

going to feel like for Humanity to play

00:09:32

against AGI what's going to happen is

00:09:35

not some dramatic battle that you know

00:09:36

the Terminators rise up and try to

00:09:38

destroy Humanity no it will be things

00:09:41

get more and more

00:09:43

confusing more and more jobs get

00:09:45

automated faster and faster more and

00:09:47

more technology gets built which no one

00:09:48

even quite knows how the technology

00:09:50

works there will be mass media movements

00:09:53

that don't really make any sense like do

00:09:54

we really know the truth of what's going

00:09:56

on in the world right now even now with

00:09:58

social media do you or I really know

00:10:00

what's going on well how much of this is

00:10:02

fake how much of it is generated you

00:10:04

know with AI or other methods we don't

00:10:06

know and this will get much worse

00:10:08

imagine if you have extremely

00:10:09

intelligent systems much smarter than

00:10:12

humans that can generate any image any

00:10:14

video anything trying to manipulate you

00:10:17

well and being able to develop new

00:10:18

technologies to interfere with politics

00:10:21

the way I expect it will go is that

00:10:23

things will seem like mostly normal just

00:10:25

like weird just like things are getting

00:10:27

weirder and weirder and then one day

00:10:30

we will just not be in control anymore

00:10:32

just it won't be dramatic there won't be

00:10:34

a fight there won't be a war it will

00:10:36

just be one day the machines are in

00:10:40

control and not us and and even if there

00:10:42

is a fight yeah sorry to even if there

00:10:44

is a fight or a war they've handed us

00:10:47

the gun and the bullets and we've done

00:10:48

it I mean it's us that might might do

00:10:51

all of this precipitated by being

00:10:54

controlled in some way absolutely

00:10:56

possible I don't think an AI would need

00:10:57

to use humans for that cuz you know it

00:10:59

could develop extremely advanced

00:11:00

technology but it's totally possible

00:11:02

humans are not secure it is absolutely

00:11:04

possible to manipulate humans like you

00:11:06

know everyone knows this you humans are

00:11:08

not immune to propaganda not immune to

00:11:10

mass movements imagine if you know an an

00:11:13

AGI gives Kim Jong Un the call and says

00:11:16

hey I'm going to make your country run

00:11:18

extremely well and tell you how to build

00:11:19

super weapons in return do me this favor

00:11:22

I mean Kim jamong is going to think

00:11:23

that's great and it's very easy to gain

00:11:27

power if you're extremely intelligent if

00:11:29

you're capable of manipulating people of

00:11:32

developing new techologies weapon

00:11:34

trading to on the stock market to make

00:11:35

tons of money well yeah you can do

00:11:38

whatever you want so you're sounding the

00:11:41

alarm Jeffrey Hinton seen as the founder

00:11:45

or father or Godfather of AI he's

00:11:47

sounding the alarm and has distanced

00:11:50

himself from a lot of his previous

00:11:53

statements others in the mainstream are

00:11:56

coming out heavily credentialed people

00:11:58

who who are the real deal when it comes

00:12:00

to a AIS saying we need guard rails we

00:12:04

need regulation we need to be careful

00:12:05

maybe we should stop

00:12:07

everything yet open AI Microsoft Deep

00:12:12

Mind these are companies but then you

00:12:14

have governments investing in this

00:12:16

everybody's still

00:12:17

rushing

00:12:19

forward hurtling forward towards a

00:12:23

possible Doom why are they still doing

00:12:26

it despite these very legitimate and

00:12:28

strong warning is it only about the

00:12:30

bottom line and money and competition or

00:12:32

is there more to it this is a great

00:12:34

question and I really like how you

00:12:36

phrase you said they were rushing

00:12:38

towards because this is really the

00:12:39

correct way of looking at this it's not

00:12:42

that it is not possible to do this well

00:12:44

it is not that it's not possible to

00:12:46

build safe AI I think this is possible

00:12:48

it's just really hard it takes time it's

00:12:51

the same way that it's much easier to

00:12:52

build a nuclear reactor that melts down

00:12:54

than to build a nuclear reactor that is

00:12:56

stable like of course this is just hard

00:12:59

so you need time and you need resources

00:13:01

to do this but unfortunately what we're

00:13:03

we're in the situation right now is

00:13:05

we're currently in a situation right now

00:13:07

where at least here in the UK there is

00:13:10

currently more regulation on selling a

00:13:12

sandwich to the public than to develop

00:13:15

potentially lethal technology that could

00:13:18

kill every human on earth this is true

00:13:20

this is this is the current case and a

00:13:23

lot of this is because of slowdown it's

00:13:25

just you know governments are slow

00:13:27

people don't want and vested interest

00:13:29

you make a lot of money by pushing AI

00:13:32

pushing AI further makes you a lot of

00:13:35

money it gets you famous on Twitter you

00:13:37

know look how much the like these people

00:13:39

are rock stars you know people like Sam

00:13:40

alman's a rock star on Twitter you know

00:13:43

people love these people they're like oh

00:13:44

yeah they're bringing the future they're

00:13:46

making big money so they must be good

00:13:48

but like I mean it's just not that

00:13:50

simple unfortunately we're in a

00:13:52

territory where we all agree somewhere

00:13:55

in the future there's a precipice which

00:13:59

we will fall down if we continue we

00:14:01

don't know where it is we don't maybe

00:14:04

it's far away maybe it's very close and

00:14:06

my opinion is if you don't know where it

00:14:08

is you should stop well other people who

00:14:10

you know gain money power or just

00:14:13

ideological points like a lot of these

00:14:15

people is very important to understand

00:14:17

do this because they truly believe like

00:14:19

a religion they believe in transhumanism

00:14:22

in in the Glorious future where AI will

00:14:26

love us and so on like so there's many

00:14:29

reasons but I mean yeah I mean the

00:14:30

cynical take is just I could be making a

00:14:33

lot more money right now if I was just

00:14:35

pushing AI I could get a lot more money

00:14:37

than I have right now how do we do

00:14:39

anything about this without just

00:14:41

deciding

00:14:43

to cut the uny internet cables and blow

00:14:46

up the satellites in space and just

00:14:48

start again how do you actually because

00:14:50

this is a technical problem and it's

00:14:53

also a moral and ethical problem so

00:14:56

where do you even begin right now or is

00:14:58

it too

00:14:59

late so the weirdest thing about the

00:15:02

world to me right now as someone who's

00:15:04

deep into this is that things are going

00:15:07

very very bad we have you know crazy you

00:15:12

know just corporations with zero

00:15:14

oversight just plowing billions of

00:15:16

dollars into going as fast as possible

00:15:19

with no oversight with no accountability

00:15:22

which is about as bad as it could be but

00:15:24

somehow we haven't yet lost it's not yet

00:15:29

over it could have been over there's

00:15:31

many things where it could be over

00:15:32

tomorrow but it's not yet there is still

00:15:35

hope there is still hope I don't know if

00:15:37

there's going to be hope in a couple

00:15:38

years or even in one year but there

00:15:40

currently still is Hope Oh wait hold on

00:15:42

one year I mean

00:15:43

that's come on man I mean we're probably

00:15:47

going to put out this interview like a

00:15:48

couple of weeks after we record it a few

00:15:51

months will pass we could all be dead by

00:15:53

the time you I know this gets 10,000

00:15:55

views I mean just just for explain this

00:15:58

timeline line one year why one year why

00:16:00

why is it going so fast that even one

00:16:02

year would be too far ahead explain that

00:16:04

I'm not saying one year is like

00:16:06

guaranteed by any means I think it's

00:16:07

unlu unlikely but it's not impossible

00:16:09

and this is important to understand is

00:16:12

that Ai and computer technology is an

00:16:14

exponential it's like covid this is like

00:16:17

saying in February you know a million

00:16:20

covid infections that's impossible that

00:16:22

can't happen in six months and it

00:16:24

absolutely did this is kind of how AI is

00:16:28

as well exponentials look slow they look

00:16:31

like you don't go have one infected two

00:16:33

infected four infected that's not so bad

00:16:37

but then you have 10,000 20,000 40,000

00:16:41

you know

00:16:42

100,000 yeah you know within a single

00:16:45

week and this is how te this technology

00:16:47

works as well is that as our computers

00:16:50

get there's something called Moors law

00:16:52

which is it's not really a lot it's more

00:16:53

like an observation that every two years

00:16:56

our computers get about you know there's

00:16:58

some details but about twice as powerful

00:17:01

so that's an exponential and our Tech

00:17:03

and it's not just our computers are

00:17:04

getting more powerful our software is

00:17:06

getting better our AIS are getting

00:17:08

better our data is getting better more

00:17:10

money is coming into this field we are

00:17:12

on an exponential this is why things can

00:17:14

go so fast so while I'm not like you

00:17:17

know it would be weird if we would all

00:17:19

be dead in one year it is physically

00:17:22

possible you can't rule it out if we

00:17:24

continue on this path the powerful

00:17:27

people who can do something about this

00:17:30

especially when it comes to regulation

00:17:31

when you saw those Congressman speaking

00:17:33

to Sam Alman they didn't seem to know

00:17:36

what the hell they were talking about so

00:17:38

how frustrating is it for you that the

00:17:40

people who can make a difference have

00:17:42

zero clue about what's really going on

00:17:44

and and more important than that they

00:17:47

didn't seem to want to actually know

00:17:50

they had weird questions that made no

00:17:52

sense and so you're thinking okay these

00:17:55

guys are in charge I mean no wonder the

00:17:56

AI is going to come and wipe us all out

00:17:58

maybe maybe we deserve

00:18:00

it well I wouldn't go that far but um

00:18:04

this used to annoy me a lot this used to

00:18:06

be extremely frustrating um but I've

00:18:09

come to I've come to peace with it to a

00:18:10

large degree because the thing that I've

00:18:12

really found is that understanding the

00:18:14

world is hard understanding complex

00:18:17

topics and technology is hard not just

00:18:18

because they they're complicated but

00:18:20

also because people have lives and this

00:18:22

is okay this is normal people have

00:18:23

families they have responsibilities they

00:18:26

have there's a lot of things people have

00:18:28

to do deal with and I don't shame people

00:18:30

for this you know like you know I have

00:18:32

turkey you know with my family over

00:18:34

Thanksgiving and whatever and you know

00:18:35

my aunts and uncles look they have their

00:18:37

own lives going on they maybe don't

00:18:39

really have time you know to listen to

00:18:41

me give them a rant about it so I don't

00:18:43

so I have a lot of love and a lot of

00:18:45

compassion for that things are hard this

00:18:48

is of course doesn't mean it that solves

00:18:50

the problem but I'm just trying to say

00:18:52

that like it is of course frustrating to

00:18:55

some degree that there are no adults in

00:18:57

the room this is this is how I would see

00:18:59

it is that there is sometimes a belief

00:19:03

that somewhere there is someone who

00:19:05

knows what's going on there's an adult

00:19:07

who's got under control you know someone

00:19:09

in the government they've got this under

00:19:11

control and as someone who's tried to

00:19:14

find that person I could tell you this

00:19:15

person does not exist the truth is is

00:19:18

the fact that anything works at all in

00:19:19

the world is kind of a miracle it's kind

00:19:21

of amazing that anything works at all

00:19:23

with how chaotic everything is but the

00:19:25

truth is is that there are quite a lot

00:19:27

of people who like

00:19:28

who want the world to be good you know

00:19:30

they might not have the right

00:19:31

information they might be confused they

00:19:34

might be getting lobbied by various

00:19:35

people with bad intentions but like most

00:19:38

people want their families to live and

00:19:41

have a good life most people don't want

00:19:44

bad things to happen most people want

00:19:47

other people to be happy and safe and

00:19:50

luckily for us most normal people so not

00:19:54

Elites not necessarily politicians or

00:19:56

technologists most normal people yeah do

00:19:59

have the right intuition around AI where

00:20:02

they see like wow that seems really

00:20:03

scary let's be careful with this and

00:20:07

this is what gives me hope so when I

00:20:09

think about politicians and them not

00:20:10

being in charge I think this is now our

00:20:13

responsibility as citizens of the world

00:20:15

that we have to take this into our own

00:20:16

hands we can't wait for people to save

00:20:18

us we have to make them save us we have

00:20:20

to make these things happen we have to

00:20:21

you know we have to make our voices hurt

00:20:24

we have to say hey how the hell are you

00:20:26

letting this happen like one of the

00:20:29

Beautiful Things is that you know to a

00:20:31

large degree politicians can be moved

00:20:35

they can be reasoned with and they can

00:20:36

be moved by the voters you can vote them

00:20:38

out of office that's a good argument for

00:20:40

democracy that's a great argument for

00:20:41

democracy that's that's wonderful you

00:20:43

know democracy is the worst system

00:20:45

except for all the other ones yeah um so

00:20:49

to the point of people's feeling and and

00:20:52

I asked about this at the very beginning

00:20:53

that intuitive feeling of like

00:20:55

something's up here there's something

00:20:57

ominous

00:20:59

there did seem to be a little bit of a

00:21:00

plateau with something like chat GPT so

00:21:04

initially people were very anxious very

00:21:07

surprised very wowed by what this thing

00:21:09

could do it could write your University

00:21:11

thesis and whatever it could you know do

00:21:13

all these these fancy gimmicks they seem

00:21:15

like magic tricks but then once the hype

00:21:20

died down a little bit uh people began

00:21:23

to input new things ask maybe better

00:21:26

questions and you could see some of the

00:21:28

limitations of something like you know

00:21:30

chat G GPT and its uh forerunners and

00:21:34

that led a lot of people to say well I

00:21:36

mean okay sometimes this thing just

00:21:37

sounds like a PR department or an HR

00:21:39

department in a company sometimes it it

00:21:42

actually it's there to detect plagiarism

00:21:44

but sometimes it feels like a

00:21:46

plagiarized like college um paper which

00:21:50

led to and this is anecdotally a lot of

00:21:53

friends of mine going ah maybe this

00:21:54

thing maybe we're okay for a while

00:21:56

because this thing has severe

00:21:58

limitations address that for me because

00:22:00

a lot of people are still sort of like

00:22:01

well I know there was the hype but now

00:22:03

I'm not so sure tell me about that so

00:22:07

there is a story I'm not sure if the

00:22:08

story is actually true or not but it's a

00:22:10

good metaphor where if you take a frog

00:22:13

and you put it into a pot of water you

00:22:16

know cold pot of water the Frog will sit

00:22:18

there happily if you slowly and turn up

00:22:20

the heat on your pot the Frog will sit

00:22:23

there there no problem and if you do it

00:22:25

very slowly you very slow slly slowly

00:22:28

increase the temperature the Frog will

00:22:30

get used to the temperature and won't

00:22:32

jump out until the water boils and the

00:22:34

frog dies I think this is what is

00:22:37

happening with people is that um people

00:22:41

are extremely good at making things

00:22:44

which are crazy normal is that if it's a

00:22:47

normal thing if it's a thing all your

00:22:49

friends do then it just becomes normal

00:22:51

this is like during war why people can

00:22:53

Slaughter other people because if all

00:22:55

your friends are doing it well it's

00:22:56

normal it's a yeah you slaughter people

00:22:58

it's normal you killing people is fine

00:23:01

this is how it can happen and the same

00:23:02

thing applies here is that well okay you

00:23:05

can talk to your computer now like sure

00:23:07

we can argue about oh chat jpt it's not

00:23:09

that smart you could talk to your

00:23:10

computer like slow down if this was a

00:23:13

sci-fi movie from 20 years ago everyone

00:23:16

would be yelling at the screen like what

00:23:18

the hell are you doing like this thing

00:23:19

is obviously like crazy like what the

00:23:21

hell is going on but because it's you

00:23:24

know available now you know cheaply

00:23:26

online he doesn't feel special so the

00:23:30

way to address this is I think a lack of

00:23:34

coordinated campaigning effort what I

00:23:36

mean by this is is that the general when

00:23:39

we think about our civilization not just

00:23:41

individual people when we think about

00:23:42

our civilization how does our

00:23:44

civilization deal with problems how does

00:23:47

it decide which problems to address

00:23:50

because there's always so many problems

00:23:53

you could be putting your effort on how

00:23:55

does it decide which one to pay

00:23:56

attention to and this is actually very

00:23:58

complicated and it can be because of a

00:24:01

natural catastrophe or a war or whatever

00:24:04

it can be because of some stupid fashion

00:24:06

hype just like some viral video on Tik

00:24:08

Tok makes everyone freak out sometimes

00:24:10

yes but usually if you actually want

00:24:13

your civilization to address a problem a

00:24:16

big problem it takes long hard grinding

00:24:20

effort from people trying to raise this

00:24:24

to saliency to raise it to attention

00:24:27

because again people have lives you know

00:24:29

like most people don't have time to go

00:24:32

go online and read huge books about AI

00:24:34

safety and like oh how do we integrate

00:24:36

chat gbt or how do we deal with like the

00:24:38

safety TR they don't have time for that

00:24:40

of course they don't and I'm not trying

00:24:42

to judge these people I understand it's

00:24:44

not their job in a good World there

00:24:46

should be a group of people that deals

00:24:49

with this the problem is they don't

00:24:52

really exist before we go I'm glad you

00:24:55

mentioned that people don't know where

00:24:57

to look look if there was one resource

00:24:59

that you could Point people in the

00:25:02

direction of so that they can educate

00:25:04

themselves about the reality of the

00:25:08

situation and can bring themselves up to

00:25:11

speed that would be

00:25:13

what there's not one who I think has the

00:25:16

whole thing which is a big problem

00:25:17

someone should make that resource if

00:25:19

someone make that resource please let me

00:25:21

know but what I would probably Point

00:25:22

people towards is control AI which is a

00:25:26

group of people who I'm also involved

00:25:27

with who are campaigning for exactly

00:25:30

these issues who are trying to bring

00:25:31

Humanity together to solve these

00:25:34

problems because this is a problem that

00:25:36

not you or me can solve no human can

00:25:39

solve these problems we're dealing with

00:25:40

right now this is a problem that

00:25:42

Humanity has to solve right that our

00:25:44

civilization needs to solve and I think

00:25:45

our civilization can do this but it

00:25:48

won't do it without our help it won't

00:25:50

help it won't happen without us working

00:25:52

together so if there's one thing I can

00:25:54

go go on Twitter or Google or whatever

00:25:56

go to control AI

00:25:58

and support them listen to what they to

00:26:00

say and this is the campaign I'm behind

00:26:04

well I support them okay we'll put the

00:26:06

link also in the YouTube description uh

00:26:08

if anybody wants to check it out Conor

00:26:10

you have a brilliant mind and I'm really

00:26:12

grateful that we got to talk thank you

00:26:14

very much for joining us on the

00:26:15

interview thank you so much take

00:26:26

care

Description:

Lauded for his groundbreaking work in reverse-engineering OpenAI's large language model, GPT-2, AI expert Connor Leahy tells Imran Garda why he is now sounding the alarm. Leahy is a hacker, researcher and expert on Artificial General Intelligence. He is the CEO of Conjecture, a company he co-founded to focus on making AGI safe. He shares the view of many leading thinkers in the industry, including the godfather of AI Geoffrey Hinton, who fear what they have built. They argue that recent rapid unregulated research combined with the exponential growth of AGI will soon lead to catastrophe for humanity - unless an urgent intervention is made. 00:00 AI guru 02:11 What is AGI, and what are the risks it could bring? 03:51 "Nobody knows why AI does what it does" 05:41 From an AI enthusiast to advocating for a more cautious approach to AI development 07:58 What does Connor expect to happen when we lose control to AGI? 11:40 "People like Sam Altman are rockstars" 14:38 Connor's vision of a safe AI future 15:24 Imran: "One year left?" 17:26 "Normal people do have the right intuition about AI" 20:58 ChatGPT, limitations of AI, and a story about a frog 24:53 Control AI Connor recommends we all visit Control AI: https://controlai.com/deepfakes Control AI / Twitter: @ai_ctrl Control AI / TikTok: @ctrl.ai Control AI / Instagram: @ai_ctrl Subscribe: https://www.trtworld.com//subscribe Livestream: http://trt.world/ytlive Facebook: http://trt.world/facebook Twitter: http://trt.world/twitter Instagram: http://trt.world/instagram Visit our website: http://trt.world

Preparing download options

Popular

HD video

Only sound

All

* — If the video is playing in a new tab, go to it, then right-click on the video and select "Save video as..."

** — Link intended for online playback in specialized players

Questions about downloading video

How can I download "Why this top AI guru thinks we might be in extinction level trouble | The InnerView" video?

http://unidownloader.com/ website is the best way to download a video or a separate audio track if you want to do without installing programs and extensions.
The UDL Helper extension is a convenient button that is seamlessly integrated into YouTube, Instagram and OK.ru sites for fast content download.
UDL Client program (for Windows) is the most powerful solution that supports more than 900 websites, social networks and video hosting sites, as well as any video quality that is available in the source.
UDL Lite is a really convenient way to access a website from your mobile device. With its help, you can easily download videos directly to your smartphone.

Which format of "Why this top AI guru thinks we might be in extinction level trouble | The InnerView" video should I choose?

The best quality formats are FullHD (1080p), 2K (1440p), 4K (2160p) and 8K (4320p). The higher the resolution of your screen, the higher the video quality should be. However, there are other factors to consider: download speed, amount of free space, and device performance during playback.

Why does my computer freeze when loading a "Why this top AI guru thinks we might be in extinction level trouble | The InnerView" video?

The browser/computer should not freeze completely! If this happens, please report it with a link to the video. Sometimes videos cannot be downloaded directly in a suitable format, so we have added the ability to convert the file to the desired format. In some cases, this process may actively use computer resources.

How can I download "Why this top AI guru thinks we might be in extinction level trouble | The InnerView" video to my phone?

You can download a video to your smartphone using the website or the PWA application UDL Lite. It is also possible to send a download link via QR code using the UDL Helper extension.

How can I download an audio track (music) to MP3 "Why this top AI guru thinks we might be in extinction level trouble | The InnerView"?

The most convenient way is to use the UDL Client program, which supports converting video to MP3 format. In some cases, MP3 can also be downloaded through the UDL Helper extension.

How can I save a frame from a video "Why this top AI guru thinks we might be in extinction level trouble | The InnerView"?

This feature is available in the UDL Helper extension. Make sure that "Show the video snapshot button" is checked in the settings. A camera icon should appear in the lower right corner of the player to the left of the "Settings" icon. When you click on it, the current frame from the video will be saved to your computer in JPEG format.

What's the price of all this stuff?

It costs nothing. Our services are absolutely free for all users. There are no PRO subscriptions, no restrictions on the number or maximum length of downloaded videos.