Generative A.I - We Aren’t Ready.

Kyle Hill
4 Mar 202416:10

Summary

TLDRThe video explores the concept of the 'Dark Forest Internet,' where the proliferation of generative AI has made it increasingly difficult to discern genuine human interactions and content online. Drawing parallels to the three-body problem in Chinese sci-fi, it argues that the internet is becoming a hostile environment where human users are forced to retreat into private spaces to avoid digital predators like bots, trolls, and synthetic content. The video warns that without robust verification systems, the internet risks being overrun by AI-generated misinformation and manipulation, challenging our ability to pass a 'reverse Turing test' and prove our humanity in an increasingly synthetic digital world.

Takeaways

  • 👽 The Dark Forest theory suggests that intelligent civilizations remain silent to avoid being preyed upon by more advanced ones, similar to how real users on the internet hide from digital predators.
  • 🌳 The internet is becoming a 'Dark Forest' due to the rise of bots, AI, misinformation, and other digital threats, making it harder to discern authentic human interactions.
  • 🤖 Generative AI, like ChatGPT, is rapidly generating synthetic content at an unprecedented scale, accelerating the 'darkening' of the internet.
  • 🕵️‍♀️ The Turing test, which measures a machine's ability to exhibit human-like behavior, has been effectively passed by large language models like ChatGPT.
  • ⚖️ A 'Reverse Turing Test' is emerging, where AI systems must prove they are human to distinguish themselves from synthetic content.
  • 🚨 Without robust verification systems, the internet risks being overrun by AI-generated spam, scams, and misinformation.
  • 👥 Meeting in person, institutional verification, triangulating objective reality through communication, and embracing algorithmic incoherence are proposed ways for humans to signal their authenticity online.
  • 🎭 Creating unique internet culture, such as memes and slang, could help humans stay ahead of AI and maintain a distinct online presence.
  • ⚡ While AI promises amazing outcomes like improved education and medical breakthroughs, it also poses significant risks, such as voice cloning scams, if not approached cautiously.
  • 🔮 The script warns that humanity is not ready for the rapid proliferation of AI and urges careful consideration of the next steps to avoid getting lost in the 'Dark Forest' of the internet.

Q & A

  • What is the 'Dark Forest' theory presented in the script?

    -The 'Dark Forest' theory refers to the idea that intelligent civilizations intentionally hide their presence and remain silent in the cosmos to avoid being preyed upon by more advanced civilizations. The script draws a parallel between this theory and the current state of the internet, where real human users are increasingly retreating into private spaces to avoid digital predators like bots, advertisers, and misinformation.

  • How is generative AI contributing to the 'Dark Forest' internet?

    -According to the script, generative AI models like ChatGPT are flooding the internet with synthetic content at an unprecedented scale. Companies and bad actors can easily create thousands of AI-generated accounts, websites, articles, and social media posts, making it increasingly difficult to discern real human-generated content from synthetic content.

  • What is the 'reverse Turing test' mentioned in the script?

    -The 'reverse Turing test' refers to a modification of Alan Turing's original imitation game, where the objective is reversed. Instead of machines trying to prove they are human, technologies and language models will be tasked with identifying real human-generated content amidst the growing synthetic content on the internet.

  • What are some practical advice given in the script for humans to signal their humanity online?

    -The script suggests several tactics, including: 1) Meeting people in real life ('meat space'), 2) Institutional verification of one's identity, 3) Triangulating objective reality by communicating with others, 4) Becoming 'algorithmically incoherent' by using internet-specific language and memes that AI models cannot keep up with.

  • Why does the script mention the potential need for institutional verification of one's identity?

    -The script suggests that institutional verification of one's identity, although dystopian, may become unavoidable in an age of generative AI. This could involve having to verify one's identity in person before registering a website, writing an article, or even posting on social media to ensure that the content is human-generated and not synthetic.

  • How does the script describe the potential negative outcomes of generative AI proliferation?

    -The script warns of potential harmful outcomes, such as phone scammers using synthesized voices to scam people, deepfakes of public figures causing chaos, and the inability of platforms like Twitter to sort synthetic content from real content. The author expresses concerns that humanity is not ready for the rapid advancement of generative AI.

  • What is the author's perspective on the future of human culture online in the age of generative AI?

    -The author suggests that human culture, such as internet-specific language, memes, and in-group dialects, may be one aspect that could outpace AI culture and help signal one's humanity online. However, the overall tone is one of concern and a warning about the potential dangers of generative AI proliferation if not properly addressed.

  • How does the script connect the 'Dark Forest' theory to the current internet landscape?

    -The script draws a parallel between the 'Dark Forest' theory in which intelligent civilizations hide their presence to avoid predators, and the current state of the internet where real human users are retreating into private spaces and curated communities to avoid digital predators like bots, misinformation, and synthetic content.

  • What is the significance of the 'capture' example given in the script?

    -The 'capture' example is used to illustrate a form of reverse Turing test, where systems like websites need to determine if the user is a real human or a bot. The script suggests that as generative AI proliferates, such systems will become increasingly important to protect online spaces from being overrun by synthetic content and spam.

  • How does the script characterize the potential positive outcomes of generative AI?

    -While the script primarily focuses on the potential negative consequences of generative AI, it acknowledges that the technology could lead to amazing outcomes, such as providing free world-class education to every child and potentially finding a cure for cancer through the use of large intelligent systems.

Outlines

00:00

🌐 The Dark Forest Theory of the Internet

This paragraph introduces the Dark Forest Theory of the Internet, as proposed by Yansy Strickler. It explains that just like the Dark Forest Theory in the sci-fi novel 'The Three-Body Problem' where civilizations hide from each other to avoid being preyed upon, human users on the internet are increasingly retreating to private spaces to avoid digital predators like bots, advertisers, and trolls. The declining realness of the web is attributed to this phenomenon.

05:00

🤖 The Rise of Generative AI and Its Impact

This paragraph discusses the rapid advancement of generative AI technology, which can create new text, images, videos, and sounds from training data. It highlights how generative AI is exacerbating the Dark Forest Internet by enabling the creation of synthetic content at an unprecedented scale. The example of an SEO Heist, where AI was used to generate 1,800 articles to divert web traffic, is cited as an indication of what's to come. The paragraph emphasizes the need to address this challenge before the internet becomes overwhelmed with misinformation and synthetic content.

10:03

🧠 Passing the Reverse Turing Test

This paragraph explores the concept of the Reverse Turing Test, where the objective is for machines to prove that they are human. It contrasts with the original Turing Test, where the goal was for machines to pass as thinking entities. The paragraph discusses practical advice from cultural anthropologist Maggie Appleton on how humans can signal their humanity online in the age of generative AI. Strategies include meeting in person, institutional verification, triangulating objective reality, and becoming algorithmically incoherent by using internet-specific language and memes.

15:05

🚨 A Cautionary Conclusion

The final paragraph serves as a cautionary conclusion. While acknowledging the potential benefits of AI in areas like education and healthcare, it expresses concern about the harmful outcomes that could arise if the proliferation of generative AI is not carefully managed. The paragraph warns that society is not ready for the challenges posed by this technology and emphasizes the need for thoughtful consideration of the next steps to prevent getting lost in the darkness of the Dark Forest Internet.

Mindmap

Keywords

💡Three-Body Problem

The Three-Body Problem is a famous unsolved problem in physics, referring to the impossible task of predicting the motion of three gravitationally interacting bodies over time. In the context of the video, it refers to the Chinese sci-fi novel 'The Three-Body Problem' by Cixin Liu, which introduces the concept of the 'Dark Forest' theory of the universe - that intelligent civilizations hide from each other due to the threat of destruction if revealed.

💡Dark Forest Theory

The Dark Forest theory, as explained in the video, is a hypothetical stance that the universe is teeming with intelligent life, but all civilizations remain undetected and non-communicative to avoid being preyed upon by more advanced ones. This theory serves as a metaphor for the Internet, where real human interactions are increasingly hidden from the noise, bots, and malicious entities that have overtaken the open web.

💡Generative AI

Generative AI refers to artificial intelligence models capable of generating new data, such as text, images, audio, or video, based on the training data they have ingested. The video focuses on the implications of generative AI models like ChatGPT, which can produce human-like content at an unprecedented scale, exacerbating the problem of distinguishing real from synthetic information online.

💡Turing Test

The Turing Test, proposed by Alan Turing in 1950, is a test to determine if a machine can exhibit intelligent behavior indistinguishable from a human. The video suggests that generative AI has effectively passed this test, blurring the lines between human and machine-generated content, and necessitating a 'Reverse Turing Test' to identify genuine human outputs.

💡Reverse Turing Test

A Reverse Turing Test is a modified version of the original Turing Test, where the goal is for machines or technologies to determine if an entity is human or not. The video posits that as generative AI proliferates, we will need systems akin to a Reverse Turing Test to separate authentic human content from synthetic outputs across various online platforms.

💡SEO Heist

An 'SEO Heist' refers to the tactic mentioned in the video, where a company used AI to rapidly generate thousands of articles based on a competitor's website structure, effectively stealing search engine traffic and rankings. This example illustrates how generative AI can be leveraged for unethical practices that undermine authentic online content and experiences.

💡Human Signaling

Human signaling refers to the methods or tactics proposed in the video for humans to distinguish themselves from AI-generated content online. These include meeting in physical spaces, institutional verification of identity, triangulating objective reality through shared experiences, and embracing algorithmic incoherence by using internet-specific language and memes that AI models cannot replicate.

💡Algorithmically Incoherent

To be 'algorithmically incoherent' means to create content or communication that does not follow the statistical patterns and coherence expected by language models and AI algorithms. The video suggests that embracing internet slang, memes, and rapidly evolving cultural references can serve as a way for humans to signal their authenticity in contrast to the generalized outputs of generative AI.

💡Dark Forest Internet

The 'Dark Forest Internet' is the central metaphor used in the video, likening the modern internet landscape to the Dark Forest theory from the sci-fi novel. Just as civilizations hide from each other in the cosmic dark forest, real human interactions and authentic content on the internet are increasingly retreating into private, curated spaces to avoid the noise, bots, and synthetic outputs proliferating in the open web.

💡Misinformation

Misinformation, or the spread of false or misleading information, is a major concern highlighted in the video with the rise of generative AI. The ability to rapidly generate realistic-looking content at scale, without the need for human involvement, raises alarming possibilities for the dissemination of misinformation and disinformation campaigns on a unprecedented level across the internet.

Highlights

In the three-body problem Chinese sci-fi, author Leu Sushin offers a solution to the famous Fermi Paradox - the universe is a 'dark forest' where intelligent life is hidden and hostile, as broadcasting one's presence invites predation from advanced civilizations.

Yansy Strickler's 'Dark Forest theory of the internet' explains the declining realness of the web, where real human users are hiding in private spaces to avoid digital predators like bots, advertisers, and trolls.

Generative AI's rapid spread is making the internet's 'dark forest' even darker and more dangerous, as it becomes harder to distinguish real human content from synthetic AI-generated content.

Misinformation expert Nahschik estimates that the majority of online content will be synthetic within the next year, driven by the exponential growth of tools like ChatGPT.

Companies are already using AI to rapidly generate content at scale, outpacing human capabilities, as demonstrated by an 'SEO Heist' that stole web traffic from a competitor.

Passing the famous Turing test, where a machine convincingly imitates human intelligence, is no longer the primary challenge - instead, humans now face a 'reverse Turing test' of proving their own humanity amidst the AI-generated content.

As AI proliferates, non-human systems will increasingly be tasked with identifying genuine human content through 'reverse imitation games', such as CAPTCHA tests currently used to filter out bots.

Without reliable systems to distinguish synthetic from human-generated content, the internet risks becoming a cacophony of AI-driven noise and misinformation.

Maggie Appleton offers practical advice for 'human signaling' online: meet people in the physical world, pursue institutional verification of identity, triangulate objective reality through communication, and create 'algorithmically incoherent' internet culture like memes and slang that outpaces AI models.

Showing up in the physical world and meeting people face-to-face may be the quickest way to prove one's humanity to others in an AI-driven online environment.

Institutional verification of identity, potentially through in-person checks before posting online, may become a necessary if dystopian solution to combat AI-generated misinformation.

Communicating grounded experiences that reflect unique human traits like sensory perception and community belonging can help distinguish humans from current AI models trained on limited datasets.

Creating 'algorithmically incoherent' internet culture through memes, slang, and evolving language could signal one's humanity by outpacing the generalization tendencies of language models.

While AI promises amazing outcomes like widespread education and medical breakthroughs, we are not ready for the potential harms like voice cloning scams and misinformation, necessitating careful consideration of next steps.

The rapid proliferation of generative AI risks plunging us into an online 'dark forest' where it becomes difficult to discern reality from AI-generated content, requiring new strategies to preserve human authenticity and signal our humanity.

Transcripts

00:01

in the three-body problem Chinese sci-fi

00:04

author Leu Sushin offers a solution to

00:06

the famous firm Paradox the universe is

00:10

not empty of life rather the universe he

00:13

writes is a dark Forest there is life

00:17

throughout it but it's both hidden and

00:20

hostile why because any intelligent life

00:23

foolish enough to broadcast its presence

00:25

to the cosmos is immediately prayed upon

00:28

by other more advanced civilizations

00:30

patience and so the fmy Paradox isn't

00:33

one life is out there but it's

00:36

intentionally

00:39

silent as sci-fi as the theory sounds

00:42

you're more familiar with the idea than

00:43

you think every day we live more and

00:46

more of our lives on the internet in the

00:48

digital and every day that space is

00:51

flooding more and more with Bots

00:53

advertisers trolls data scrapers

00:55

clickbait influencers and Mindless

00:58

social media mobs looking for today's

01:00

main

01:01

character the internet feels steadily

01:03

more lifeless but that's because like

01:06

those alien civilizations the real human

01:09

users are hiding in private apps servers

01:12

and RSS feeds lest they be beset by

01:15

these digital

01:17

Predators this is yansy strickler's Dark

01:20

Forest theory of the internet something

01:22

to explain the declining realness of the

01:25

web and I know you feel it it's hard not

01:28

to when most photos are shopped

01:30

influence is bought engagement is a

01:33

meaningless number and every article

01:35

reads worse than a high schooler's first

01:37

essay unfortunately with the now

01:40

Unstoppable spread of generative AI the

01:44

forest is about to get a lot darker and

01:47

a lot more

01:51

dangerous when you and I last spoke

01:53

about generative AI or AI that can

01:56

generate new text images videos sounds

01:59

from training data I was merely

02:01

speculating on how this new technology

02:04

might change your digital life but you

02:08

know it already is don't you when's the

02:11

last time you clicked on anything

02:14

believed any headline or any social

02:16

media post not wanting to go through the

02:20

time and effort to check whether or not

02:22

it's ground truth whether or not it has

02:24

the right context whether or not it's

02:27

made by an actual person how could you

02:29

not feel overwhelmed in an increasingly

02:32

lifeless and dangerous

02:36

internet the Dark Forest internet and

02:38

generative AI is probably why you've

02:40

retreated to the places you still get

02:42

provably human interactions with people

02:45

you might even know private spaces like

02:47

text messages emails discords and slacks

02:51

there have always been advantages to

02:53

these more curated spaces of course but

02:55

as cultural Anthropologist Maggie

02:57

Appleton points out generative Ai and

02:59

large language models like chat GPT are

03:02

going to force us further into our

03:04

digital bunkers and impenetrable silos

03:07

echoey though they may be because the

03:09

Dark Forest internet is exponentially

03:12

expanding misinformation expert nah

03:14

schik estimates that the majority of

03:16

online content will be synthetic within

03:19

the next year case in point chat gbt and

03:23

its users are currently generating more

03:25

text than has ever appeared in every

03:28

physical book ever written every two

03:31

weeks so by now um language models have

03:34

turned uh into lots of very easy to use

03:36

products right you don't really need any

03:38

technical skills to use them so these

03:40

are a bunch of like very popular

03:42

copywriting apps that are out there in

03:43

the world here's just one of the

03:44

examples of what's coming from

03:46

Appleton's talk the expanding Dark

03:48

Forest and generative AI the reason I'm

03:50

making this video imagine that some

03:53

political lobbyists spin up 1,000

03:56

decently intelligent AIS that can

03:58

generate text and video and then they

04:00

tell each of them to go be influencers

04:03

the Bots then instantaneously make

04:05

accounts on most social media platforms

04:07

they generate their own websites they

04:09

publish independent books with

04:11

synthesized voices they make many

04:14

documentaries on YouTube they host each

04:16

other on podcasts individually they all

04:19

appear to make a reasonable human amount

04:21

of content but taken together these

04:24

human lobbyists have created an

04:26

automated content ring of lifeless

04:29

engagement at a scope and scale that

04:31

would take any one human a lifetime to

04:33

create and

04:35

curate this is all possible with the

04:37

technology that we have right now and it

04:40

appeared almost out of nowhere in just a

04:43

few months um but the point is that this

04:45

is incredibly easy to do at this point

04:47

with with no technical skills Appleton's

04:49

example isn't just hypothetical

04:51

companies are doing this right now in

04:54

the November of 2023 Elon musk's mistake

04:57

user Jake Ward tweeted this we pulled

05:00

off an SEO Heist that stole 3.6 million

05:03

total traffic from a competitor AI

05:05

allowed Ward to quickly and easily

05:07

export a competitor's sitemap turn their

05:09

list of URLs into article titles and

05:12

then in a matter of just hours create

05:15

1,800 articles from those titles that

05:18

directed web traffic away from that

05:20

competitor now imagine this kind of

05:23

strategy for every single corner of the

05:26

internet every business every chatbot

05:29

every influencer pulled off faster and

05:31

more efficiently than any human could

05:33

even

05:34

conceptualize this is what's

05:38

coming the Dark Forest

05:41

expands just over a year ago we weren't

05:44

sure any technology or indeed chatbot

05:47

could pass the famous Turing test but

05:50

now faster than we were ready for and

05:52

without our consent humanity is faced

05:56

with maybe a more interesting question

05:59

how how will we humans out here in the

06:02

dark looking for light pass the reverse

06:06

Turing

06:09

test Alan turing's seminal 1950 paper

06:12

Computing machinery and intelligence

06:15

opens with a simple question can

06:18

machines think turns out thinking is

06:21

pretty hard to Define so instead he

06:24

proposed a simple test an imitation game

06:28

he called it imagine that you're an

06:30

interrogator of two entities your job is

06:33

to determine through written text alone

06:36

which is the human and which is the

06:38

computer if a computer can fool you into

06:41

thinking it is the human then that

06:44

machine will have passed what came to be

06:45

known as the Turing test since touring

06:49

his test was the popularly understood

06:52

Benchmark for thinking machines and for

06:54

those of us outside the wired walls of

06:57

computer science the Turing test never

06:59

really seemed passable it was like The

07:02

Uncanny Valley problem for graphics

07:04

technology computers are always getting

07:07

closer and closer to rendering

07:08

believable humans but you can always

07:10

tell something is off turns out primate

07:13

brains naturally selected over millions

07:15

of years are pretty good at recognizing

07:18

faces but everything changed with chat

07:21

GPT suddenly hundreds of millions of

07:24

people around the world realized that

07:26

the Turing test had been passed

07:30

the large language model you can learn

07:32

exactly how it works in another video

07:34

now rates higher than human doctors on

07:36

bedside Manor scores better than 90% of

07:39

lawyers on the bar and 99% of graduate

07:42

students on the gr unsurprisingly every

07:46

sector of human endeavor that trades in

07:48

on human intelligence which is all of

07:50

them is now rushing to incorporate this

07:52

technology in some way I've said that

07:55

because of Technology like chat GPT

07:58

everything is about to change

08:00

but I don't just mean humans losing

08:01

their jobs or falling in love with

08:03

chatbots I mean that turing's

08:05

fundamental assumption in his imitation

08:07

game that a human will have to decide

08:10

what passes for human is no longer valid

08:13

in all cases a reverse Turing test is a

08:17

modification of the original where the

08:19

objective of one or more of the roles

08:21

has been reversed as AI proliferates

08:24

into every corner of the Dark Forest

08:26

internet other non-human machines will

08:28

increase inas inly be tasked with a

08:30

reverse imitation game where language

08:33

models and other technologies will try

08:35

to prove that they're human capture is

08:38

an example of a reverse Turing test it

08:41

may soon not be the case but right now

08:44

there are still no systems sophisticated

08:46

enough to reliably read and reproduce

08:49

distorted images of text and so the

08:52

capture computer decides that any

08:54

successful deciphering must be done by a

08:57

human a test like this is important

09:00

because without it you can imagine

09:02

websites being overrun by spam scams and

09:06

Bots but it seems inevitable now that

09:09

they will be the specific phrase that

09:12

llms use to identify themselves as an AI

09:16

language model has started to show up in

09:18

Amazon reviews Yelp reviews tweets and

09:21

Linkedin posts everywhere specifically

09:24

because we don't have the capture

09:26

equivalent protecting those spaces

09:29

to these systems the AI may as well be

09:32

human the time is coming and coming very

09:36

soon when we will need captal likee

09:38

systems in place socially politically

09:41

commercially to determine what is

09:44

actually human generated content lest

09:46

everything you see online just add to

09:49

the cacophony of a dead internet we

09:52

don't have these systems yet and if and

09:55

when we do they are likely to lag

09:57

dangerously behind what massively

09:59

incentivized generative AI will be able

10:02

to do by that point do you really think

10:05

a platform like Twitter will be able to

10:08

sort the synthetic from the Simeon and

10:11

make the Right Moves here when it's

10:13

basically just posting its way

10:15

through the

10:17

apocalypse so how do you tell other

10:21

humans and thinking machines that you

10:24

are in fact human on a dark Forest

10:28

internet

10:30

Maggie Appleton has some practical

10:32

advice for human signaling online in an

10:35

age of generative Ai and the first is

10:39

Right Here show up in meat space meet

10:43

other people go outside and actually

10:46

touch grass something no doubt was lost

10:50

when our digital lives became more

10:52

important than our physical ones so

10:54

let's go out there and reclaim it this

10:58

is not as fun or as accessible as the

11:01

internet before it went dark but it's

11:03

probably the quickest and easiest way to

11:06

prove that you are a human to other

11:09

humans the institutional verification of

11:12

your humanness feels the most dystopian

11:15

idea but it may be the most unavoidable

11:18

it would have to be something that goes

11:20

beyond a blue check mark that you pay a

11:22

rich man Bay before maybe before

11:25

registering a website writing an article

11:27

or even posting to social media you have

11:30

to show up in person somewhere and

11:32

verify that you are in fact you and not

11:35

just an image or voice that could be

11:37

easily

11:38

generated internet culture grew up

11:40

around hating this idea in principle but

11:43

it may be impossible to avoid when any

11:46

bad actor for free can fool millions of

11:49

people into thinking that the pope has

11:52

drip or deep fake of President declaring

11:55

nuclear

11:56

war the last two tactics Appleton offers

11:59

takes advantage of the fact that current

12:01

AI models are indeed machines machines

12:04

trained on certain data sets from

12:06

certain places and at certain times

12:09

knowing this we can come together online

12:11

and triangulate objective reality with

12:14

each other to prove our Humanity our

12:17

brains are constantly producing models

12:19

of the world and check those models

12:22

against sensory input current AI models

12:24

cannot do this they don't know any facts

12:27

people or world events that came into

12:29

being after their most recent round of

12:31

training and they don't feel the world

12:33

like you do they don't belong to

12:35

communities they don't reflect on

12:37

themselves and their history or enjoy

12:39

the richness of sensory experience by

12:43

communicating with each other online in

12:45

a way that reflects the aspects of

12:47

humanity apart from intelligence we can

12:50

simultaneously have more grounded

12:52

interactions and be sure that a human is

12:55

on the other side of the screen finally

12:58

we can distinguish ourselves in the dark

12:59

Forest by becoming what Appleton calls

13:02

algorithmically

13:04

incoherent large language models like

13:06

chat GPT only work by sequencing words

13:09

that are the most statistically likely

13:11

to go together in response to a prompt

13:13

based on vast amounts of training data

13:16

because of this these models hedge and

13:19

generalize in other words they basic

13:23

this gives real humans an opportunity to

13:25

reclaim one aspect of Internet culture

13:27

that is still the most fun creating

13:30

internet specific culture no language

13:33

model will be able to keep up with the

13:35

pace of weird internet lingo and memes

13:38

Appleton writes using jargon euphemistic

13:41

Emoji unusual phrases ingroup dialects

13:44

and memes of the moment will help signal

13:47

your Humanity end quote it's possible

13:51

that human culture could continue to

13:53

outpace AI culture like teenager culture

13:56

outpaces their parents Maybe Riz and Gat

14:01

will end up saving something uniquely

14:03

human online I wish I was able to end

14:06

this warning on a happier note I really

14:08

do and make no mistake I am 100% certain

14:12

that Ai and its proliferation will lead

14:15

to some amazing outcomes I think it

14:18

could provide free worldclass education

14:21

to every child in the world and it would

14:23

just be on their phone I think if

14:26

anything is going to find a cure for

14:27

cancer it's going to be one of these

14:29

large intelligent systems but for me it

14:33

is much easier to think of many more

14:36

much more harmful outcomes phone

14:40

scammers using your synthesized voice to

14:42

scam your parents

14:58

from a single

15:00

photo we are not ready for what's coming

15:04

and we need to think about our next

15:06

steps very carefully lest we all get

15:11

lost in the

15:19

[Music]

15:27

Darkness

15:29

[Music]

15:57

for

16:01

[Music]