Generative A.I - We Aren’t Ready.

in the three-body problem Chinese sci-fi

author Leu Sushin offers a solution to

the famous firm Paradox the universe is

not empty of life rather the universe he

writes is a dark Forest there is life

throughout it but it's both hidden and

hostile why because any intelligent life

foolish enough to broadcast its presence

to the cosmos is immediately prayed upon

by other more advanced civilizations

patience and so the fmy Paradox isn't

one life is out there but it's

intentionally

silent as sci-fi as the theory sounds

you're more familiar with the idea than

you think every day we live more and

more of our lives on the internet in the

digital and every day that space is

flooding more and more with Bots

advertisers trolls data scrapers

clickbait influencers and Mindless

social media mobs looking for today's

main

character the internet feels steadily

more lifeless but that's because like

those alien civilizations the real human

users are hiding in private apps servers

and RSS feeds lest they be beset by

these digital

Predators this is yansy strickler's Dark

Forest theory of the internet something

to explain the declining realness of the

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

to when most photos are shopped

influence is bought engagement is a

meaningless number and every article

reads worse than a high schooler's first

essay unfortunately with the now

Unstoppable spread of generative AI the

forest is about to get a lot darker and

a lot more

dangerous when you and I last spoke

about generative AI or AI that can

generate new text images videos sounds

from training data I was merely

speculating on how this new technology

might change your digital life but you

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

last time you clicked on anything

believed any headline or any social

media post not wanting to go through the

time and effort to check whether or not

it's ground truth whether or not it has

the right context whether or not it's

made by an actual person how could you

not feel overwhelmed in an increasingly

lifeless and dangerous

internet the Dark Forest internet and

generative AI is probably why you've

retreated to the places you still get

provably human interactions with people

you might even know private spaces like

text messages emails discords and slacks

there have always been advantages to

these more curated spaces of course but

as cultural Anthropologist Maggie

Appleton points out generative Ai and

large language models like chat GPT are

going to force us further into our

digital bunkers and impenetrable silos

echoey though they may be because the

Dark Forest internet is exponentially

expanding misinformation expert nah

schik estimates that the majority of

online content will be synthetic within

the next year case in point chat gbt and

its users are currently generating more

text than has ever appeared in every

physical book ever written every two

weeks so by now um language models have

turned uh into lots of very easy to use

products right you don't really need any

technical skills to use them so these

are a bunch of like very popular

copywriting apps that are out there in

the world here's just one of the

examples of what's coming from

Appleton's talk the expanding Dark

Forest and generative AI the reason I'm

making this video imagine that some

political lobbyists spin up 1,000

decently intelligent AIS that can

generate text and video and then they

tell each of them to go be influencers

the Bots then instantaneously make

accounts on most social media platforms

they generate their own websites they

publish independent books with

synthesized voices they make many

documentaries on YouTube they host each

other on podcasts individually they all

appear to make a reasonable human amount

of content but taken together these

human lobbyists have created an

automated content ring of lifeless

engagement at a scope and scale that

would take any one human a lifetime to

create and

curate this is all possible with the

technology that we have right now and it

appeared almost out of nowhere in just a

few months um but the point is that this

is incredibly easy to do at this point

with with no technical skills Appleton's

example isn't just hypothetical

companies are doing this right now in

the November of 2023 Elon musk's mistake

user Jake Ward tweeted this we pulled

off an SEO Heist that stole 3.6 million

total traffic from a competitor AI

allowed Ward to quickly and easily

export a competitor's sitemap turn their

list of URLs into article titles and

then in a matter of just hours create

1,800 articles from those titles that

directed web traffic away from that

competitor now imagine this kind of

strategy for every single corner of the

internet every business every chatbot

every influencer pulled off faster and

more efficiently than any human could

even

conceptualize this is what's

coming the Dark Forest

expands just over a year ago we weren't

sure any technology or indeed chatbot

could pass the famous Turing test but

now faster than we were ready for and

without our consent humanity is faced

with maybe a more interesting question

how how will we humans out here in the

dark looking for light pass the reverse

Turing

test Alan turing's seminal 1950 paper

Computing machinery and intelligence

opens with a simple question can

machines think turns out thinking is

pretty hard to Define so instead he

proposed a simple test an imitation game

he called it imagine that you're an

interrogator of two entities your job is

to determine through written text alone

which is the human and which is the

computer if a computer can fool you into

thinking it is the human then that

machine will have passed what came to be

known as the Turing test since touring

his test was the popularly understood

Benchmark for thinking machines and for

those of us outside the wired walls of

computer science the Turing test never

really seemed passable it was like The

Uncanny Valley problem for graphics

technology computers are always getting

closer and closer to rendering

believable humans but you can always

tell something is off turns out primate

brains naturally selected over millions

of years are pretty good at recognizing

faces but everything changed with chat

GPT suddenly hundreds of millions of

people around the world realized that

the Turing test had been passed

the large language model you can learn

exactly how it works in another video

now rates higher than human doctors on

bedside Manor scores better than 90% of

lawyers on the bar and 99% of graduate

students on the gr unsurprisingly every

sector of human endeavor that trades in

on human intelligence which is all of

them is now rushing to incorporate this

technology in some way I've said that

because of Technology like chat GPT

everything is about to change

but I don't just mean humans losing

their jobs or falling in love with

chatbots I mean that turing's

fundamental assumption in his imitation

game that a human will have to decide

what passes for human is no longer valid

in all cases a reverse Turing test is a

modification of the original where the

objective of one or more of the roles

has been reversed as AI proliferates

into every corner of the Dark Forest

internet other non-human machines will

increase inas inly be tasked with a

reverse imitation game where language

models and other technologies will try

to prove that they're human capture is

an example of a reverse Turing test it

may soon not be the case but right now

there are still no systems sophisticated

enough to reliably read and reproduce

distorted images of text and so the

capture computer decides that any

successful deciphering must be done by a

human a test like this is important

because without it you can imagine

websites being overrun by spam scams and

Bots but it seems inevitable now that

they will be the specific phrase that

llms use to identify themselves as an AI

language model has started to show up in

Amazon reviews Yelp reviews tweets and

Linkedin posts everywhere specifically

because we don't have the capture

equivalent protecting those spaces

to these systems the AI may as well be

human the time is coming and coming very

soon when we will need captal likee

systems in place socially politically

commercially to determine what is

actually human generated content lest

everything you see online just add to

the cacophony of a dead internet we

don't have these systems yet and if and

when we do they are likely to lag

dangerously behind what massively

incentivized generative AI will be able

to do by that point do you really think

a platform like Twitter will be able to

sort the synthetic from the Simeon and

make the Right Moves here when it's

basically just posting its way

through the

apocalypse so how do you tell other

humans and thinking machines that you

are in fact human on a dark Forest

internet

Maggie Appleton has some practical

advice for human signaling online in an

age of generative Ai and the first is

Right Here show up in meat space meet

other people go outside and actually

touch grass something no doubt was lost

when our digital lives became more

important than our physical ones so

let's go out there and reclaim it this

is not as fun or as accessible as the

internet before it went dark but it's

probably the quickest and easiest way to

prove that you are a human to other

humans the institutional verification of

your humanness feels the most dystopian

idea but it may be the most unavoidable

it would have to be something that goes

beyond a blue check mark that you pay a

rich man Bay before maybe before

registering a website writing an article

or even posting to social media you have

to show up in person somewhere and

verify that you are in fact you and not

just an image or voice that could be

easily

generated internet culture grew up

around hating this idea in principle but

it may be impossible to avoid when any

bad actor for free can fool millions of

people into thinking that the pope has

drip or deep fake of President declaring

nuclear

war the last two tactics Appleton offers

takes advantage of the fact that current

AI models are indeed machines machines

trained on certain data sets from

certain places and at certain times

knowing this we can come together online

and triangulate objective reality with

each other to prove our Humanity our

brains are constantly producing models

of the world and check those models

against sensory input current AI models

cannot do this they don't know any facts

people or world events that came into

being after their most recent round of

training and they don't feel the world

like you do they don't belong to

communities they don't reflect on

themselves and their history or enjoy

the richness of sensory experience by

communicating with each other online in

a way that reflects the aspects of

humanity apart from intelligence we can

simultaneously have more grounded

interactions and be sure that a human is

on the other side of the screen finally

we can distinguish ourselves in the dark

Forest by becoming what Appleton calls

algorithmically

incoherent large language models like

chat GPT only work by sequencing words

that are the most statistically likely

to go together in response to a prompt

based on vast amounts of training data

because of this these models hedge and

generalize in other words they basic

this gives real humans an opportunity to

reclaim one aspect of Internet culture

that is still the most fun creating

internet specific culture no language

model will be able to keep up with the

pace of weird internet lingo and memes

Appleton writes using jargon euphemistic

Emoji unusual phrases ingroup dialects

and memes of the moment will help signal

your Humanity end quote it's possible

that human culture could continue to

outpace AI culture like teenager culture

outpaces their parents Maybe Riz and Gat

will end up saving something uniquely

human online I wish I was able to end

this warning on a happier note I really

do and make no mistake I am 100% certain

that Ai and its proliferation will lead

to some amazing outcomes I think it

could provide free worldclass education

to every child in the world and it would

just be on their phone I think if

anything is going to find a cure for

cancer it's going to be one of these

large intelligent systems but for me it

is much easier to think of many more

much more harmful outcomes phone

scammers using your synthesized voice to

scam your parents

from a single

photo we are not ready for what's coming

and we need to think about our next

steps very carefully lest we all get

lost in the

[Music]

Darkness

[Music]

for

[Music]