Thanks for watching!

Hey,

welcome to another episode of Cloud

Unplugged.

We have a few stories.

I know you're going to cover these,

Salman, some of these.

We've got the EU sovereign cloud contract

for one hundred and eighty million to talk

through.

We've got the big copyright settlement

against Anthropic called the BART versus

Anthropic.

I know we have some random stories,

two random stories.

I think you've got one, I've got one.

We'll save those for the end as a

little bit of a teaser and a bit

random.

So yeah,

shall we crack on or anything you want

to add?

Yeah,

I'm just going to say a bunch of

companies keep getting hacked.

So if we get to it,

if we have time,

this is just business as usual,

just companies getting hacked.

So maybe we can talk about those as

well.

But yeah,

let's just jump into the first one.

Exactly.

Just another day, another hack.

That's it.

That's how it rolls nowadays in twenty

twenty six.

So, yes,

I guess the sovereign cloud award,

the one hundred eighty million not gone to

a US based company.

Do you want to talk that through and

then maybe why you think that's happened?

Yeah,

so the EU has awarded a hundred and

eighty million euros to a bunch of

European companies like Thales,

S three and S,

Scaleway and a few other companies for a

six year contract to build

european sovereign cloud and microsoft aws

and azure google currently holds seventy

percent of the cloud market in in eu

as well so they didn't get any piece

of that and that's perhaps to the point

that you know the us cloud act that

came out that allows the american

authorities to access data uh but that's

held by us companies no matter where it

is

So, you know, this is EU has said,

why don't we start building our own cloud?

So, you know, legal compliance,

data protection,

and all of that can be transparent rather

than relying on an American company.

So that's come out in the last couple

of weeks, which is big news.

I don't know what your take on that

is.

Yeah,

I think because there was talk around them

setting up a company,

like a European-based company,

and then that didn't really make any

difference because the holding company was

still going to be,

could be like Alphabet or Amazon.

So actually,

it still came under the same jurisdiction

and really didn't really make much

difference.

I believe.

So then, yeah,

I guess this was then they've decided not

to go US at all.

I don't know if you've heard of the

FedRAMP stuff where there was like a big

investigation into how Microsoft do their

cloud.

So it's like federal risk and something

management program.

I can't really remember.

But basically the US government basically

gatekeeper for cloud security and they did

loads of investigation and they were like

trying to figure out how Microsoft did

things and I think they spent something

like five years or some insane amount of

time trying to work things out on like

how secure

their cloud was,

but actually in the end they just gave

up because they couldn't really get to the

bottom of it and then just announced that

it's kind of fine anyway and you can

kind of use it.

And then I think they figured out that

they had like,

I think something like the people in China

maybe were having access to things or

something or other.

There was those things that kind of came

out of the woodwork.

following it on um but yeah so maybe

that's a part of it too like essentially

can't really trust even if you did do

a secure cloud i guess there's a question

on how secure secure really is for

government data and whether truly um

they're being transparent and honest

anyway but yeah yeah

think the interesting thing also yeah i

agree with about transparency but the

interesting thing here which i found is

mistral ai i don't know if you come

across them as a french ai company yeah

so they are in in their consortium that

won this hundred and sixty million euros

to do this so i think europe is

signaling that

We're trying to build our own sovereign AI

stack as well because now, yes,

cloud is an important thing,

but having your own GPUs so you can

train your own models and be in charge

of your own destiny is probably also very

important because we spoke about it

previously how AWS and how Google and

Microsoft are trying to break the NVIDIA

monopoly by having their own chips.

So I guess it's good for consumers as

well that these companies are hopefully

will be building the...

cloud that lives in european sovereign

land i know a lot of people have

to say this and and i think going

to these conferences and speaking to

people within europe the um the feeling is

that we should have our own well as

europe should have our own uh cloud in

which we can host our

I'd be transparent about how things are

being hosted.

So I think it's... I mean,

the thing is, right,

the EU is putting up €EU €EU to

build our own sovereign club and reduce

reliance on US providers.

Now, people might say,

is one hundred and eighty million euros

enough to rival the cloud giants.

And well,

let's just say it's like trying to out

pizza the hut.

Right.

But you have to start somewhere,

maybe with a sovereign slice.

So that's basically what's happening.

So we Europe is taking a stance and

starting up.

And of course,

there's a lot of other smaller companies

and large sized companies as well that are

out there.

But I don't know if we ever see

like a hyperscaler type of provider come

out of Europe.

I don't know what you think.

I think it's going to be a bit

tricky to see because they hold a lot

of real estate where all this stuff is

hosted.

I don't know if we'll see a European

come out.

I don't really understand how it's going

to work, though, because to build a cloud,

what does that mean?

Do you know what they mean by build

a cloud?

Are all these independent companies...

going to somehow come together and unify

in like a common way of

like managing and upgrading and deploying

things to the cloud or they've just

basically diversified the pot and you're

going to have five different types of

solutions to maybe work with that are all

different.

What's the end game exactly?

This is the Airbus equivalent of cloud,

John.

You know how Airbus,

we have all the European companies come

together and make a plane that can rival

Boeing.

from europe that's the same so they will

work together of course they will have

some uh some competing technologies but

that's the same as what was done with

aerospace monopoly the that boeing used to

hold so it's a it's an interesting time

And it is an exciting time that there's

more people entering the market.

So I'm looking forward to what comes out

of it.

We'll keep an eye on it.

Of course, John,

we can let people know how it progresses,

perhaps in the next few episodes,

maybe in the fifteenth or so.

We'll let you know where we are with

this.

I mean, yeah,

we should probably move on a little bit,

but I just don't know.

Why don't they just invest in?

It's a bit controversial for me to say,

but why not just invest in one company?

to create an Amazon or a GCP equivalent,

but high investment.

And then poach loads of the people from

Amazon, Google, and Azure,

who obviously have driven data centers and

know it and make the package really

compelling,

sponsor them to come over here.

Because in the end, it's talent-based.

not necessarily company based.

So it's like, you know,

you probably want to get the talent in

that's got the understanding of cloud and

the cloud services and certain things,

not just necessarily the companies that

hold the talent,

if you see what I mean.

Yeah,

but then some of these companies have

expertise in some of these areas already.

Why wouldn't you bring them together and

use them so you can create one solution,

I guess?

But giving it to one provider,

one company,

probably be a bigger gamble than dividing

it between six or seven.

And I understand what you're saying.

I think it's more of a gamble.

Really, I think the reverse.

Only because...

like they're all going to be trying to

have a slice of the pie.

They're probably looking at like,

you know,

a hundred and eighty million isn't that

much.

I mean,

like one department inside of government

spends more than that in one of the

cloud providers a year anyway.

So you're like you've basically put like a

single department in the UK spends more

than that in a year with Amazon or

whatever.

Right.

So you're like, well,

you've basically divided up of quite a

small amount of financial value.

And they're all going to be picking at

that pie and trying to work out how

to proportionate it,

all trying to work out what the

contractual model is around it.

And how do we all work together?

What's the commercial legalities around it

all?

Who owns what?

Who's prime in this contract is like a

prime that's managing it all?

Or we all are equal?

And how do decisions get made?

So in the end, it's like,

how does it legally commercially work for

me is way more complicated than saying

maybe as Europe,

we all take a little bit of the

EU's funding, put it into some big bucket.

So we all contribute.

We set up like an independent company that

we've all got a stake in proportionally.

And then we just steal those of talent.

I mean, literally,

my idea is probably crap as well and

crazy,

but it just feels like this is going

to be harder for people to agree

commercially, I mean,

not necessarily on the technicality of

what needs to be done,

but the commercial element around it feels

harder to agree upon.

Yeah,

but this is the EU way of doing

things, right?

Like we have these companies that are

split out.

Perhaps maybe, John,

you want your company to win this contract

and do the whole thing.

But I think this is just a statement.

We are in this.

So yeah,

that's why I'm saying this on this podcast

today,

because I'm a little bit bitter about it,

actually.

I was joking.

I'm just having a prejudice.

Okay,

so you kept it all in until we

got into the podcast and just started.

Yeah, I'm messing up.

It's just a statement of intent, John.

As you correctly pointed out,

€EU €EU is not a lot of money,

but it's a statement of intent.

So I think it's a good start.

It's a good start.

It is.

So another story, which is quite a big,

I think this is quite a big one.

I think this is going to be happening

a lot, I imagine.

And there's probably going to be other

lawsuits.

But the BART versus Anthropic.

So one point five billion copyright

settlement has been kind of finalized.

It's like one is like three hundred

million has to be paid.

And another four hundred and fifty million

gets paid in September twenty,

twenty seven.

This is to do with like Libgen and

Polybi,

which is like these pirated book

repositories that were online.

Apparently the model was trained on these

basically other authors books through like

a pirated route, I guess,

to basically get the data.

And then it's basically a new class of

action was then put in place against the

Meta as well.

And Mark Zuckerberg personally,

I believed,

I think he was named personally because he

gave authorization for LibGen scraping,

basically, I believe.

And so there's like a personal one,

the one against Meta.

But anyway,

essentially what they're saying is you've

stolen

from us, really, through nefarious means,

and trained your model on it,

and now you're competing with us from our

work, really.

What do you think about that?

So, of course,

we should say before we start,

we don't have any legal expertise, right?

Because this is a legal matter.

So there's two legal terms here.

Well,

but the lawyers have got an opinion on

it because that's what's happening.

So there's two terms which are being used,

fair use and transformative use.

Fair use is if you legally purchase a

book and you use that book for training

purposes, sharing in a course,

writing a review about it, that's allowed.

The other one is a transformative use,

which is you take your book and you

create something else out of it which does

not stop the sales of the original book.

Inspiration.

Take inspiration from it.

Yeah, exactly.

Exactly.

It's like sampling a track.

I took a track.

You can sample a track and you can

create a new one.

So what the judges said is the fair

use, the books that you bought...

The books that you bought and you train

the model on, that's kind of fair use.

Some of the people are not happy with

it.

The publisher is not happy with them

because what you're not doing with that is

stopping the sales of the original book.

You took the information,

you trained yourself.

Basically,

you turned it into the AI learning process

was considered as transformative use.

So it analyzed the pattern,

it created something new.

You can't go to ChatGPT or Claude and

say, can I read the War and Peace?

It's not going to give you the text

to read the War and Peace.

It'll give you a summary,

which is allowed.

So

Which is fine.

I don't know if some of the publishers

are not happy with that.

But the thing about these pirated books,

that's the thing that's caused the issue

for them to say, right,

now you're allowing pirated books to be

downloaded and trained.

That's not allowed.

But it still doesn't answer the question,

is training your model using books,

is that fair use or not?

The court says it is.

People don't really agree that this is

fair news.

This is something new.

It's kind of like two things are being

mixed up.

You might remember Napster, right?

I know it's before your time, John,

but Napster back in the day where you

can turn songs illegally and listen.

And then Spotify came in and then they

said, okay, you know what?

You can stream the music,

but you have to pay like ten pounds

a month.

And the singers will only get like zero

point zero five cents of every stream,

every million things that they stream.

it's yes it's been things have been agreed

upon that payouts need to be given up

but i don't know this opinion i don't

know if this is gonna go away anytime

soon so basically wasn't though the main

issue that they basically used pirated

yeah as in like it's which is technically

like you just said with like napster or

like or any torrent bit torrent you know

I mean,

BitTorrent itself obviously is not the

issue,

but basically it allows people to download

the software,

obviously using other software, sorry,

download the assets using other software,

obviously published by others on servers,

BitTorrent servers,

which then means most of those would be

like movies or books or whatever else that

people decide to share to save people

money.

But they've done it as a company.

Apparently they then trained on pirated

information which they hadn't bought.

And so that is obviously surely is illegal

because that is illegal.

But the case when it started wasn't just

about the pirated companies, right?

It's about the fair use as well.

But that question is still open because

according to that case,

transforming the information to something

new that's allowed.

So the publishers somehow still need to

get their information back.

They still need to get their investment

back.

So yes, the pirated ones,

that's absolutely bad.

But I don't know if we got to

this is the final time,

the last time you're going to hear about

this.

What we've seen is that AI companies may

be allowed to train on copyrighted books

under fair use,

but only when you obtain the book legally.

However, if you buy it...

Didn't New York Times have the same thing?

Didn't New York Times win a court case?

I'm sure they had one as well.

They probably did.

So if you get the book legally,

you are allowed to train your models.

But buying the book and teaching your AI

from it might be okay and downloading it.

But if you download it from

totallynotpiratedbooks.com,

that's not allowed.

So just buy a book, one book,

and train your model.

Yeah,

that does seem a bit unfair because

training a model,

like me training a model on a book,

not so much risk.

Just as a person, I mean,

say I want to do model building.

It's only me that uses that model.

which is the same as me referring to

the book manually myself anyway,

there's not very much risk.

Me though, publishing the book,

say I then took that book,

then published that book on the internet

for others to then read at high scale

um because i've like say this podcast is

like going through the roof and we've got

bazillions of followers right me then

publishing the book or narrating the book

on this podcast with our bazillion viewers

surely can i do that can i can

i read each chapter out loud to bazillions

of people on a podcast and that's not

copyright

That's not fair use, right?

So this is you falling into fair use.

But the model that's trained on the book,

which is a service for lots of people

to then use,

that's very different to a single model

that an individual is using.

Yeah, because now, according to the court,

they decided that's transformative way,

right?

So you've transformed the information.

So that's fine.

The only thing they got caught upon is

downloading the pirated books.

So basically,

buying the book legally and training your

AI on it is fine.

But if you download it from...

piratebay.com that's not fine and train

your model so you can still you can

train the model yeah that's that's what

they came up with but i don't think

this is the last we're going to hear

of this we will still it will still

keep yeah there'll be loads of this and

we also got a little confession from you

for using napster back in the day to

download uh software you see you've seen

very important i've seen uh i've seen i've

seen the the the social what is that

movie but the facebook movie

oh the social dilemma is it or something

the social network the social that's the

one it is actually a dilemma having us

uh having having some of these social

media is a dilemma but anyway yeah that's

that's what that is yeah so uh i

guess the random story getting into the a

little bit of a random story what what

is your uh

Random story of the week.

My random story,

which is not too random anymore,

is that Stanford and Imperial College

carried out an Internet Archive study,

and they found that by the middle of

last year,

about thirty five percent of the newly

published websites were generated or

assisted and almost twenty percent were

fully generated.

So basically,

the research is linked to what they call

a dead Internet theory,

that the idea that bots and AI are

dominating the online content.

So imagine this.

Imagine this.

You go into a library.

And you start looking at the books.

Every third book is written by somebody

who doesn't sleep, doesn't eat,

doesn't pay taxes,

doesn't get writer blocks.

How original it is, I don't know.

But this is what's happening right now.

Now,

do you care if the information that you

get is useful or

transform your life i'm not sure does this

provide value that's fine but a lot of

the information online is now going to be

ai generated i've seen a lot of posts

and comments on the posts that are

generated by ai and people actually

accepting that they

put a AI generated comment,

like their bots reply to LinkedIn posts.

So this is a reality is not so

random.

But this is a reality.

That's what's mental.

It didn't.

There's loads of like,

there's some people that were MPs here

that I think also wrote a book,

or maybe might have been the US I

can't really remember.

they basically wrote a book with um with

ai it was like it was really obvious

everyone was then commenting on it saying

this is like the most ai written book

ever published um because it's super super

obvious that it is written by ai so

yeah i guess but it's just going to

be not very the worth or value of

something that can produce something

really easily it's like

It's not high value, is it really?

I don't know how popular those things are

going to be.

I wouldn't want to read an AI generated

book.

Well,

if you don't know it's an AI generated

book and you read it and you liked

it, what would you do?

You probably never see me again.

You'd be like, where did John go?

I'd just be in the house all the

time,

paranoid that people can tell when they

see me that that guy likes AI books.

Somehow it would translate maybe in my

eyes.

When people look at me,

they'll be able to see.

So I'd probably never be able to leave

the house, I think.

I'd be locked up inside.

Would you accept textbooks like,

for example, a match textbook?

There was a there was a thing going

around where somebody published a match

textbook for a school and there was an

equation that they're helping us solve.

And there's a table underneath.

It says, oh,

if you want me to explain a different

way, let me know.

Right.

So they just copied and pasted whatever

the model said.

So would you accept a textbook perhaps for

like school?

Maybe.

Yeah,

I think I think anything that's I think

creativity.

is something i mean obviously figuring

things out and that's that's not something

necessarily we specifically have created

it's that it's it's the calculation of the

things around us it's the things that

already existed nature or you know the

parameters different but something that

somebody else has created or manifested

you know through creativity is a very

personal thing that is affiliated with

that individual their perception of the

world that

as another human being should be very

relatable.

And so the characters or the person,

wherever that stemmed from,

including music or art or whatever else,

I do feel there is a protection because

it's got a sense of being mocked when

AI does it.

There's a mockery to it that's like it

doesn't really understand any of those

things that are quite humanistic.

So therefore the value of it's like a

little bit insulting when something

pretends to understand those things and

then just can knock it up rather than

factual things.

But yeah, I've got a really quick one.

I'll get it out of the way.

But there's basically a wolf in Korea that

escaped from a zoo and was on the

rampage.

And some guy basically created an AI

generated image of this wolf in his area

and then posted it,

I think on social media.

And then everybody freaked out and all the

schools closed in the area.

And I think hospitals were obviously

panicking because of the wolf and all

these other things.

So I think there was a lot of

police and people reacting to it.

And they basically put him in jail.

He's gone to jail in South Korea for

basically causing a disruption by using AI

to pretend that the wolf was in an

area that it wasn't really in.

The rule of law around AI is coming

for people nowadays.

It's hit or miss.

You never know what's going to happen.

They're probably looking for the wolf on

the wall street.

All they found was a wolf made by

AI on the streets.

Some guy.

I mean,

I suppose if I dressed in a wolf

outfit and was prowling around the

streets...

might not get arrested.

You know, how convincing you are,

like pretending to be a wolf.

John,

no one's ever seen a six foot three

wolf, right?

So if you dressed up as a wolf.

Not that they haven't, but they soon will.

They soon will, my friend.

John in a six foot,

John in a wolf costume.

I'd love to see that.

yeah i'll have to uh but yeah ai

ai uh generated um wolves are a no-go

in south korea so don't even think about

it um fair enough but anyway so i

know it's a bit of a a quicker

one for us uh this week um but

but um back at back in action since

my holiday but

Yeah,

we'll be tuning in next week as well.

And we shall speak to everybody soon.

Thanks for listening.

Speak to you soon.

Bye-bye.

Cheers, bye.