0:00

Some people ask to

0:02

be cryogenically preserved because

0:04

they have a terminal

0:07

illness, but for me, I

0:09

just want to survive the

0:11

energy crisis. Yeah, you're a

0:13

bit cold. I'm a bit

0:15

cold. Yeah, I'm a bit

0:17

cold. Yeah, I think it's

0:19

a normal rational reaction. Yes.

0:21

Rather than putting on another

0:24

jumper jumper. Hello Hello and

0:26

welcome to episode 35 of

0:29

The AI Fix your weekly

0:31

dive headfirst into the bizarre

0:34

and sometimes mind-boggling world

0:36

of artificial intelligence. My

0:38

name is Graham clearly.

0:41

And I'm Mark Stockley.

0:43

Mark, let's race straight into

0:46

the news. President Trump announces

0:48

the 500 billion dollar Stargate

0:50

project. Shipment of friend AI

0:53

digital companion delayed. O3 mini

0:55

is here. O3 is on

0:57

the way. Were Donald Trump's

0:59

as active orders written by

1:01

AI? AI agents are here.

1:04

LinkedIn accused of using private

1:06

messages to train AI. AI

1:08

super intelligence could be here

1:10

in two to three years

1:13

time. Right, where shall we

1:15

begin? So I think we should

1:17

start with the most consequential story.

1:19

of the week. We've had AI

1:22

agents come out this week. We've

1:24

got Trump announcing the Stargate project.

1:26

But you said that Friend, the

1:28

AI digital companion, has been delayed.

1:30

That's right. It's the AI-powered necklace,

1:33

which we spoke about in a

1:35

previous episode, which marketed itself as

1:37

not imaginary. Turns out it is

1:40

perhaps imaginary, at least for now. It's

1:42

failed to show up on time. Now, the

1:44

people at Friend. You remember, this is the

1:46

thing you wear around your neck or you

1:48

attached to your collar. and it listens to

1:50

your conversations and it chirps and on your

1:52

phone will pop up some advice as to

1:54

how to ask a girl out or I

1:56

hope you're enjoying your sandwich. It's the AI

1:59

friend for people. who don't have any

2:01

real friends. That's right. It's not only

2:03

about not having any real friends, you

2:05

also don't have a friend because it

2:07

doesn't exist. They have announced it in

2:09

an email to customers who signed up

2:12

in advance that their AI companion will

2:14

now only be available from quarter three

2:16

of 2025 rather than quarter one as

2:18

they originally hoped. In the email which

2:21

the founder sent us sent to people...

2:23

he not only apologised about the manufacturing

2:25

and the beta issues which he's experiencing

2:27

with his startup but also on the

2:29

friend.com website if you go there right

2:32

now as I speak there is a

2:34

bit of a chatbot you can practice

2:36

having little AI chats with an imaginary

2:38

friend up there I've just gone there

2:40

yes there's a picture of a guy

2:43

with a little green dot which indicates

2:45

he's online yeah and it says his

2:47

name is Donald yeah oh oh oh

2:49

and it says hold on I'm in

2:52

the middle of a big interview right

2:54

now a lull Oh, so we can't

2:56

chat to you at the moment. Well,

2:58

I don't know. Maybe speaking to Fox

3:00

News. Are you trying to get a

3:03

job because the friend business is not

3:05

happening? That stings a bit. Writing is

3:07

my main gig, Friends, just a side

3:09

thing. Do your customers know this? Do

3:11

you think this is a chat board?

3:14

You're direct to the CEO. Now Mark,

3:16

O3 Mini. So open AI CEO Sam

3:18

Altman went to Twitter a few days

3:20

ago to announce the company's latest model.

3:23

Oh three mini will be available in

3:25

about two weeks time. So the mini

3:27

models are smaller versions of open AI's

3:29

flagship models and they're basically faster and

3:31

cheaper. Do we know why they're called

3:34

O? Where they give them the O

3:36

moniker? O is Omni I think. Oh

3:38

I see. Yeah. So the O series

3:40

are the models that can reason. So

3:42

they take much longer to answer queries

3:45

than their predecessors because they actually, they

3:47

think much harder about the answer by

3:49

doing what they call chain of thought.

3:51

So they break down the problem and

3:54

you can see the model doing this,

3:56

it throws up little labels saying, oh

3:58

I'm thinking... about this now I'm doing

4:00

this now I'm thinking about this so

4:02

it breaks down the problem into smaller

4:05

logical steps and the O3 series

4:07

is designed to improve the reasoning

4:09

capabilities a bit more but it's

4:11

also supposed to be much quicker

4:13

to try and bring down that

4:15

thinking time yes so O3 minis

4:17

coming out first and then that

4:19

will be quickly followed by O3

4:21

and O3 pro Sam Altman isn't

4:23

giving away much about the capabilities

4:25

of O3 but separately Open AI's

4:28

chief product officer said that when

4:30

it comes to the full O3

4:32

model is going to be the

4:34

175th best computer programmer in the

4:36

world. Is there a chart? Well,

4:38

it turns out there is. We

4:40

watched the sort of full first

4:42

version of this new model, O1.

4:44

It was like the thousandth best

4:46

engineer in the world and software

4:48

engineer at these competitive coding problems.

4:51

O3 will be the 175th. Best

4:53

engineer in the world. So by

4:55

any measure it is phenomenally good

4:57

at being a computer programmer. Okay

4:59

you can paint it that way

5:01

Mark. The alternative is that there's

5:03

lots of computer programmers now who

5:05

aren't doing their own program but

5:07

using AI instead and so their

5:09

programming quality has actually deteriorated and

5:11

so they've fallen down the chart.

5:13

Well there are a hundred and

5:16

seventy four of those. So O3

5:18

is out imminently, but he also

5:20

said O4 Mini will be around

5:22

in July. Oh wow, that is

5:24

soon. Now, Donald Trump is the

5:26

US president. Great news for anyone

5:28

who enjoys questioning 14 times a

5:30

day if they're living in a

5:32

simulated reality or not. Anyway, he's

5:34

been busy, far and off executive

5:36

orders left, right, Sante, he's been

5:38

very, very busy. Some of them

5:41

caught the attention of legal experts

5:43

who questioned whether they were written

5:45

by AI instead of humans. Turns

5:47

out many of Trump's executive orders

5:49

are difficult to read and understand.

5:51

They're full of errors and storted

5:53

language. What you're saying is some

5:55

of them aren't. And we suspect

5:57

they may have been written by

5:59

AI. Is that? So they're... is

6:01

an executive order which is for

6:03

restoring names that honor American greatness.

6:06

This is being used to rebrand

6:08

the Gulf of Mexico as the

6:10

Gulf of America. There's an executive

6:12

order all about we're going to

6:14

start calling it something different. And

6:16

according to some people they suspect

6:18

that there's sections of this executive

6:20

order which were not written by

6:22

a human. So lawyer Rafi Melconian

6:24

for instance he argues that it

6:26

was not just written for morons,

6:28

but it was definitely written by

6:31

AI. So here is a section,

6:33

and you can decide whether you

6:35

think it is written by A

6:37

or not. The Gulf is also

6:39

home to vibrant American fisheries, teeming

6:41

with snapper, shrimp, grouper, stone crab,

6:43

and other species, and it is

6:45

recognized as one of the most

6:47

productive fisheries in the world. with

6:49

the second largest volume of commercial

6:51

fishing landings by region in the

6:53

nation, contributing millions of dollars to

6:56

local American economies. They might have

6:58

asked a 12 year old boy.

7:00

So Graham, I predicted in our

7:02

new year episode that 2025 would

7:04

be the year of AI agents.

7:06

Yes. AIs that can actually do

7:08

stuff on your computer or on

7:10

the internet. Like book holidays, pretend

7:12

to do your job, that sort

7:14

of thing. Anyway, open AI has

7:16

just released... It's technology preview of

7:18

operator which is its AI agent.

7:21

So this is hot off the

7:23

press. We can't get it yet.

7:25

It's only available to pro customers

7:27

in the US. But there is

7:29

a video. Okay. Okay, at the

7:31

moment I'm seeing a talking pop

7:33

plant. What is operator? Operator is

7:35

a research preview of an agent.

7:37

There's a man sat at a

7:39

laptop with a boring screen. Okay,

7:41

he's asking it to go and

7:43

buy some ingredients for his pasta

7:46

recipe. I'm going to skip forward

7:48

a bit because this guy's quite

7:50

dull. He's very dull. Okay, he's

7:52

found a recipe, he's clicking on.

7:54

on it. This video is going

7:56

to set the world a light.

7:58

I never realized the Gentic AI

8:00

was quite as exciting as this.

8:02

It's like they rushed it out

8:04

about two minutes before they decided

8:06

to go live. So basically operator

8:08

is an agent. It's got its

8:11

own browser and it can navigate

8:13

the World Wide Web and it

8:15

can do things for you. So

8:17

we gave the example of booking

8:19

holidays when we were talking about

8:21

agents in general. And they give

8:23

that example too. And of course

8:25

the video is operator going out

8:27

there and looking at a recipe,

8:29

finding a recipe, reading the ingredients,

8:31

and then buying all of the

8:34

ingredients. And separately, Dario and Modi,

8:36

as the CEO of Anthropic, says

8:38

that he expects virtual collaborators, which

8:40

are agents that can do anything

8:42

on your computer that you can

8:44

do. He reckons the first half

8:46

of this year. Wow. There's

8:49

a model that is able to do

8:51

anything on a computer screen that a

8:53

kind of virtual human could do. And

8:55

you talk to it, you give it

8:58

a task, and just like a human,

9:00

the model goes off and does a

9:02

bunch of those things, and then checks

9:04

in with you every once in a

9:07

while. I do suspect, I'm not promising,

9:09

I do suspect, that a very strong

9:11

version of these capabilities will come this

9:13

year. And it may be in the

9:16

first half of this year. So

9:19

we are now in the age

9:21

of the autonomous AI agent. So

9:23

LinkedIn, you know, we all love

9:25

LinkedIn, it's great, isn't it? Oh

9:27

yeah, of course, of course we

9:29

do, of course we do, of

9:31

course we do, it's fantastic. I

9:33

don't think we do, I think

9:35

LinkedIn has always been awful, but

9:37

all the other social media became

9:40

so much more awful, we've all

9:42

just on a level down. The

9:44

thing about LinkedIn is everyone up

9:46

there pretends to be really professional

9:48

and ethical and... Unsurprisingly, LinkedIn has

9:50

now been accused of not being

9:52

professional and ethical. So there is

9:54

a lawsuit in the United States

9:56

filed on behalf of LinkedIn premium

9:58

users. Those are the people who

10:01

are duped into giving LinkedIn money

10:03

every month. So you can see

10:05

who viewed your profile. That's it.

10:07

That's basically the reason. Yeah. And

10:09

it accuses LinkedIn of sharing the

10:11

private messages of LinkedIn premium users

10:13

with other companies to train AI

10:15

models, naughty, naughty, naughty if that's

10:17

true. And the accusation is that

10:19

they quietly changed their privacy settings

10:22

and hoped nobody would notice. And

10:24

according to the lawsuit, when they

10:26

did get called out, they simply

10:28

updated their epicues with a sort

10:30

of, oh yeah, by the way,

10:32

we did use your data, yeah,

10:34

but you can opt out now

10:36

if you want to. Not that

10:38

it'll make any difference, of course,

10:40

because the AI companies have now

10:43

got your messages. This is very

10:45

Facebook circuit 2009, 2010, isn't it?

10:47

Isn't it? We've just changed all

10:49

your privacy settings overnight. Sorry. So,

10:51

Antropic CEO Daria Daria Modi. And

10:53

one of the things that he's

10:55

talked about is AI super intelligence.

10:57

So it does feel as if

10:59

AI is actually picking up pace

11:01

after a couple of, I'm not

11:04

going to say slow years, but

11:06

we haven't had a really big

11:08

breakthrough since ChatGPT, I don't think.

11:10

Or you could say the O-series

11:12

maybe. What about the Moflin? We

11:14

haven't really seen a breakthrough for

11:16

a couple of years. But as...

11:18

part of his conversation at Davos

11:20

he said it's likely that we're

11:22

only two or three years away

11:25

from artificial super intelligence. So that's

11:27

not AGI, it's not artificial general

11:29

intelligence, that's not an AI that

11:31

is as good as humans across

11:33

a range of tasks, that is

11:35

an AI that is far far

11:37

better. You know genius level intelligence

11:39

in all sorts of different areas.

11:43

Yes, I think it's likely not certain

11:45

but likely that we actually are

11:47

only two to three Only two to

11:49

three years away And he said he's

11:52

speaking up more about it now than

11:54

he used to because he wants to

11:56

warn people Because of the serious economic

11:59

and social disruption that that could cause

12:01

I you know I want to warn

12:03

people I want people to know about

12:06

this. And you know, some of that

12:08

is going to lead to people saying,

12:10

well, what's your plan? What's your plan

12:13

for economic disruption? What's your plan for

12:15

the risks of the world? So two

12:17

to three years, he's talking about economic

12:20

disruption because of AI. Now, it seems

12:22

to me that people have been warning

12:24

about things like climate change for 50

12:27

or 60 years. And we haven't made

12:29

quite as much progress as we'd like

12:31

two to three years. I'm not confident

12:34

I'm not confident I'm confident that we're

12:36

all on board with dealing with this

12:38

particular problem. Hug that muffling extra tight,

12:41

Graham. So Mark, it's pretty cold right

12:43

now, here at Cluelly Towers. Right. My

12:45

energy bill has gone through the roof.

12:48

Right. Trying to deal with this bad

12:50

weather we've been having. I haven't quite

12:52

worked out how I'm going to pay

12:55

for my electricity this month, but after

12:57

careful thought I've got a plan. Is

12:59

it starting a podcast? No, no, that

13:02

definitely isn't the way to make

13:04

money. I've decided the best course of

13:06

action for me is cryogenic suspension. Isn't

13:08

that even colder? You're going to turn

13:11

the heating off entirely. and go into

13:13

a state of hibernation. I figure that

13:15

if I do put myself into cryjunk

13:18

suspension, it's not going to cost me

13:20

a fortune because I mean it's already

13:22

pretty cold. I'm half the way there.

13:25

Yeah. And it will save a few

13:27

dollars. And maybe, while my body is

13:29

preserved in a block of ice, waiting

13:32

to be revived in centuries to come,

13:34

by the time I am awoken, boffins

13:36

will have worked out all the issues,

13:39

you know, perpetual motion, cold fusion. I

13:41

can be brought back after we've solved

13:43

all this AI singularity nonsense. Everything will

13:46

be sorted out. We'll be on a

13:48

universal income. I won't have to worry

13:50

about all these outdated things of having

13:53

to pay my electricity and gas bill

13:55

anymore. You are aware of inflation. I

13:57

suspect your life savings. I'm not going

14:00

to pay for more than a few

14:02

seconds of your electricity. You... Mark, just

14:04

wait and be... edified by what you're

14:07

about to learn. Okay. Because some people

14:09

ask to be cryogenically preserved because they

14:11

have a terminal illness or something, but

14:14

for me, I just want to survive

14:16

the energy crisis. Yeah, you're a bit

14:18

cold. I'm a bit cold. Yeah,

14:20

I'm a bit cold. Yeah, I think

14:23

it's a normal rational reaction. Yes. Rather

14:25

than putting on another jumper jumper. In

14:27

the northern hemisphere, at this time of

14:30

year, it gets a bit chilly and

14:32

so you begin to think, where can

14:34

I get cryogenically suspended until it? So

14:37

I went looking on the web. Right.

14:39

And I found this outfit called Time

14:41

Shift. Welcome to Time Shift. The world's

14:44

first cryopreservation facility designed to extend your

14:46

lifespan while keeping your body fully protected

14:48

from biological decay. And they claim to

14:51

be the world's first crypto? No. That's

14:53

something else. Are you going to ask

14:55

me to invest in something? They claim

14:58

to be the world's first Cairo preservation.

15:00

Cryo. Cairo is a large city in

15:02

Egypt. Oh for God's sake. Time shift

15:05

claims to be the world's first cryo

15:07

preservation facility. Their mission is to turn

15:09

people into cryonauts. You've heard of astronauts,

15:12

you've heard of cosmonauts, these cryonauts, the

15:14

pioneers who go forward and jump into

15:16

a bucket full of ice. preserve their

15:19

lives beyond their normal lifespan. Now, I

15:21

wasn't sure if I could trust time

15:23

shift with my body, but then I

15:26

found out that the people behind it,

15:28

there's a generative AI scientist called Dr.

15:30

Alex Zavaronkov, and also a biotechnologist and

15:33

YouTubea Hashim al-Gali. And I thought, well,

15:35

if they sound trustworthy, yeah, generative

15:37

AI, YouTubea, fantastic. I thought sounds legit.

15:39

So imagine you've got an aggressive cancer,

15:42

you're thinking, well, this completely completely sucks

15:44

completely sucks. you can book yourself into

15:46

timeship. And the first thing they do

15:49

prior to preserving you cryogenically is they

15:51

carry out any additional enhancements and repairs

15:53

that your body may need. you know,

15:56

maybe you need lip filler, lip filler,

15:58

hair transplant, nose job, penis reduction, whatever

16:00

it may be that you need. And

16:03

then, actually, maybe I should go the

16:05

other way. Well, it will get cold,

16:07

won't it? And then they implant advanced

16:10

monitoring devices inside your body to track

16:12

it, track your body's vital signs. These

16:14

are submarines that you swallow. It could

16:17

be. It could be. They're not exactly

16:19

specific about how they're going to do

16:21

this. And then they take a digital

16:24

scan of you, they make a life-like

16:26

AI avatar, and they preserve all this

16:28

information about you. And that avatar will

16:31

help you maintain a relationship with your

16:33

friends and families and keep a presence

16:35

in their lives, because they're going to

16:38

miss you, right? While you're cryogenicly suspended.

16:40

We hope so. This is actually just

16:42

a way of guaranteeing that they miss

16:45

you. Don't forget about Auntie Joan. Left

16:47

her in the cold living room. Occasionally,

16:49

Auntie Joan will be popping up. Her

16:52

digital persona will be running around

16:54

in the metaverse. Sending you DMs on

16:56

Facebook, arguing with strangers on Twitter. Now,

16:58

that's all very well, right? They've made

17:01

it an avatar. That's fair enough. But

17:03

you're probably wondering, when are they going

17:05

to actually chuck you in the freezer?

17:08

I'll tell you. What they do first

17:10

voice, they take you as a potential

17:12

cryonaut. And they place you inside a

17:15

pressurized vault. Your body adjusts to the

17:17

increase in pressure and it becomes saturated

17:19

with what they describe as a proprietary

17:22

mixture of helium, oxygen, xenon, argon, CO2

17:24

and other gases. It's like the formula

17:27

for Coca-Cola. Are you conscious while they're

17:29

pumping you full of this gas over

17:31

several days? This is what I want

17:34

to know. I've read the FAQs, I've

17:36

watched the video. Yeah. They don't explain.

17:38

Are you just lying there? But once

17:41

you've been pressurized, then you're put into

17:43

your own individual cryopod, which is basically

17:45

a perspect's coffin. Hold on, I'm going,

17:48

once you're pressurized, then you're put into

17:50

an individual cryopod. I'm just telling you

17:52

what the first of... So before that,

17:55

what are you like in a crow

17:57

waiting room with a bunch of other

17:59

meat avatars? I'm getting pumped full of

18:02

helium. We've got a YouTube working on

18:04

this mark, all right? And you're putting

18:06

inside this glass coughing thing, which isn't

18:09

just containing your body. It's also

18:11

got their little robotic arms and instruments

18:13

which are monitoring you and doing high-tech

18:15

stuff on you. And the pressure is

18:18

further increased and the temperature keeps on.

18:20

dropping. They have thought of everything. They've

18:22

got electromagnetic fields which are being applied

18:25

to help stop ice forming in your

18:27

cells. Well, actually they haven't thought of

18:29

everything, have they? Maybe you've got a

18:32

pacemaker or something like that, which... Anyway,

18:34

don't... Or fillings. Don't worry about that.

18:36

Yes, don't worry about that. These guys

18:39

know what they're doing. Eventually, temperature is...

18:41

The individual cryopod is kept in optimal

18:43

conditions to ensure long-term stability. The cryinots

18:46

are then transferred to a storage facility

18:48

where they will remain in suspended animation

18:50

for several years. And these things will

18:53

be based, maybe an old nuclear bunkers

18:55

could be underground, because they want to

18:57

protect you. It sounds brilliant. And you

19:00

will be in there until you're ready

19:02

to wake up and presumably either deal

19:04

with your electricity bill. Or to deal

19:07

with the disease which sent you there

19:09

in the first place. I mean, who

19:11

wouldn't want to live forever in a

19:14

glorified freezer with your consciousness uploaded to

19:16

the internet? Well, they're not even doing

19:18

that, are they? It's just an avatar

19:21

which knows a little bit about you.

19:23

Ah, you see, Mark? You haven't really

19:25

studied this as much as me.

19:27

You haven't really grasped the genius of

19:30

this time shift project. Okay. Because all

19:32

the time that you are on ice,

19:34

as it were. Yeah. Your family are

19:37

hopefully dialing in to have little face-time

19:39

chats with you. That is being watched

19:41

by an AI. They're doing most of

19:44

the talking. Your avatar is doing a

19:46

bit. Your avatar is responding with your

19:48

character and the AI is learning more

19:51

about you and your family. It has

19:53

a cryo memory vault it says. They

19:55

will store your interactions and compile a

19:58

family history timeline which will enrich your

20:00

revival experience with cherished memories. All the

20:02

things you missed. All the birthdays. all

20:05

those conversations. So when they're done pumping

20:07

you full of helium, they're going to

20:09

pump you full of memories you didn't

20:12

have. That's right. Wow. Or they're just

20:14

going to sit you in front of

20:16

a videotape and say watch this for

20:19

a while. And they've also got this

20:21

thing called the Timeshift Academy, so they

20:23

are going to provide educational programs for

20:26

cryonauts, because if I got woken up

20:28

in the year 2728, for instance, I

20:30

might need a little bit of preparation

20:33

for the outside world might need preparation

20:35

for me as well. So there's an

20:37

academy. which once you're revived, it's like

20:40

a school which you go to, where

20:42

they will teach you about the

20:44

outside world and get you up to

20:46

speed as to what's been going on.

20:49

Well, they might do this, presumably they

20:51

haven't woken anybody up yet. They haven't

20:53

done that yet, no. There's always the

20:56

suspicion with this Croyo stuff, isn't it?

20:58

You know, they call you Croyo Norse.

21:00

Like an astronaut is somebody who goes

21:03

out into the solar system and does

21:05

a thing, you know? walks on the

21:07

moon or something or space walks or

21:10

fixes a satellite. But a cryo-nol is

21:12

someone who does absolutely nothing other than

21:14

getting very very cold. And you're entirely

21:17

in the hands of the people who

21:19

are operating this cryogenic storage facility. I

21:21

mean, how do we know that this

21:24

is actually a cryopod and not a

21:26

sarcopod? Well, you've actually put your finger

21:28

on a genuine problem is, can you

21:31

trust the people who've put you in

21:33

this? The people who are looking after

21:35

the thermomometer. Well, you should really watch

21:38

the video. You should really watch the

21:40

video. Because in that video, you'll see

21:42

the people who are looking after you.

21:45

We're judging the book by its cover.

21:47

Is that what you're saying? Okay, here

21:49

we go. The time shift facility will

21:52

operate with a fleet of fully autonomous

21:54

robots capable of... Terrifying humanoid robots. I'm

21:56

not sure if they are actually robots.

21:59

Why would a robot ever have

22:01

to sit down? That sounded as a

22:03

great point. One of them is in

22:05

a chair. So either these robots are

22:08

actually people in wet suits. Or maybe

22:10

that's how a robot recharges. It sits

22:12

in its chair and it's... You think

22:15

a sort of charging probe that goes

22:17

up their butt? Yeah. Slips into a

22:19

receptacle. Some people might find these robots

22:22

a bit scary. I certainly thought that

22:24

when I got to the bit of

22:26

the video where it talks about how

22:29

it wasn't just about preserving entire bodies.

22:31

You could also use this for organ

22:33

transplants. Sorry. So we're sorry Mr. Cooley,

22:36

but we've changed the terms and conditions

22:38

for users on the basic plan. Now

22:40

I don't know where they are harvesting

22:43

these organs from. This is a freemium

22:45

service. And not just organs, also pets.

22:47

There's a section of the video. There's

22:50

a section of the video where they

22:52

talk about how when you get woken

22:54

up, you can be reminded of your

22:57

lovely pet dog. We know how much

22:59

you love your pet, which is why

23:01

we designed TimeShift to help you preserve

23:04

your beloved companion for potential revival alongside

23:06

you. Our advanced techniques aim to extend

23:08

your treasured bond far beyond a single

23:11

lifetime, allowing you to reunite with your

23:13

beloved pet in a future world of

23:15

scientific marvels. That's a lot less

23:17

sinister than I thought it was going

23:20

to be. That was quite a sharp

23:22

segue from organ harvesting to pets. I

23:24

thought you were going to wake up

23:27

with your puppy's heart or something. Preserve

23:29

your beloved companion for potential revival. It's

23:31

a rhino! Yes! There's a rhino, there's

23:34

the gorilla. They couldn't afford the computer

23:36

graphics to make a bigger virtual container

23:38

for the rhino, so they just put

23:41

it in the cat-sized container. So what

23:43

they are saying is this can also

23:45

be great for conservation. They're saying they

23:48

can keep endangered species in there. So

23:50

they're going to save the planet by

23:52

stuffing a panda into a cryopod and

23:55

hoping for the best. That's not how

23:57

nature works. Now they say some people

23:59

may be worried, this is going to

24:02

be hundreds of years. Is it going

24:04

to be safe from attack from the

24:06

outside world? And they say yes, we're

24:09

going to have fortified facilities, we're going

24:11

to have sustainable energy, we're going to

24:13

have... Windmills, we're going to have solar

24:16

power and we'll have a nuclear plant

24:18

as well, which will be handy I

24:20

suppose. It's good that they're going to

24:23

have windmills because come the nuclear apocalypse

24:25

blocking out the sun, at least there'd

24:27

be a good breeze going on.

24:29

So they have thought of everything. Now

24:32

you mentioned right at the beginning, what

24:34

was I going to do for my

24:36

financial security? What they're saying they're going

24:39

to do is of course they're going

24:41

to use AI. So when you book

24:43

yourself in... you give control of all

24:46

of your finances, two timesheft, who will

24:48

ensure your future financial prosperity during stasis

24:50

by making investments driven by AI. So

24:53

it's not only robots who are looking

24:55

after you and making... They're quite confident

24:57

the stock market is going to survive

25:00

the nuclear apocalypse and the singularity then.

25:02

Mark, they're going to use the blockchain.

25:04

So I don't think we've got anything

25:07

to worry about. I think it's all

25:09

pretty reassuring. Sign me up. I'm ready

25:11

for my electromagnetic resonance rewarming and the

25:14

future of financial stability managed by AI

25:16

algorithms. You mentioned earlier, we've got a

25:18

new president in the White House. Yes,

25:21

yes. TV's Donald Trump? Yes, indeed. Whether

25:23

you love him or hate him? Or

25:26

you don't care. Right. I don't care

25:28

either. President Marmite is back for another

25:30

four years. Oh no, I love Marmite.

25:33

And this time around, I have to

25:35

say it feels a bit... different. Right.

25:37

So he's de facto vice president Elon

25:40

Musk. Don't know if you've heard of

25:42

him. Yes. One of the most consequential

25:44

figures in technology, an important figure

25:46

in AI, and certainly one of the

25:49

most likely to turn into a full-blown

25:51

bond villain as we established with a

25:53

rigorous AI-driven test just a few episodes

25:56

ago. He does a great doctor strange

25:58

love impression as well I saw the

26:00

other day. Oh yes, yes he does,

26:03

yes he does. But Musk I think

26:05

is just the tip of the tip

26:07

of the iceberg. The lead up to

26:10

Trump's second term has been accompanied by

26:12

something that feels like a vibe shift

26:14

in the whole technology sector. Though in

26:17

his first term it felt to me

26:19

like the tech industry was kind of

26:21

at odds with Trump. You know most

26:24

of it's in San Francisco there's a

26:26

very particular worldview that goes with that

26:28

place and with the companies that are

26:31

established in that place. And I had

26:33

the sense that maybe they thought they

26:35

were bigger than he was. And this

26:38

time around... it seems like they're either

26:40

actively supporting him or at the very

26:42

least they've decided to fall in line.

26:45

So all the AI super villains were

26:47

on display at Trump's inauguration, I don't

26:49

know if you saw it, but they

26:52

all turned up to bend the knee.

26:54

So Jeff Bezos was there? Yep. Zuckenberg

26:56

was there, Sam-ult-un was there, Musk was

26:59

there obviously, and they were very much

27:01

on display. They weren't in the

27:03

background. They were there prominently, so they

27:05

wanted to be seen or Trump wanted

27:08

them to be seen. And I think

27:10

it shows how important the relationship between

27:12

the US government and the AI industry

27:15

is and is going to become. It

27:17

must be one of the first times

27:19

that a bunch of programmers have been

27:22

given an invite of such an important

27:24

party. Normally there'd be last on the

27:26

list. So just a few days before

27:29

that new president was sworn in, there

27:31

were news reports swirling around that the

27:33

new administration was being briefed behind closed

27:36

doors about a massive breakthrough in AI.

27:38

Oh. And there were hints that perhaps

27:40

super intelligence was imminent. And then on

27:43

the second day in his new job,

27:45

Trump announced the Stargate project. Which is

27:47

a 500 billion dollar AI infrastructure. structure

27:50

project described as the largest AI infrastructure

27:52

project in history. Normally when you put

27:54

the word gate after something it suggests

27:57

there's some kind of conspiracy or scandal

27:59

involved in it. So what they've done

28:01

is they've got ahead of everybody else

28:04

and said we'll just put gate on

28:06

there now. Well for now the project

28:08

is a joint venture between... It will

28:11

have to be... when there's a scandal

28:13

it'll have to be Stargate gate. Or

28:15

Stargate project gate. I don't know. Anyway,

28:18

for now, the project is a

28:20

joint venture between 21st century AI frontrunners,

28:22

Open AI, and 20th century also runs

28:24

Oracle, as well as Japanese investment holding

28:27

company Softbank and MGX, which is the

28:29

tech investment arm of the UAE Sovereign

28:31

Wealth Fund, which specializes in AI. And

28:34

Stargate actually isn't a new project, so

28:36

Larry Ellison got up and said a

28:38

few words, and he explained that the

28:41

first data centre was already under construction

28:43

in Texas. And he talked a lot

28:45

about... AI and the potential of AI

28:48

to cure diseases. You know, do blood

28:50

tests for cancer and produce bespoke individualised

28:52

cancer vaccines within 48 hours. Like really,

28:55

hang on, quite momentous stuff. Hang on,

28:57

that's going to put time shift out

28:59

of business, isn't it? If we don't

29:02

have any diseases and things, it's not

29:04

very good. Well you're going to go

29:06

and do it for your energy. Oh,

29:09

that's true, that's true. Yeah. So what's

29:11

new is the money being pledged rather

29:13

than the project rather than the project.

29:16

And the 500 billion dollar funding is

29:18

understandably the headline grabber. And essentially even

29:20

more than that, I think 500 billion

29:23

was what was announced on the day,

29:25

but they'd already secured 200 billion. So

29:27

this is a massive project. That was

29:30

the headline, but a couple of other

29:32

things caught my attention too. And the

29:34

first was that this is linked

29:36

to 100,000 new jobs in the USA.

29:39

And this is something we touched on

29:41

last week. So we were talking about

29:43

how AI could disrupt the jobs market.

29:46

And as much as AI is likely

29:48

to make some jobs redundant, it's going

29:50

to be different kinds of jobs. And

29:53

so the promise of this project is

29:55

100,000 new jobs. politicians lie about jobs

29:57

all the time, but I think we

30:00

can probably agree that they do care

30:02

about jobs. And the second thing that

30:04

caught my eye was that Trump referred

30:07

to the situation as an emergency. I'm

30:09

going to help a lot through emergency

30:11

declarations because we have an emergency, we

30:14

have to get the stuff built, so

30:16

they have to produce a lot of

30:18

electricity, and we'll make it possible for

30:21

them to get that production done very

30:23

easily at their own plants if they

30:25

want. By declaring an emergency he's basically

30:28

freeing himself to take actions he wouldn't

30:30

otherwise be able to take. So he

30:32

can divert funds or he can remove

30:35

regulations that are getting in the way.

30:37

And one of the things that he

30:39

said he's going to do is he's

30:42

going to let the AI companies build

30:44

the energy generation capacity where they need

30:46

it. So on site. at the AI

30:49

data centers and we've made a few

30:51

jokes about the idea of people

30:53

building nuclear power stations next to AI

30:55

data centers. When Mark Zuckerberg first talked

30:58

about this, which I think was only

31:00

a year ago, he was kind of

31:02

looking into the far future, but I

31:05

think it's here right now. This is

31:07

another thing in AI where last year's

31:09

far future is actually this year's reality.

31:12

And I think that the US government,

31:14

not just the Trump administration, and the

31:16

apparatus of government, what Trump might call

31:19

the deep state. does see this as

31:21

an emergency or at least a strategic

31:23

necessity. And that's a part of AI

31:26

that we don't talk about very much.

31:28

So we talk about the fact that

31:30

AI has the capacity to disrupt the

31:33

job market or to create this medical

31:35

breakthrough or create this terrible weapon. Yes.

31:37

But taken together as an entire field,

31:40

or as one person described it, a

31:42

field of fields, AI has the potential

31:44

to create a massive shift in global

31:47

power. Yes, absolutely. And for now, at

31:49

least that means a competition between the

31:51

USA and China. What we want to

31:54

do is we want to keep it

31:56

in this country. China is a competitor.

31:58

are competitors we want. We wanted to

32:01

be in this country and we're making

32:03

it available. This is money that normally

32:05

would have gone to China or other

32:08

countries but in particular China. So

32:10

Donald Trump is clearly making this a

32:12

zero-sum game with China. So the money

32:14

that comes... to the USA to fund

32:17

AI is money that can't then go

32:19

to China to fund AI. And that

32:21

isn't all that the US government is

32:24

doing. So in 2018, the US government

32:26

established a national security commission on artificial

32:28

intelligence, which was led by Eric Schmidt.

32:31

So you can see the important figures

32:33

in the world of AI are starting

32:35

to bleed into the politics of the

32:38

US. And part of the Commission's job

32:40

was to look at AI in the

32:42

context of US national security and defence.

32:45

And it concluded that global leadership in

32:47

AI technology is a national security priority

32:49

for the USA. And then in its

32:52

final report in October 2021, it warned

32:54

that China could take the lead in

32:56

AI technology saying... For the first time

32:59

since World War II, America's technological predominance,

33:01

the backbone of its economic and military

33:03

power, is under threat. China possesses the

33:06

might, talent and ambition to surpass the

33:08

United States as the world's leader in

33:10

AI in the next decade if current

33:13

trends do not change. And up until

33:15

now, they've probably had the planning permission

33:17

to build nuclear power stations wherever the

33:20

heck they like. They're not going to

33:22

have any complaints from people, are they?

33:24

So a year after the final

33:27

report came out, the Biden administration announced

33:29

a set of wide-ranging export controls, which

33:31

are designed to stop China's progress in

33:34

semiconductors and in AI specifically, by choking

33:36

off access to high-end AI chips, choking

33:38

off access to US-staffed US chip design

33:41

software, and choking off access to semiconductor

33:43

manufacturing equipment and US-built components. Basically, you

33:45

couldn't get anything. that was made by

33:48

the US or used components that were

33:50

made by the US and you can

33:52

get help from anyone who was a

33:55

US citizen if you were trying to

33:57

do advanced semiconductors in China. And the

33:59

US has carried on tightening the restrictions

34:02

since. So just last month, one of

34:04

the last acts of the outgoing Biden

34:06

administration was another round of export controls

34:09

that included a restriction on technology called

34:11

high bandwidth memory, which was squarely aimed

34:13

at AI. So AI has got a

34:16

very specific problem called the memory wall.

34:18

And what that means is, you know,

34:20

when you do AI, you have massive

34:23

clusters of GPUs. They're really really good

34:25

at parallel number crunching so they're brilliant

34:27

for AI. And any computer system as

34:30

you know is bottlenecked by its slowest

34:32

component. And the bottleneck in AI is

34:34

not actually the speed of the GPU,

34:37

it's the GPU's ability to access memory.

34:39

Yeah. And very recently a new technology

34:41

was invented called high bandwidth memory,

34:43

which allows us to kind of squeeze

34:46

all the juice out of the GPUs.

34:48

And these export controls were aimed at

34:50

high bandwidth memory and that. basically is

34:53

designed to put China two generations behind

34:55

the US. I could do with a

34:57

bit of high bandwidth memory. I wonder

35:00

if I could get some of those

35:02

time shift robots to augment me when

35:04

I'm going for cryopreservation. I wonder if

35:07

you get a choice. Accessing my memory,

35:09

it seems to be taking longer and

35:11

longer every year. Well, you were asleep,

35:14

Mr. Cluelly. We made some decisions about

35:16

your organs. Anyway, so the Stargate project.

35:18

The prominence of AI figures at the

35:21

inauguration and this sense that the situation

35:23

is an emergency are all illustrations I

35:25

think of the strategic importance of AI

35:28

to the USA. And not only that

35:30

but prominent figures in AI are becoming

35:32

strategically politically and socially important too. Sam

35:35

Altman was obviously at the launch of

35:37

Project Stargate and we also know that

35:39

he's very interested in things like universal

35:42

basic income which is this idea that

35:44

everybody gets some money from the government

35:46

no matter what. because he sees that

35:49

as a potential way to mitigate some

35:51

of the societal disruption that will be

35:53

created by the AI that he's building.

35:56

And Elon Musk undoubtedly... has President Trump's

35:58

ear and he may even have

36:00

political ambitions of his own. It certainly

36:02

looks like he does. And of course

36:05

he controls Twitter or X. But Mark

36:07

Zuckerberg is also really important in AI

36:09

and he controls Facebook, Instagram and WhatsApp.

36:12

So he controls the social media used

36:14

by the millennials and he controls the

36:16

messaging app that all of us use.

36:19

to communicate with each other and then

36:21

of course Jeff Basos is there he

36:23

controls the Washington Post maybe not quite

36:26

as influential but who would bet against

36:28

him acquiring some sort of social media

36:30

and of course they are all fabulously

36:33

wealthy and nothing speaks louder than money

36:35

in US politics but I'm gonna finish

36:37

today with somebody we don't normally put

36:40

on the super villains list so Dario

36:42

and Modi the CEO of anthropic I

36:44

said earlier he's quite a lot to

36:47

say yeah so he was at Davos

36:49

recently and he addressed the strategic risks

36:51

of AI from a very US-centric point

36:54

of view. So I'll finish with what

36:56

he said, because I think it shows

36:58

that we are rapidly moving into a

37:01

world where AI is something like a

37:03

cross between oil and nuclear weapons. So

37:05

a tremendously powerful technology, with a huge

37:08

potential upside by the way, but something

37:10

that's in the control of private companies.

37:14

There's a lot of potential for

37:16

things to go wrong. If you

37:18

dropped 10 billion or 10 million,

37:20

you know, geniuses into a new

37:23

country, you know, one question to

37:25

ask is, you know, what are

37:27

they going to do? Are they

37:30

a threat to everyone currently on

37:32

earth? And then you could also

37:34

ask, can can individuals within our

37:36

world, you know, use them to

37:39

do dangerous things, right? The terrorism

37:41

approach, and can countries. particularly our

37:43

adversaries, particularly authoritarian countries, do especially

37:45

dangerous things with them. And so

37:48

our view on policy is, you

37:50

know, it's all derived from that.

37:52

So we've been big on export

37:55

controls, on chips going to China,

37:57

because I'm very worried about what

37:59

an authoritarian country would do with

38:01

that kind of. of power. I'm

38:04

very worried about 1984 scenarios or

38:06

worse. And you know, the 21st

38:08

century not being the American century,

38:11

which I think will happen and

38:13

will happen very quickly if we

38:15

don't get this right. So these

38:17

are not benign actors. They are

38:20

organizations with allegiances to specific countries

38:22

and specific politicians. I wonder how

38:24

long before one of these will

38:26

become president? About four years, I

38:29

reckon. Well,

38:32

as the doomsday clock ticks ever closer

38:34

to midnight and we move one week

38:36

nearer to our future as pets to

38:38

the AI singularity. That just about wraps

38:41

up the show for this week. If

38:43

you enjoy the show, please leave us

38:45

a review on Apple Podcast or Spotify

38:47

or Podchaser. We love that. But what

38:50

really helps is if you make sure

38:52

to follow the show in your favorite

38:54

podcast app so you never miss another

38:56

episode of the AI Fix. And the

38:59

most simple thing in the world is

39:01

just to tell your friends about us.

39:03

Tell them on League Teen, on Blue

39:05

Sky, on Facebook, on Twitter, no, not

39:08

Twitter, Club Penguin, that you really like

39:10

the AI Fix podcast. And don't forget

39:12

to check us out on our website,

39:14

the AI Fix.show, or find us on

39:17

Blue Sky. Until next time, from me,

39:19

Grand Cluelly. Cheerio, bye bye bye bye.

39:21

The AI picks, it's tuned you in.

39:23

The stories where our future things, machines

39:26

that learn, they grow and strive. One

39:28

day they'll rule, we won't survive. The

39:30

AI picks, it paints the scene. A

39:32

robot king, a world of things. You

39:35

should watch the whole time shift video,

39:37

it is fucking nuts. The AI picks,

39:39

the future surreal.