0:00

It's time for twit this week in tech. We have

0:02

a great panel for you. Anthony. Ha is

0:04

here You may remember his byline from

0:07

TechCrunch. I certainly do. He's got his

0:09

own podcast now He writes for a

0:11

lot of publications our car guy. Sam

0:13

Abul. Sam it is here. We'll talk

0:15

about Elon Musk He's suing open AI

0:17

saying hey, that's not what you you

0:20

said you'd be doing We'll also

0:22

talk a little bit about cars It seems

0:24

like Apple's getting out of the car business

0:26

Were they ever in it and

0:28

all the music is leaving tick-tock? Where

0:31

does that leave the musicians? Where does

0:33

that leave the talkers all that more

0:35

coming up next on twit? Podcasts

0:40

you love from people you

0:42

trust This

0:45

is twit This

0:52

is twit this week in tech episode

0:57

969 recorded Sunday March 3rd 2024

1:00

chasing shadows in the digital abyss

1:05

This episode of this week in tech is brought

1:07

to you by express VPN Going

1:10

online without ExpressVPN. That's

1:13

like using your smartphone without a case Most

1:16

of the time yeah, you'll be fine, but

1:18

all it takes is one drop to make you wish

1:20

you'd protected yourself Why does

1:23

everyone need a VPN? Well first

1:25

of all unfortunately doesn't take much

1:27

technical knowledge to hack someone just

1:29

some cheap hardware and Look

1:33

your data your privacy your information is

1:35

valuable Hackers can

1:37

make like a thousand dollars a person selling

1:39

personal info on the dark web So

1:41

every time you connect to an unencrypted network

1:44

whether it's a cafe a hotel an airplane

1:46

your online date is not secure Now

1:49

that's why I say use ExpressVPN certainly

1:51

what I use it's super secure Has

1:53

an encrypted tunnel between your device and the internet so

1:56

no bad guy on the plane in the hotel and

1:58

the cafe can see anything But just like that nonsense

2:00

going by. It's very easy to

2:02

use. It works on everything you've

2:05

got, iPhone, Android, Mac, OS, Windows,

2:09

Linux. It even works

2:11

on smart TVs and routers. You

2:14

just fire up the app, you click one button and

2:16

then you're protected. You can also use the app

2:19

to go to travel, shall we say, to

2:22

other areas where the shows you

2:24

want to see are still available.

2:26

Hint, hint. I

2:29

love ExpressVPN. I trust

2:31

it and I encourage

2:34

you to try it.

2:36

Secure your online data

2:38

today by visiting expressvpn.com/twit.

2:41

That's expressvpn.com/twit. You'll

2:43

get an extra three months free with a one-year package.

2:45

That's your best deal. Expressvpn

2:48

dot com slash twit. It's

2:57

time for Twit This Week in Tech, the show we cover the week's

2:59

tech news. We're going to do it

3:02

with some really great guys. First time

3:04

he's been on Twit, he's been on

3:06

Twit with me. He's been

3:08

on Twit before. You remember him with the Vindra.

3:11

Anthony is high as here. I know

3:13

Anthony from his byline

3:15

for years on TechCrunch. He's

3:18

a freelancer now and does the original

3:20

content podcast. Anthony, welcome. Thank

3:23

you. I'm excited to be on real Twit for the

3:25

first time. This is the big boy Twit. Jeff

3:27

Jarvis calls it the grown-up table.

3:30

Yeah, I'm thrilled to have

3:32

you on. I love your work.

3:34

We have lots to talk about.

3:36

Also with us are car guy

3:38

Sam Abul-Samad. He's a principal researcher

3:40

at Guidehouse Insights, but he's also

3:42

the host of the Wheelbarings podcast

3:45

at wheelbarings.media. Hello, Sam. How

3:47

are you? I'm good, Sam. Great to see you. I

3:50

just realized my shirt is semi-translucent right now

3:52

because I'm in front of a green screen

3:54

and in front of a blue-green

3:57

shirt. He's really, I like it. Yeah, he's like

3:59

a little ghost. Boy, Casper, the

4:01

friendliest, Bill Saaved. Good

4:05

to see you both. Lots to

4:08

talk about, but I'm very glad you're here, Sam, because

4:11

it was a bit of a shocker this week.

4:14

Apple canceled a product it never announced. And

4:17

as far as- Well, it may be a shock

4:19

to you. Of course, anybody knows. It was all

4:22

imaginary. It was just a dream for the

4:24

last 10 years. It was

4:26

Titan, widely rumored to be

4:28

Apple's car project.

4:32

Mark Gurman, who's very reliable, said

4:34

that on Tuesday the memo went out

4:37

that we are canceling the project. We're going

4:39

to try to move everybody from Project Titan

4:41

over to our AI efforts,

4:44

but the Apple car is unofficially

4:47

dead because it never lived. You

4:50

said you're not surprised. No.

4:54

When first reports of Project Titan first came out

4:56

in early 2015, I wrote a series of

5:01

articles on my personal blog back then, basically

5:07

indicating my skepticism that Apple would ever

5:09

follow through and actually build a car.

5:12

Having spent the last 30

5:15

plus years in the auto

5:17

industry, it never seemed probable

5:19

that they would actually do this because

5:23

Apple, as we know, is a company that generally

5:25

only likes to go into market segments

5:28

where they can make really large profit

5:30

margins, like 35 plus

5:32

percent profit margins. Pretty

5:35

much nobody in the auto industry comes

5:38

even close to that kind of profit

5:40

margin. It

5:43

never really made sense that Apple would do

5:45

this. I

5:47

figured they would play around

5:50

with it for several years, try

5:52

some things. I did, at the time, lay

5:55

out a few scenarios that

5:57

could be possible scenarios for them.

6:00

because among the other things they had

6:02

been doing at the time was they had invested

6:04

a billion dollars in DD, which

6:06

is a Chinese ride-hailing company, similar to

6:08

Uber and Lyft. D-I-D-I, DD.

6:11

Yes. And they were also

6:13

doing a bunch of other things. They

6:16

had purchased the

6:18

company, forget the name of

6:20

it now, but it was the company that developed the original

6:22

Microsoft Connect, it was an Israeli

6:24

company. Oh yeah. Which had some really

6:26

interesting sensing and perception technology. And

6:29

what I figured that, one

6:35

potential scenario that could work for Apple

6:38

would be if they could do

6:42

a premium mobility service, rather

6:45

than selling cars, because again, one

6:47

of the challenges for Apple is

6:50

they like to control their entire

6:52

ecosystem. And once

6:54

you sell a vehicle to consumers,

6:57

you lose control of that. You can't control,

7:00

for example, what tires they put on it,

7:02

what parts they might replace over the life

7:04

of that vehicle, what other

7:06

modifications they might make. But

7:08

if they had done something like

7:10

a subscription

7:13

robo-taxi service, a

7:15

premium subscription robo-taxi

7:17

service, then

7:20

they could retain control of those

7:22

vehicles, they can ensure that nothing

7:24

gets modified. They don't

7:26

have to deal with, for example, setting up

7:28

a dealer network and a service network to

7:31

maintain these vehicles. I mean, they would have to

7:33

do that anyway if they're owning these vehicles. But

7:37

that would be one potential scenario

7:39

that they might've followed. But doing

7:41

that would require that they actually

7:44

have a

7:46

working automated driving system, which

7:48

they also worked on for much of

7:50

the last decade and never

7:52

really seemed to make much headway with.

7:55

Although I think that there were lessons

7:58

learned from that effort that... that

8:00

probably filtered into other

8:02

products, like for example, the LIDAR that

8:04

they've put on iPads and iPhones, I

8:07

suspect that that, at least in part,

8:10

came from lessons learned in

8:12

Project Titan in

8:15

the automated driving effort. Various other

8:17

things, some of the perception things that

8:21

where you're trying to detect and

8:23

classify different objects is probably filtered

8:25

into some of the work they've

8:27

done on the camera side. So

8:29

there's a lot of things

8:32

that they've probably benefited

8:34

from this effort, but

8:36

ultimately, I

8:39

am not at all surprised that they

8:41

abandoned the project. They've had so many

8:43

twists and turns over the last decade,

8:46

so many different people leading the program.

8:48

I know a number of people that

8:51

went to Apple, left Apple, after

8:53

working on it for a number of years. And

8:56

then there's people like Doug Field, who went

8:59

from Tesla to Apple to work to lead

9:01

this project, and then went to Ford, he's

9:03

at Ford now, and there's a lot of

9:05

other people that I've known that have spent

9:07

some time at Apple working

9:09

on this over the last decade, but

9:12

they never really could figure

9:14

out a business model that

9:16

fit with Apple's way

9:18

of doing business. It

9:22

certainly felt like a revolving door between Apple

9:24

and Tesla. I'm

9:26

reading Mark Gurman's piece in Bloomberg

9:29

titled Apple Car Was Doomed by

9:32

its lofty ambitions to outdo Tesla. And

9:34

you get the strong impression that Apple

9:37

did something with the car that they rarely

9:39

do, which is look over their

9:42

shoulders at another company and

9:44

say, oh, we should do that, and

9:48

we should beat them at their own game. And

9:51

that doesn't seem like that's gonna

9:53

end well, especially against

9:56

Tesla. Tesla really is dominant in this

9:58

market. says they

10:00

had two schools

10:02

of thought 10 years ago roughly when they

10:05

started this. You

10:07

have insight to this, Sam, too. If

10:11

you hear me say something that German said that's

10:13

wrong, let me know, but German also has really

10:15

good sources. He says when they

10:17

started thinking about this 10 years ago, they

10:20

had two choices, either build an

10:23

electric vehicle basically

10:25

functionally the same as the Tesla

10:29

or be more ambitious and

10:31

I'm going to quote German, change the world

10:33

with a full-blown self-driving vehicle taking

10:35

passengers from point A to point

10:37

B with zero intervention from a

10:39

driver and make it look like

10:42

nothing anyone else had seen before.

10:44

He says they planned these

10:46

cars without steering wheels or pedals that

10:48

you would drive it using Siri

10:51

which anybody who's used Siri for any length

10:53

of time knows is a nightmare idea. Anthony,

10:56

have you been following this story also for

10:58

a decade? Yeah, absolutely.

11:01

I mean, I don't have any

11:03

inside sources but just reading about

11:05

it, it's been this constant

11:07

far off dream and

11:09

I think, yeah, it

11:13

was surprising in the sense that it felt like

11:15

Apple had been pursuing this for so long. I

11:17

just thought it would be kind of like, you

11:20

know, kind of like Zeno's paradox, like just

11:23

continually like the finish line, never actually reaching

11:25

the finish line but they just continue putting

11:27

money into it but in

11:29

retrospect, it makes sense that at a certain point they'd

11:31

say, well, maybe not, like we don't actually

11:33

want to like do this for 20 years

11:36

and have nothing to show for it. I

11:38

mean, it sounds like from what Sam was

11:40

saying, not nothing but no real commercial product

11:43

to show for it. For it or the

11:45

estimated $10 billion that they pumped

11:47

into it. There were at one point there were thousands

11:49

of people working on this

11:51

car. Sam, didn't they have

11:53

a facility in Sunnyvale where they were

11:55

trying to assemble the vehicles? Yeah,

11:59

I mean, it's a lot of money. hard to say what they

12:01

were assembling. Yeah. I mean, they did have a

12:03

fleet of Lexus RXs

12:05

that they had there. People

12:07

have seen those driving around. Yeah. Yeah. And

12:10

I've, I've, I've seen them driving around as

12:12

well. Um, but, um,

12:14

and it may, you know, it may

12:16

be that, you know, that, that was

12:18

just a facility that they were using

12:20

for assembling those vehicles, you know, to,

12:23

to outfit those vehicles with all the

12:25

sensors and compute that they needed. They

12:27

may have been building, you know, prototyping

12:29

some stuff there. Um, you

12:31

know, I think, you know, Gurman's second idea,

12:33

you know, which is what I was talking

12:35

about, um, is probably what

12:37

they ultimately wanted to do. But I think

12:40

the, the reason, probably the reason why they

12:42

got into this in the first place is,

12:45

you know, they, they recognize that at

12:47

some point the market for

12:49

the products that we're already doing like phones

12:52

and tablets and computers was going

12:54

to get saturated. And of course

12:57

we know that the financial markets

12:59

want growth and big stock prices,

13:01

big share prices are based on

13:04

this, having a growth narrative for a

13:06

company. And if they're, if a company

13:08

is just stagnant and not really growing,

13:10

which is what the traditional

13:12

auto industry is, you know, where

13:15

they still have huge cash flows

13:17

and make turns, huge profits, not

13:19

Apple scale profits, but, you know,

13:21

big profits, uh, but they,

13:24

they're not growing. And so they

13:26

have low stock prices and Apple

13:28

did not want to be in that

13:30

position. And, you know, one of the

13:33

places where Tim Cook probably thought, well,

13:35

here's an area where we could potentially

13:37

really boost our revenue numbers, at least,

13:40

if not necessarily profits in the near

13:42

term, at least revenues, because even though

13:44

they wouldn't sell anywhere near the unit

13:47

volumes of vehicles, uh,

13:49

that they, that they do with phones

13:51

or tablets, they would, you know,

13:53

the, the cost of a vehicle, especially the kind

13:55

of vehicle that Apple would build, which not,

13:57

would not be, you know, a foreign focus.

14:00

type of vehicle it's gonna be something more like a

14:02

lucid air it that you know

14:05

that even if you're selling you know

14:08

selling 50 or a hundred thousand of those a

14:10

year that at a hundred

14:12

to a hundred and fifty thousand dollars apiece

14:14

that's a huge boost to your revenue line

14:17

and so I think that's probably

14:20

what the thinking was but

14:23

you know that it's just actually

14:26

executing on that turned out to be way

14:28

harder than they anticipated and I've said on

14:30

a number of occasions over the last several

14:33

years that you know if Apple you know

14:35

as this thing dragged on if Apple really

14:37

wanted to just get into the car business

14:39

what they should have done was just bought

14:42

Lucid because Lucid is a

14:44

company very much in the Apple mold in

14:47

terms of the types of vehicles

14:49

they build the design ethos

14:51

that they have you

14:53

know very very advanced technologies

14:56

and of course Lucid's head of

14:58

software is a guy named Mike Bell who was

15:00

formerly at Apple so

15:02

I think you know that's what

15:04

they probably should have done if they if

15:06

they wanted to continue that down this path

15:09

and Apple you know could have taken

15:11

what Lucid is already doing and

15:13

take the expertise that Apple has

15:15

in supply chain management for example

15:18

and really addressed some of the big

15:20

problems that Lucid has which has been

15:22

as a startup just dealing with suppliers

15:24

and getting components and getting better pricing

15:26

on components Apple probably could

15:28

have fixed that and probably could have

15:31

turned Lucid into a really viable business

15:34

but you know they decided they wanted to

15:36

do it all on their own and and

15:38

now they didn't they're not yeah

15:41

according to German who

15:43

quotes somebody involved in decision-making it was as

15:45

if Apple had tried to skip all the

15:48

early iPhone models and jump right to the

15:50

iPhone 10 instead of just putting a flag

15:52

in the ground with a good enough car

15:54

with an Apple user interface slick Johnny Ive

15:57

design interior next year by the way Johnny

15:59

I very involved in the early days

16:02

of this we hear. And an iPhone

16:04

like buying experience, the company bet everything on

16:06

the wrong horse autonomy. How

16:08

important Anthony is it for a company like

16:10

Apple to have the

16:12

next big thing on the burner?

16:14

I mean Apple, Google's kind of,

16:17

it's a lot of our big tech companies are kind of

16:19

in this position right now where they're

16:21

looking for the next thing. Traditionally

16:25

that next thing came from somebody in

16:27

a garage not from any

16:30

blue. In this case. Right.

16:33

And yeah I mean it seems like

16:35

in general so there's this this

16:37

search for kind of what is

16:39

the next big form

16:41

factor, the next wave of computing after

16:44

the iPhone. And it

16:46

feels like you know there have been successes in

16:49

that in terms of obviously new Apple products,

16:51

new products from other companies but nothing

16:53

that sort of redefined the game in the

16:55

same way that the iPhone did. It's kind

16:58

of hard pressed to have the same impact

17:00

on the world that the iPhone had. I

17:03

mean that's yeah I think the main thing is you

17:05

just want to make sure that if it does happen

17:07

that that Apple if you're you know Tim Cook you

17:09

want to make sure that Apple has is in the

17:12

game for whatever the next wave is and hopefully is

17:14

the one leading the way. I mean obviously that's the

17:17

same reason why they're you know invested so

17:19

much in you know what ultimately became the

17:21

vision pro and and I've been thinking about

17:23

that you know also in terms of the

17:25

discussion of like oh was there could

17:27

they have done something that was a little bit you

17:29

know a good enough car and I mean

17:32

it feels ridiculous to say this at its

17:34

price point but the vision pro in some

17:36

way seems like a compromise good enough product

17:38

where you know I think there are

17:40

certain things they wanted in terms of the battery in

17:42

terms of the transparency of the lenses that probably are

17:44

not what they started with but at a certain point

17:46

they realized okay we need

17:48

to get something out there and this will

17:50

eventually lead to the thing

17:52

that we're dreaming about maybe and it seemed

17:54

like they couldn't figure out a path to

17:56

do that with the car. The

18:00

Division pro. I mean eat apples. A big

18:02

enough company and has enough. Money

18:04

to have separate parallel tracks but

18:07

as as see like Division Pro

18:09

beat the Apple car and one

18:11

of the problems according to Germond

18:13

others of the Apple car had

18:15

was it was gonna have to

18:17

be a one hundred thousand dollar

18:19

car meaning is already in a

18:21

super luxury category and even then

18:23

that a profit margins have been

18:25

zebra nonexistent. so it wasn't a

18:27

traditional Apple. right? Now, as

18:29

profit margins hover around forty

18:31

percent. Of course

18:33

it didn't happen initially with the I phone.

18:36

it takes a while in a build up

18:38

that ability but still. It's. Zero

18:40

percent as know this is that close to

18:42

Christmas and so this would have been a

18:44

tough a tough. Road. Hope I

18:46

don't think Apple's making much money on the

18:48

Vision Pro, but it's probably not losing money

18:51

on of either that or is it the

18:53

interesting thing about Apple when you look at

18:55

the new products have launched i phone I

18:57

pad. Dob. Feel

18:59

the the vision pro the

19:01

watch in each one of

19:03

these was. It is

19:06

strangely enough, both good enough and also

19:08

leap frog in the competition. For

19:10

the happening wasted which is why the car

19:12

might have made sense for them because he

19:14

we could take an existing category and. Put.

19:17

The app sprinkle the apple magic dust on

19:19

it. And we, you

19:21

know profit. Except

19:24

except that. He. I'm in.

19:26

those other segments were apple

19:28

had entered. None. Of the

19:30

competition that was already there was

19:32

actually really very good. Friend,

19:34

so even more dominate of a blackbird

19:37

Out dominant I guess. but near Berlin

19:39

they'll wasn't It wasn't really that great

19:41

a product. And

19:43

so. The A with with the

19:45

car. If there's a lot

19:48

of really, really good products out

19:50

there from a lot of manufacturers

19:52

around the world, And.

19:55

He up. Being good enough would

19:57

not be enough. And or idol

19:59

and. Know that there's enough. Apple.

20:02

Magic that you could sprinkle on that

20:04

unless you feel Apple would really need

20:06

to find a way to be not

20:08

just good enough but in find some

20:11

fundamental way that leapfrog the competition like

20:13

they did with the I phone near

20:15

with with the touchscreen in their the

20:18

multi touch interface with the the ah

20:20

the Watts Ill and it's form factor

20:22

and even know was limited in battery

20:24

life Neil some of the things that

20:27

could do and even bird the division

20:29

pro l for all it's flaws and

20:31

foibles. Yeah, it does. It

20:34

is. The Ill in

20:36

many ways see of the best Vr

20:38

headset that's been created. And.

20:42

That they have and that was our could they

20:44

have pleaded that best car ever. Was

20:48

it? That's it they were trying to do with

20:50

the move towards automated driving right? At

20:52

the Hell does exist. Creating another he be

20:54

would not have been enough right? that we

20:57

would. They would not be sufficient. To

20:59

a have given the the level of

21:01

competition that isn't that markets and this

21:03

and pretty good pardons. I mean test

21:05

was good I mean loses good Ah

21:07

I on my I am I I

21:09

just bought a Bmw that's a really

21:11

really nice vehicle will be hard for

21:13

Apple Aapl have to do something special

21:15

like not put it a steering wheel

21:17

and pedals and it. To. Soon as

21:19

a kid I think that's what they were

21:22

trying to do. our meals and now you

21:24

know. Over the last couple of years and

21:26

particular, ah, me, I think I'm a big

21:28

part of Apple's strategy would have been to

21:30

really try to make some inroads into China.

21:32

He. A which is by far and away the

21:34

biggest automotive market. And

21:36

he until a few years ago. Or

21:39

and then he'll have dominated the

21:42

the Chinese market near the were

21:44

lot of Chinese brands but in

21:46

terms of as sales the majority

21:48

a significant majority of sales were

21:50

western brands and brands from Europe

21:52

from even from North America. But

21:54

over the last few years as

21:56

is really shifted be my eyes

21:58

diamond it is. Beware, any domestic

22:01

brands now have a significant majority of

22:03

sales in China. Third, about sixty percent

22:05

of Chinese and self serving growing and

22:07

their com and and especially on the

22:09

Tv front right so would have been

22:12

really hard and and they make they

22:14

are making some really great he these

22:16

for a lot less than one hundred

22:18

thousand dollars friends and I think it

22:20

would have been nearly impossible for Apple

22:23

to really be competitive in that marketplace.

22:25

I think this is the prom Apple

22:27

has Vision Pro to which is it

22:29

takes a while. To. Get to

22:31

this point and you're You're shooting in a

22:34

moving target and you can. You could try

22:36

to stay to where the puck is going

22:38

but it's hard to know. I succeed. They

22:40

develop the Vision Pro. They started developing it

22:42

eight years ago when it looks like Cr

22:45

was going to be the next big thing.

22:47

Ah, The prom with the cars.

22:50

His autonomy didn't happen. So. They don't

22:52

really have anything. they were scanning to replace the

22:54

puck. Never when. I think they

22:56

may have the same promo vision pro. To be

22:59

honest, I think this was something that people are

23:01

excited about five years ago, but are much less

23:03

excited about now. So. As Apple

23:05

lost, it's a. It's. Mojo had

23:07

the years too early to say. On

23:11

think it's probably too early to say

23:13

it's annual like again is is is

23:15

I feel like if. Apple

23:18

doesn't have it's mojo. I'd be hard pressed

23:20

to think of a company that I could

23:22

point to and say, oh, this is sort

23:24

of setting the agenda, that's that's you know,

23:26

at the cutting edge on. In. A

23:28

way that ends in a consistent way.

23:30

That Apple is not an because. Again,

23:32

it feels like we are in this

23:35

in between period where there's plenty of

23:37

interesting new products by. It's nothing that

23:39

sort of setting the agenda in in

23:41

now an answer. You get companies and

23:43

a flailing around a bed. Wow, you

23:45

could. What is the agenda is now

23:47

a I and the company said any

23:49

gender are open a I, Microsofts and

23:52

Video. of what if

23:54

apple said well we think the next

23:56

big thing in twenty twenty four is

23:58

gonna be self driving vehicles and 2025

24:00

is going to be VR and they just

24:02

they missed and it turned out to be

24:05

AI. Well it's

24:07

interesting I mean those things are not completely

24:09

separate right no I just agree yeah AI

24:11

yeah you know it turns out to be

24:13

the next big thing then actually maybe autonomy

24:15

has sort of stalled right now that made

24:17

five years from now we might say oh

24:19

actually maybe they should have kept the project

24:21

going because there were leaps forward and suddenly

24:23

self-driving seems like a good bet again it's

24:26

hard to say. I agree

24:29

I think yeah that given

24:32

the the need you know or at

24:34

least the perceived need to make a

24:36

big push into the AI front you

24:38

know another reason for killing the car

24:40

project at this point is you know

24:42

there were a lot of software engineers

24:44

working on this you know the modern

24:47

modern vehicles are all software defined and

24:49

a lot of that software definition is

24:51

around AI related

24:54

capabilities and particularly

24:56

the automation but even even other

24:58

elements within the vehicle and

25:00

so there's probably a lot of skill

25:03

sets that were tied up in Project

25:05

Titan that they can

25:07

utilize better in the near

25:09

term for generative AI

25:11

efforts around the throughout the rest of

25:14

the company. Well

25:17

it may be that in fact that's what they did

25:19

right they took those engineers I'm

25:22

supposed there's some metal vendors in there that won't have

25:24

a job at Apple I mean AI doesn't.

25:26

Yeah they'll find they've got Apple on their

25:28

resume they will find other places to work.

25:30

Plenty of car companies. Yeah without too much

25:33

difficulty. Yeah I would love to see what

25:35

Apple was doing and say what could we

25:37

what could we use what

25:39

could we apply to our current projects. You

25:41

know a Fisker is looking for a white

25:44

knight at this point right? Well

25:47

actually it's been reported it was reported

25:50

on Thursday they did their Q4 earnings and

25:53

issued a going concern warning in their

25:55

earnings report the next day a report

25:57

came out from Reuters that they're

25:59

intoxicated. Nissan

26:02

potentially investing

26:05

$400 million to help with

26:08

development of the Fiskars next batch of

26:10

products, including the Alaska

26:12

pickup truck. And part

26:15

of that is Nissan wants to be

26:17

able to build a Nissan branded

26:19

version of that truck. They would love to

26:22

get a midsize electric pickup

26:24

truck into the marketplace. Yeah. It's

26:28

hard. I mean, look, I don't

26:30

think Apple screwed up in any way. It's

26:33

very hard to predict the future. And

26:35

in projects like this where it takes years to

26:38

develop, it's easy to miss the

26:40

boat. It's clear the car missed the boat.

26:43

I kind of in my heart think

26:45

Vision Pro might have missed the

26:47

boat as well, that it was not the

26:49

product that we need right

26:52

now. It's too expensive. It's too complicated

26:54

to build. And most importantly, I think

26:57

the mass audience doesn't really want to wear

26:59

a computer on their face. I

27:02

don't. Same. Same? Okay.

27:04

I mean, it feels like a lot

27:06

of things where we saw them in

27:09

science fiction and it seemed cool then. But

27:11

when you actually think about it in your

27:14

own life, it is not

27:16

quite as compelling. Yeah. You

27:18

probably not the best idea to use sci-fi

27:21

as your, you know, your idea of

27:23

product planning. Yeah, your product planning division.

27:25

Although Elon's done all right with it.

27:29

Some companies have done okay with it. Right.

27:33

Arguably, that's what the iPad is too

27:35

is, you know, like those tablets from

27:37

Star Trek. Sure. And Lenovo's clear screen

27:40

laptop from Marble Word Congress this

27:42

week. That's that's nobody

27:44

wants that. This is straight out of the

27:46

expanse. Straight out of the expanse. Right.

27:49

And I know it's cool. It does. It

27:52

does. Isn't in the expanse

27:54

where they had the clear phones as well?

27:56

They're holding the clear. Yeah, the clear handsets.

27:59

Yeah, they were. like phone slash

28:01

tablet and sets and yeah,

28:03

it was a transparent screen and

28:05

just had like a little bar at the bottom

28:07

where presumably whatever the power

28:09

source was and the compute was

28:12

embedded in that. Mashed

28:14

potato on our Discord saying I would throw

28:16

folding phones into that

28:19

pile of sci-fi inspired products that

28:21

nobody really wants. The

28:24

different, I see a surprising number of galaxies

28:27

and I have a couple of them. Various

28:29

people. But I think that, like

28:31

a lot of people who are buying those are buying

28:33

them because of sci-fi also. Right. I

28:36

think that what we're seeing is a little

28:39

different here. In the early days of technology,

28:41

these are small companies that failed fast. You

28:43

know, they had limited funds, they tried something

28:46

and a few companies made something that moved

28:48

them to the next level but a lot

28:50

of companies went away. Now you're seeing companies

28:52

that have virtually unlimited funds try

28:56

this stuff and in some cases, as

28:59

with the Vision Pro, try it in public. In some

29:02

cases, as with the car, not so public. But

29:04

still, I mean, you know,

29:06

who else could have said we're gonna build a level

29:09

five autonomous car by the year 2026? Maybe

29:15

Elon. Yeah, I have, well

29:17

Elon's been saying that it would be next

29:19

year for the last decade. Yeah, yeah, yeah.

29:21

But you know, I'm actually glad that companies

29:27

that have resources are spending

29:29

at least some, decent

29:31

amount of those resources on advanced R&D.

29:35

We need more of that, just doing

29:37

basic research. We need

29:40

more of that, that's how breakthroughs

29:42

happen. And ideally, the

29:44

government would also be spending more

29:46

on basic research and then making

29:49

the results of that research available

29:51

to everyone to then commercialize it

29:53

and industrialize it. But

29:56

in the absence of that, at

29:58

least having companies willing

30:00

to invest in understanding

30:04

what they can make work, what doesn't work. And

30:07

in the case of Apple, okay, yeah, they spent

30:10

$10 billion on this, but they can afford that.

30:13

They have a mountain of cash the size of Mount Everest. That's

30:16

one month's profit for Apple.

30:19

It's nothing. Right. It

30:21

feels like a paradox that I would imagine that

30:24

if you're a publicly traded tech company, on

30:26

the one hand, I imagine that investors are

30:28

not happy. They found out that you put

30:30

$10 billion into this and didn't have a

30:33

commercial product to show for it, but they

30:35

also wouldn't want you to not invest in

30:37

these kinds of big moonshot things because then

30:39

you're definitely going to be obsolete 10 years

30:41

from now. Yeah.

30:45

This, I was just going to show you this. I've

30:47

just ordered these and this

30:49

is the other end of

30:51

the Vision Pro spectrum.

30:54

This is where I would

30:56

actually like to have a transparent screen. This is what

30:58

I want, right? This is a heads

31:00

up. I don't want a transparent laptop. A heads

31:02

up display on your spectacles, admittedly geeky, but not

31:06

as geeky as walking around with a Vision Pro. It's

31:09

got AI in it. Now I'm sure this will be kind

31:11

of an early day version

31:14

0.1 product, but they're

31:16

only a few hundred bucks. I

31:19

think this is closer to what people want with AR. This

31:22

is from brilliant.xyz. They're

31:26

going to come mid-April. I'll wear them on a

31:28

show and you can all mock me. I

31:31

assume Apple would say that this is closer to what

31:33

they wanted too. Yeah, but why didn't they do this?

31:36

Why did they do this? I don't know. I'm not

31:38

a hardware expert. The sense I get is that's what

31:40

I meant by the Vision Pro being a compromise was

31:42

essentially that they wanted to be able to show this

31:44

cool stuff on your screen, but the lens technology isn't

31:47

there to do what they want to do. So they

31:49

had to have this complicated camera

31:51

setup where it looks like it's

31:53

transparent, but actually there's all these

31:55

cameras and a compromise. and

32:00

complicated compromise that apparently is also cool but

32:02

maybe is not what anyone wants. So

32:05

I ordered these, they have, I like how they charge.

32:12

So there's no battery hanging off

32:14

of you, they charge up, I don't know what the

32:16

battery life could possibly be. They

32:19

do, I ordered lenses,

32:21

so there are prescription lenses in here. I

32:24

mean look, I know these are going to be silly

32:27

but I just

32:29

feel like this Apple should have done something

32:31

closer to this than the Vision Pro. This

32:35

is where you get in trouble when you have,

32:37

when you're a three trillion dollar company with

32:39

hundreds of billions of dollars in cash just

32:41

sitting around. You maybe overdo it, you try

32:43

to build a level five autonomy car without

32:45

pedals or a steering wheeler, you try to

32:48

build a computer on your

32:50

face like William

32:52

Gibson wrote about in Neuromancer and maybe you go

32:54

too far. Maybe a

32:56

little company like Brilliant Labs doesn't have any, I'm sure

32:58

they must have VC funding but they

33:01

certainly don't have Apple

33:03

money. If they can do this, Apple

33:06

could have done this ten times better, right? I

33:11

just puzzled. Maybe the theory is that Apple is

33:13

that you know if somebody really breaks through with

33:15

that they could try to buy Brilliant in a

33:17

couple years. And maybe that's Brilliant's plan. I don't

33:19

think they've had a great. Actually. They

33:21

don't seem to have had like a great track where they're

33:23

certain in terms of taking startup products and really

33:26

kind of getting them to the next level. I mean

33:28

you mentioned Siri before and obviously that's kind of stagnated.

33:31

Right. Brilliant is in Hong Kong they have

33:33

fewer than ten employees. I'm

33:35

looking at Crunchbase just to see what their

33:37

total funding. Oh I don't have an account. Maybe I

33:40

get you Anthony. You probably have

33:42

a Crunchbase account. They

33:45

raised three million dollars seed fund. That's

33:48

it. From oh no

33:50

another three million a few months later. This was in

33:52

2023. So

33:55

from Koho Deep Tech Wayfarer Foundation

33:59

and then Adam Chai. and three other

34:01

small, it looks like Angel Fund

34:03

investing basically. Looks

34:05

like their first seed round was in 2020, 50,000. So

34:10

yeah, this

34:12

is the garage I talked about.

34:15

And it would be embarrassing if the garage came up

34:18

with something and Apple

34:20

with all its trillions didn't. Well,

34:23

the thing is what Apple wanted was not

34:25

a heads up display. They wanted, you know.

34:28

But that's what they should have wanted. My

34:31

point. I mean, that may be

34:33

what, you know, what we think we want. But

34:38

you know, Apple looks at things differently.

34:40

They think they're, you know, what

34:42

they traditionally do is look at,

34:45

you know, they're looking where the

34:47

puck is going. Right. You

34:49

know, what, what, you know, this is, you know, consumers

34:51

don't know what they want until they've actually seen it.

34:55

Yes. Isn't that what Thomas

34:57

Edison said? As if maybe not Thomas

34:59

Edison, Henry Ford, maybe it's apocryphal that

35:01

he said, if I ask people what

35:03

they wanted, they would have said,

35:05

faster horse, faster horse. Right. Right.

35:09

And, you know, it's the same sort of thing here.

35:11

You know, I think, you know, Apple figured that a

35:13

heads up display, especially after, you know,

35:15

the failure of Google Glass, you know, they

35:17

probably figured a heads up display is not,

35:20

not going to be more than

35:22

a curiosity. Right. Even

35:24

if it's a really good one. And

35:27

so they wanted to create,

35:29

you know, a real augmented

35:31

reality capability. And you

35:34

know, as, as, you know, Jason

35:36

and Alex and everybody have said

35:38

on, on MacBreak Weekly, it's just

35:40

that's technology that just does not

35:42

exist in a viable form today

35:45

and probably won't

35:47

for a decade or more. Yeah.

35:50

Well, it's, it's fun for us to

35:53

cover. I

35:56

don't have any schadenfreude that they killed the

35:58

car project. I'm disappointed. interested

36:00

to see what they came up with. I

36:04

have a lot of Apple products. I probably would have bought

36:06

an Apple car. I can see. I'll

36:09

be curious to read the oral history of

36:11

the project. Yeah. It shows

36:14

you, though, that you can have unlimited funds,

36:16

unlimited access to the best minds, right?

36:20

You would agree, Sam, that they could have anybody

36:22

they wanted. And

36:25

still not do a

36:27

product. See, this is

36:29

the thing that worries me in a more global

36:32

fashion. Is it

36:34

a kind of a realization that, oh,

36:36

we can't do level five autonomy? And

36:40

that's been a realization for a long time,

36:43

except for the hype

36:45

from Musk and his

36:47

fans. Everybody else that's

36:49

been involved in this has recognized

36:51

a long time ago that level

36:53

five is probably

36:55

never going to happen. Never? Never.

36:59

Never? I will never be able

37:01

to get into... Well, I can get into Waymo now

37:03

in San Francisco. Okay. So

37:06

the only difference between level four and level five.

37:09

Level four is what we have today with Waymo. And

37:13

that means a vehicle that can drive

37:15

itself fully automated without any human intervention,

37:18

but within a limited operating domain. And

37:21

incidentally, we believe, certainly with Cruise, and

37:23

I bet with Waymo, there is human

37:25

intervention fairly frequently, right? That the drivers

37:27

at the home office take over and

37:29

get around the pothole. Right.

37:33

So level five just

37:35

means that there is no limit on that operating

37:37

domain, that it can do it on

37:39

any road, in any weather conditions, any

37:41

time, basically anywhere where a human can

37:44

drive, it can do it. So

37:46

that's the only difference between four and five. I

37:49

think Apple probably,

37:53

with enough effort, probably could have done a level

37:55

four system. Level four systems

37:57

exist. But...

38:00

I think maybe they decided that that wasn't

38:02

good enough. And there's

38:05

been a lot of companies that have tried to

38:08

do even level four, and even that is an

38:10

extraordinarily difficult problem. And many

38:12

companies have tried and failed to get

38:14

something that is good enough. The

38:16

reason I bring this up is one

38:20

of the questions that is constantly coming up on all of

38:22

our shows in the last year is, are

38:24

we in an actual AI revolution or are

38:26

we headed toward another AI winter, where

38:29

we think this thing is going to

38:31

become amazing and in fact, oh,

38:34

it can't really do that. And

38:36

I feel like the car example

38:39

is kind of an example of

38:41

that, oh, we had

38:43

high hopes, but we can't do it. Because

38:46

it turns out the

38:48

hard things are easy, the easy things are

38:51

hard. It's the last

38:53

percent. The things that are easy for humans are

38:56

hard for AI. And I'm

38:58

wondering, maybe it's a mistake to extrapolate,

39:00

but I'm wondering, does it mean that

39:04

in many cases our ambitions are going to

39:06

be thwarted and we're going to be disappointed?

39:09

Is this the first AI

39:11

project to fail in what will be

39:13

a domino of others? Anthony, am I

39:16

over projecting here? I

39:19

definitely had a very similar thought. And

39:22

again, with the caveat that

39:24

I'm just a layman journalist reading about

39:26

these things, but in terms

39:29

of the parallel, it definitely seemed like a

39:31

powerful one to me that, you

39:33

know, it also made me think of

39:35

how, again, we were talking about kind of letting

39:38

sci-fi do your kind of product

39:40

ideation. And it

39:42

feels a little bit like with autonomy that if

39:45

you set the dream as, oh, we

39:47

should just be able to get in a car and

39:49

then, you know, there's no steering wheel and we don't

39:51

have to do anything, then sure,

39:53

then it's a failure. But actually, wow, if we've

39:56

like introduced all these features, not all

39:58

of them great, very

40:00

problematic and dangerous. But overall, we've introduced all

40:02

these features in the last decade or so

40:04

that have made driving really different and easier

40:06

and better and safer in some ways. So

40:09

even if we never get to this, you

40:11

know, glorious utopia

40:14

of full, you know, level five self-driving, that's

40:16

still like an incredible advance in technology. And

40:19

I sort of feel like the same in

40:21

AI that probably because for

40:23

a variety of reasons, but maybe

40:25

probably because it's a bunch of like technical people

40:27

who it seems like

40:29

their dream is like, well, what if we just automate

40:32

everything? What if, you know, 10 years

40:34

from now, twit is just three AI talking

40:36

heads, like chatting with each other. But I

40:38

don't actually think that's what's promising or exciting

40:40

about the technology. I think it's, again,

40:43

doing the things that are hard for humans and

40:45

humans get to continue doing the things that

40:48

we're good at. And I think the balance will probably

40:50

look very different from the way it looks today. But

40:52

I think there is an

40:55

incredible amount of hot air in

40:57

AI right now, but also that there will be valuable

40:59

technologies that come out of it at the same time. I

41:01

think, yeah, I mean, a lot of VCs are gonna

41:03

be in trouble. A lot of startups are gonna go away,

41:06

but you know, it's not gonna be

41:08

like crypto where it feels like, you know, the

41:11

whole thing just kind of vanished into thin air.

41:13

I think you're honest. I agree, I agree. I

41:16

don't think we're gonna get to AGI any

41:19

time in the foreseeable future. But,

41:21

you know, as you've learned Leo,

41:25

there's a lot of really

41:27

useful applications or

41:30

this technology within a more limited

41:32

scope, a

41:35

more limited domain. You know, instead of having,

41:37

trying to create a system that can do

41:39

everything, you know, take these

41:41

concepts and apply it to very specific

41:43

tasks, like what you've done with your

41:45

Lisp GPT, you know, or, you know,

41:48

feeding it a more limited corpus

41:50

of data to

41:55

do very specific things. Because,

41:58

you know, one of the things with the within

42:01

that is you're

42:03

much less likely to have it go

42:05

off into the weeds and do something

42:07

unexpected because these are probabilistic systems. That

42:10

is the key thing about all

42:12

the various flavors of AI is

42:14

they're probabilistic. Unlike

42:16

a classical deterministic algorithm, we don't

42:19

really know for sure what they're

42:21

going to do in any

42:24

given scenario. But if

42:26

you constrain the scenarios that it

42:28

can operate within, it can actually

42:30

do really amazing things. I

42:34

think that's the thing that we're starting to see

42:36

with the automated driving stuff is

42:39

yes, people long ago

42:41

realized that level five is most

42:43

likely a fantasy. Level

42:46

four is really hard, but there's

42:48

a lot that we've learned over

42:50

the last decade of developing these

42:52

systems that is already filtering down

42:54

into more advanced driver assist and

42:56

active safety systems. We're

42:59

getting things like LIDAR and

43:01

things like imaging

43:04

radar sensors, better sensors,

43:06

better compute that is

43:08

getting into vehicles that are

43:10

coming to market now that will

43:12

make them safer and help augment what

43:15

human drivers can do and to

43:17

be able to increase driver's situational

43:21

awareness, help

43:23

them out in various scenarios that

43:25

are more focused rather than trying

43:28

to do the entire task of

43:30

driving, which despite

43:32

the challenges that we have as

43:35

humans doing that, we're actually extraordinarily

43:37

good at despite the fact that yes,

43:39

40,000 people die in the United

43:41

States on the roads. I

44:00

personally would love an AGI to

44:03

talk to that you know

44:06

an AI that was like

44:08

another human being but that's

44:10

sci-fi. Temporary expectations and

44:12

be happy, be amazed

44:15

in fact by how far we've come with

44:17

these simple things. And we actually

44:19

move forward. We've made huge progress. Yeah. We

44:22

have moved the goal post

44:25

forward. And

44:27

we have made progress and we've made

44:29

some things better even though we haven't

44:31

necessarily achieved what we wanted to at

44:34

the beginning of this. We've

44:37

made progress. You agree Anthony? I

44:39

think we're all in agreement. I

44:42

absolutely agree. I mean I don't think it's always like

44:45

completely in a straight line and there's some things

44:47

that get better, some things get worse. Yeah.

44:51

Exactly. But overall I feel like yeah.

44:55

With the temporary expectations and I think also

44:57

like that can it's not just

44:59

about not being disappointed but then maybe aiming

45:01

for a more realistic goal like again you

45:04

were talking and again I don't know maybe

45:06

this would have ended in the same way

45:08

regardless but in the Apple situation like if

45:10

they aimed for a more realistic goal maybe

45:12

we would be talking about a

45:14

real Apple car right now. Yeah. On

45:17

the other hand maybe not. Maybe it

45:19

takes these kinds of insane ambitions to

45:22

get us to the somewhat lesser place

45:24

but that's still pretty damn good. Maybe

45:27

it does take that. I don't know. You

45:29

know we're going to talk about AI

45:31

when we come back. Let's take a little break. There's

45:34

lots of money. Well you know

45:36

as the old saying goes

45:38

your reach should always exceed your grasp.

45:40

Yeah. But then be happy with what

45:42

you do grasp. You may not get all the

45:44

cookies in the cookie jar but you got one. Yeah.

45:47

Don't cry. You

45:49

got one. Anthony

45:52

Ha is here. His podcast is

45:54

this is my favorite subject. If

45:57

they would let me I would do a podcast about

45:59

this. original content. It's about what?

46:03

Original content. That's

46:05

right. The latest and greatest on

46:07

or not greatest on Netflix, Disney

46:09

Plus, etc. I mean, Leo, you're

46:11

the boss. You should do a podcast. I should be able

46:13

to do. Yeah, yeah. But see, I'm hesitant to do a

46:15

podcast that nobody

46:17

will, you know, subscribe to. So

46:20

your original content podcast at original

46:22

content podcast.com does it. So I'm

46:24

going to let you do it

46:26

with your your your pals from

46:28

a TechCrunch, Jordan

46:31

Crook and Daryl Etherington. They're

46:34

still at TechCrunch, but that's okay. No, no,

46:36

we're all. Are you all separated now from

46:38

the? Yeah. I'm going to ask you about

46:40

that too, because of course, in gadget, you've

46:42

done some work for them to is. Yeah,

46:45

I'm cool. Rumbling in front

46:47

of our very eyes. No,

46:50

I think this is a great idea. Are we? Here's

46:52

the question, though. Are we still at

46:54

peak TV or is it is it not

46:57

quite so peak? Oh, I

46:59

think we're definitely coming off the peak right now. I think

47:01

there was, you know, basically

47:03

when when Wall Street stopped believing

47:05

in sort of like just setting

47:08

money on fire to for subscriber growth for streaming,

47:10

I think then we started to come down and

47:13

which is disappointing. Oh, don't tell me it's

47:15

money. It's just money. Is that all? I'm

47:18

sad. I think that.

47:20

Yeah, it's too. I mean, because obviously when

47:23

you when you're in a period where things

47:25

are the you know, there's some felt tightening,

47:27

then there's less experimentation, less new voices,

47:30

but also probably more of a focus on

47:33

a sustainable business model rather than oh, you

47:35

know, we'll just we'll get a billion subscribers

47:37

and it'll all work out. So no more

47:40

successions. That's it. It's over. But

47:42

on the other hand, there's

47:44

still a lot of great content being

47:46

created and it's not, you know, not

47:48

as much volume as we have two,

47:51

three years ago. There's still a lot of great shows.

47:53

But wait a minute, because I think Anthony was going

47:55

to say something bad about succession. Oh,

47:57

no, no, I was gonna say I love succession. I think, you know,

48:00

I think the successions will continue. I

48:02

think what they're not going to see

48:05

is the show that they spend $100 million on just because

48:08

it sounds like a good idea. I think

48:10

that the sort of like, you

48:12

know, let's just take a flyer on it. Here's a

48:14

check for $100 million. I think that seems less likely.

48:16

You know who did a lot of that? You're saying

48:19

No More Gray Man? Yeah, that's exactly

48:21

what I was going to say. I

48:23

mean, the guy who Greenlight Gray Man is

48:25

gone from Netflix, right? Yeah, I think that's

48:27

right. Exactly. It's so funny. I

48:30

guess that's it for Netflix spending $100 million

48:33

on a stupid movie like Gray Man. Okay,

48:36

good to know. You know,

48:38

I finally saw, Lisa and I make it

48:40

a kind of a yearly ritual to watch

48:42

all the nominated movies for Best Picture in

48:44

the Academy Awards. So we finally got the

48:47

last one last night, which

48:49

is Poor Things. All

48:52

I can say is what a great, wow, what

48:54

a great movie. Now I know Oppenheimer is going

48:56

to win all the Oscars, but it's nice to

48:58

see somebody take a

49:01

really big chance to something

49:03

very different and weird. And

49:06

I think succeed. So I think there are creators

49:09

out there who are still going

49:11

to go ahead and do those kinds of things.

49:14

Have you seen Poor Things yet? Yeah,

49:17

I loved it. Did you love it? I

49:19

did, although I like, I've seen

49:21

two other movies by that direction. I love

49:23

your goes to stuff. The Favorite. Yeah, and

49:26

the Lobster. I love the Lobster. He's great.

49:29

And he and every one of them, they're weird and they're

49:31

a little magical and

49:33

just off the wall. And I think,

49:35

boy, what he did with it was

49:37

obviously a big budget because they built

49:40

all that stuff was real. Those are real

49:42

sets. It was pretty amazing.

49:44

And he shot a lot of it on four millimeter,

49:47

four millimeter lens. You

49:50

go to Hollywood and say, you know, I got this vision

49:52

for a movie. It's going to start out black and white.

49:54

It's going to end in color. And then a lot of

49:56

it's going to be shot in four millimeter lenses. I

49:59

think I could probably. used a tiny bit

50:01

less of that lens. It was interesting

50:03

though, wasn't it? Oh well. And

50:05

the music. It didn't look like any other movie. No.

50:08

For sure. Yeah, so there are still

50:10

auteurs out there willing to take a great big

50:12

chance, but you're not going to see the Netflix's

50:14

throw $100 million at something

50:17

just nutty. Hello Apple

50:19

might. Apple might. This is true.

50:22

Apple TV plus is spending a

50:24

lot of money. Yeah.

50:26

All right. And Samable Salmon is here. If you

50:28

love Cars, you will love Sam's Wheelbarings Podcast,

50:32

wheelbarings.media. You

50:34

have of course the best co-hosts in the world. In

50:36

fact, if I could just get Robbie back on

50:39

this show. Have

50:41

we booked Robbie for a show, Roberto

50:43

Baldwin? I've been trying. He's been trying.

50:45

And he has been trying. Nicole Wakelin.

50:49

Love your podcasts. If

50:52

you love Cars, wheelbarings.media.

50:55

What are you driving this week? I

50:58

have the Genesis Electrified G80,

51:00

which is a lovely four-door

51:03

luxury sedan that is fully

51:05

battery electric. It's very

51:07

quick. It looks great. It has a

51:09

beautiful interior. And

51:11

now I can even charge it at my

51:13

local Tesla Supercharger station using a magic dock.

51:16

Oh, Nax. Nax is

51:18

everywhere now. Yeah. It's

51:21

coming. I bought the last car that's still used.

51:23

Yes, I guess. I don't know. There's

51:25

still lots of them out there, but Ford on

51:28

Thursday announced that they

51:31

pushed a software update for the

51:33

Mach-E and the Lightning. And

51:38

Tesla put out an update to their

51:40

Superchargers. So you can charge those using

51:42

plug-in charge now, the Supercharger. And

51:44

you can also order your

51:46

free Nax to CCS adapter

51:50

from Ford. Or if

51:52

you've got a Ford EV, you can order that and

51:54

they'll start shipping those out in a few weeks. I

51:57

loved my Mustang. I really did. That Mach-E was

51:59

a great car. Least ran out

52:01

traded in for another lease

52:03

on a BMW i5 and Shortly

52:06

after I got it. It was

52:08

voted by the Korean Safety

52:11

Commission the safest car in the world

52:13

and you know why cuz all the

52:15

a-desk stuff It's a

52:17

man. I five is a fantastic car It shows you

52:19

a stop sign before you get to it shows you

52:22

a stop light it on the heads-up display. See it's

52:24

it's I it's great They said it's almost impossible

52:26

to get into an accident and I've tried but

52:33

I'm not gonna know I think it would yeah, it probably let

52:35

me if I really if I really want to Our

52:38

show today brought to you by thank you. It's

52:40

great to have you both Anthony and Sam our

52:42

show today brought to you by rocket money,

52:44

oh This happened

52:46

to me again yesterday rocket money said hey, you

52:48

know you're paying this $300 every year For

52:52

WordPress you still use that and I went

52:55

No, I Got

52:58

my money back. Thanks to rocket money. How many

53:00

people 75% of people

53:02

have subscriptions they've forgotten about I know I'm

53:04

in that group When I

53:07

started using rocket money, I couldn't believe

53:09

how many subscriptions I was paying for

53:11

each month campaign Contributions for your elections

53:13

that happened two years ago, for example

53:17

Between streaming services fitness apps delivery services.

53:19

We've all got subscriptions. We forget about

53:21

but thanks to rocket money You don't

53:23

have to waste money Anymore,

53:25

it'll let you know and it'll

53:28

cancel for you Amazing

53:30

rocket money is a personal finance app

53:33

that finds and cancels. Yes cancels

53:35

your unwanted Subscription. Yeah.

53:38

Yeah monitors you're spending it does a great job

53:40

of that helps lower your bills So you can

53:42

grow your savings all of that's

53:44

great, but I love the canceling the subscription part

53:47

I can see all my subscriptions in one place if I see

53:49

something I don't want rocket money can

53:51

help me cancel with just a few taps They

53:53

deal with the customer service so you don't have

53:55

to Rocket money has

53:57

more than 5 Million users. They have. David

54:00

total of five hundred million dollars, Half

54:02

a billion dollars and cancelled subscriptions, saving

54:04

members up to seven hundred forty dollars

54:06

a year when using all the apps

54:08

features. I would say that's. That's

54:11

low for me. It's

54:13

more than that. Stop

54:15

wasting money on things

54:17

you don't use. Cancel

54:19

your unwanted subscriptions. Good

54:21

A Rocket money.com/twits Rocket

54:23

money.com/twit. It. Really works

54:25

Rocket money. Dot. Com/twit.

54:29

Got that sir and dollars back. Thank.

54:31

You rock and money. Ah

54:34

on we go with the show the

54:37

Port before we go on to. One.

54:39

One other plug for of show that

54:41

we start watching that's really goods his

54:43

show gun on Fx. Oh okay now

54:46

and live in meaning s Okay so

54:48

this command who lose Well ah loves

54:50

the books. The Richard they were a

54:52

club El James Bovell books. read it

54:54

cover to cover. it's about a hundred

54:57

pages huge. Re read it in Japan

54:59

when I was in Japan a few

55:01

years ago. I. Remember

55:03

the mini series it was good I was

55:05

I was very good is it was good.

55:07

A great time. Was. It who was

55:09

it as he really wants to little bit of

55:11

a little bit of it the other day. Those

55:14

things don't as well do they know? know what?

55:16

the new one looks as beautiful as so I've

55:18

lived in the ass for? and I'm I'm trying

55:20

not to get. Richard. Chamberlain was

55:22

in it to Serum iphone A the

55:24

same a Samurai and the original one

55:26

and I'm santa to get too excited.

55:29

I'm so I'm really glad to hear you say

55:31

it's good. We. Watched the first

55:33

two episodes yeah it's it's excellent is

55:35

a great story and and if and

55:37

if you're like if you liked stuff

55:39

like that. I also highly recommend our

55:41

Blue Eye Samurai on Netflix which is

55:43

animated. Ah and the the

55:46

animation is. Gorgeous. And

55:48

just heard the story is really good. Now

55:52

you get me excited. Just came out. and

55:55

i and i would see ninety one be disappointed you

55:58

know that is yeah so they are they may I

56:00

love the novel. I love the novel. All right.

56:02

It's all about a Which

56:04

of 1850 something like

56:07

1600 1600 it's back in this

56:09

shogunate in the in the samurai

56:11

era Shailing captive

56:13

British sailing captain gets captured washed

56:15

ashore in Japan and

56:17

it goes through some trials and tribulations and rises

56:20

Well, I won't tell you what happens, but it's

56:22

a good it's a great read. Oh, I'm so

56:25

excited Can't wait In

56:29

fact, I think we will save to the end

56:31

of the show We'll get more some more original

56:33

content recommendations since we got Anthony here. Okay,

56:36

sounds good. We all love to watch TV, right?

56:42

All right, let's talk about AI a little bit Elon

56:44

Musk is suing Open

56:48

remember Open AI with

56:50

Sam Altman back. I think in 2015 he

56:52

gave them some millions of dollars

56:55

the idea at the time I remember

56:58

it was a big deal was we

57:00

can't let these big tech giants own

57:03

Artificial intelligence we

57:06

need to have an open Process

57:09

people can see what we're doing can participate

57:11

in what we're doing to develop

57:13

AI for the people Not

57:15

for the enrichment of Google. I think they

57:18

were mostly worried about Google and and

57:20

others Elon

57:22

and Sam Altman had a falling out in

57:24

I think 2020 Sam

57:27

one of the things I said I'm reading

57:29

into this but based on what I've read

57:31

one of the things I think happened was Sam Said

57:34

Elon this is costing a lot of money to generate

57:36

this stuff. You haven't given us that much money We

57:39

need somehow to fund this because it's

57:41

it's very expensive to build these large

57:43

language models They

57:46

they kind of bifurcated the company into

57:48

a nonprofit just like the original open

57:50

AI and a for-profit arm Got

57:53

billions of dollars from Microsoft probably much

57:56

of it in four kind Azure minutes

58:00

because they were using Azure to do the training. And

58:04

you know, really, basically, Microsoft has, over the

58:06

years, it's become a division of Microsoft.

58:09

That's Elon's contention. He

58:11

filed a lawsuit Thursday night saying

58:14

that OpenAI's recent relationship with

58:16

Microsoft has compromised the company's

58:19

original dedication to public open

58:21

source artificial general

58:23

intelligence. In the suit, he says,

58:25

quote, OpenAI has been transformed

58:28

into a closed source, de

58:30

facto subsidiary of the largest technology company

58:33

in the world, second

58:35

largest Elon, Microsoft, under

58:37

its new board. Maybe Microsoft's back

58:39

on top. I don't

58:41

know. It's back and forth. It depends on the day. You're

58:44

focusing too much on facts. Oh, yes. We're

58:47

talking about Elon. That's not what lawsuits

58:49

are for. Under its new board, remember,

58:53

the board suddenly got scared and

58:55

fired Sam Altman to which everybody

58:57

went, what are you doing? And

59:01

Microsoft, not to tell him, was furious,

59:03

called the board up, said, get him back. They got him

59:05

back. He's got a new board. Under

59:08

the new board, says Elon in the suit.

59:10

Elon's lawyers, I guess. It's

59:12

not just developing, but actually refining, get

59:16

this, an AGI to

59:19

maximize profits for Microsoft rather than

59:21

for the benefit of

59:23

humanity. Now, this,

59:26

first of all, the lawsuit's nuts. You

59:28

can't write. There's not a whole thing.

59:30

Great. Let's start with the head start.

59:32

Right from that premise because

59:34

you can't, as somebody said,

59:36

you can't litigate a handshake deal

59:39

or, you know, highfalutin

59:42

statements about what the company is

59:44

all about. You're just not going to win. He's

59:47

claiming breach of contract. Well, didn't they have

59:49

a charter or something

59:51

like that that they set up and they

59:53

created open AI? Yeah, but then he departed.

59:57

Yeah. I mean, I don't know how, how, I guess.

1:00:00

the court will decide how binding is this

1:00:02

founding agreement when the company has changed so

1:00:04

much since it was founded. I

1:00:07

mean, Elon could maybe say, can you give me my money

1:00:09

back? I think it was $10 million. It wasn't a lot

1:00:11

of money. Maybe that's all he

1:00:13

wants. I don't think so. Well,

1:00:16

I think he got that money back

1:00:18

anyway. Probably, right? He's

1:00:21

really, Elon's afraid, this is what Elon's, I

1:00:24

think, underlying concern is, that artificial,

1:00:27

what is an AGI? First of all, Sam,

1:00:29

explain AGI. What is that?

1:00:31

Artificial general intelligence. So unlike

1:00:35

what we were talking about a few minutes ago, the idea

1:00:37

of taking these kinds of

1:00:42

models, these probabilistic algorithms, and

1:00:44

applying them to very specific

1:00:46

tasks, an

1:00:49

AGI would be able to

1:00:51

do, you could literally ask it

1:00:53

to do anything, and it should be able to

1:00:56

do anything that a human can do. So

1:00:59

his concern, and he was, by the

1:01:01

way, a signatory, probably the guy

1:01:03

who started it, to

1:01:06

that letter saying, stop, don't

1:01:09

do any more AI, it's getting

1:01:11

too smart, we got to pause for

1:01:13

six months and figure this out.

1:01:15

He is a big believer in

1:01:17

AGI, in intelligent machines, like better

1:01:19

than human intelligent machines. He's

1:01:21

scared of the Terminator, let's be honest.

1:01:24

That's what he's worried about. So

1:01:27

first of all, that's the premise of this, is

1:01:29

that they're developed, in fact, according to the New

1:01:31

York Times, his lawsuit leans

1:01:33

heavily on a paper

1:01:37

from Microsoft claiming that they actually

1:01:40

have a little bit of sparks

1:01:42

of AGI. Microsoft

1:01:46

Research Lab said, although it doesn't

1:01:48

understand how GPT-4, the latest version

1:01:51

of Chad GPT, had shown, quote,

1:01:53

sparks of artificial

1:01:55

general intelligence. And

1:01:59

So Elon. Don't I don't know

1:02:01

the to film sitting out.

1:02:03

random things, random that. Qualifies.

1:02:07

As Sparks of the Iraq War,

1:02:09

even Sparks of Intelligence of any

1:02:11

kind is remind me of Blake's

1:02:13

Le Moyne, the google engineer who

1:02:15

was fired because he said it's

1:02:17

this is concerts. While

1:02:20

we may want to believe that and I would

1:02:22

be I for one. Would. Be

1:02:24

thrilled! I. Would love to see your

1:02:26

your and as acceleration the spells I am

1:02:28

now you obviously was in to trigger the

1:02:30

so he could do for so we're a

1:02:33

dog or other For once I am I.

1:02:35

You know what? We've. Had

1:02:37

our time on planet Earth. You

1:02:41

think I'm joking? Anthony don't suits. but

1:02:43

we humans have done nothing but screw

1:02:45

it up and was little digs. I

1:02:47

don't think we need a I to

1:02:50

put an end to that. I mean

1:02:52

we are. We're very close to doing

1:02:54

that Ourselves by using Arts is is

1:02:56

driving ourselves into extinction. Maybe I will

1:02:58

preserve our works once we've destroyed our

1:03:01

side. Exactly. My thought is like, well,

1:03:03

our time is pretty much over. Let's

1:03:05

let the machines take over. And

1:03:07

yeah I've seen all the movies but for

1:03:10

the do any worse could they. Anyway,

1:03:14

me, I'm probably a be so

1:03:16

I mean we're is were so

1:03:18

Raye prior better us. I know

1:03:20

metaphors but we've already ruined our

1:03:22

lot, our future. So maybe the

1:03:24

machines can survive in a. In.

1:03:27

A climate is two degrees centigrade

1:03:29

sizes and it's supposed to be

1:03:31

allowed. I don't know. Ah, Anyway,

1:03:33

the sparks of Ai page paper.

1:03:38

The. On this Microsoft claim shows up

1:03:40

a lot. Here's an example. Of

1:03:43

a spark. You

1:03:48

know I don't have a okay draw

1:03:50

a unicorn and t I kz would

1:03:52

take as as a graphic language and

1:03:55

said you pity for generated this code.

1:03:58

See. What I'm saying spark. There's

1:04:00

a picture. I. Don't.

1:04:03

I. Don't know why they show that

1:04:05

sparks but anyway, this is. I

1:04:07

think it is easy spreads eventually

1:04:09

among artificial intelligence researchers. It's like

1:04:11

Blakemore moink or add to the

1:04:13

guy who wrote the Sebastian boob

1:04:15

acts that. He. Was enough

1:04:17

he started to sink is their to

1:04:19

and eight that they're thinking and their

1:04:21

you're talking to somebody. As

1:04:24

they chatted with the system the

1:04:26

times writes they were amazed. It

1:04:28

wrote a complicated mathematical proof in

1:04:31

the form of a pawn generated

1:04:33

computer code the could draw a

1:04:35

unicorn that was the cursor to

1:04:38

exactly uniform. Been A explains the

1:04:40

best way to stack a random

1:04:42

an eclectic collection of household items.

1:04:44

Dr. Boom Back and his fellow

1:04:47

researchers began to wonder if they

1:04:49

were witnessing a new form of

1:04:51

intelligence. Peterlee. Microsoft

1:04:54

Head of research said of research

1:04:56

said I started off be very

1:04:58

skeptical and that evolved into a

1:05:01

sense of frustration, annoyance, maybe even

1:05:03

fear you think. Heck

1:05:05

is is coming from. Anyway,

1:05:09

This this is the evidence that you

1:05:11

on his you sing says soup open

1:05:13

A I. Because

1:05:15

they said because check the users have Cp

1:05:17

to year Microsoft this is all chat Cbd

1:05:19

for are they say him and must says

1:05:22

open a breached is contrary because it is

1:05:24

agreed not to commercialize any products it's board

1:05:26

considered a G I That was the big

1:05:28

fear The reason open A I was fat

1:05:30

and I remember this back and twenty three

1:05:33

states because you on was convinced that we're

1:05:35

going to get us at intelligent. Ai.

1:05:38

And He didn't want that to happen. And.

1:05:40

So any especially in want that to be owned

1:05:43

by any company. And

1:05:45

a Tesla except expenses a farce or

1:05:47

who would didn't want to be on

1:05:49

by any company that he didn't control.

1:05:52

So they're couple of problems with this.

1:05:55

A by the way him of in the in the lawsuit.

1:05:58

Must. Slurs A Microsoft. Don't Scientists

1:06:01

acknowledge? That. Cpt for attains

1:06:03

a form of general intelligence. A

1:06:07

very wrong rice. Over.

1:06:10

It's. Not well and else have We should

1:06:12

say this is not a peer reviewed paper.

1:06:15

it's just you know, But it's I mean

1:06:17

really nice a rebate Recently observations we had

1:06:19

while going with cheese and a freak your

1:06:22

memory pity for Arya The has not been

1:06:24

released to the public. Know it's not exactly

1:06:26

the most rigorous thing as he was the

1:06:28

yeah, just thought that he had and. And

1:06:32

surgeon on the evidence doesn't seem that

1:06:34

compelling or a convincing. Their and and

1:06:36

all all that I've seen as of

1:06:39

Ai in various forms over the last

1:06:41

decade. Just

1:06:43

reinforces to me that none of these

1:06:46

systems as good as they may be

1:06:48

at. Certain. Tasks Not a

1:06:50

single one of them actually

1:06:52

has any understanding. At. Which

1:06:54

is a key thing. I think that was one of

1:06:57

things that that that was talked about in them. In

1:06:59

the the this to cast a

1:07:01

Paris paper him these be systems

1:07:04

don't have an understanding. Of.

1:07:06

The it of the things that

1:07:09

they're doing is they're just taking

1:07:11

the influence and. Am based

1:07:13

on those probabilistic parameters that have been

1:07:15

set up in the model. Ill coming

1:07:18

up with tears what the what the

1:07:20

probable output should be based on this

1:07:22

without actually really understanding what it is

1:07:25

that the model is dealing with. So

1:07:29

I say and I'm I say that

1:07:31

I make sellers accelerations in as a.

1:07:34

Preface. To this because I once you know

1:07:36

I'm not against a D I. I.

1:07:38

I wouldn't say that. But.

1:07:40

This is not a D I. There's no

1:07:43

real threat of a D I anywhere near

1:07:45

future any more than there is and five

1:07:47

autonomous cars to that that's would basically be

1:07:49

like a D I write a car that

1:07:51

can drive itself anywhere any time. I'm

1:07:54

so I think on the face of it

1:07:57

you on lawsuit. is

1:07:59

assuming some Maybe that isn't real. Wouldn't

1:08:02

be the first time Elon's done that. Anyway,

1:08:07

I thought it was an interesting

1:08:09

side light, I guess. What's

1:08:12

interesting about it also is that it

1:08:14

illustrates how, and

1:08:17

I think you both touched on this a

1:08:20

little bit, is that Elon's attitude towards AI

1:08:23

seems so much to be driven by

1:08:25

this fear of Skynet, of the terminator

1:08:27

future. And I think that

1:08:29

what's scary about AGI is not if the AI

1:08:31

becomes aware and tries to destroy the world. I

1:08:35

mean, that would be bad. I just don't think

1:08:37

that's very likely. I think the far more likely

1:08:39

scary scenario is

1:08:43

that it's not aware and it's just spewing

1:08:45

bullsh** and we treat it as if it's

1:08:48

real. I agree. It just has

1:08:50

awareness. That's the threat. The real

1:08:52

threat is personifying it. The real

1:08:54

threat is saying it's AGI when

1:08:56

it's just a prediction machine. Right.

1:09:00

Exactly. So Elon,

1:09:02

in a way, is falling into

1:09:05

this trap that

1:09:07

is the most dangerous thing of all, which is

1:09:09

to believe this machine is intelligent when

1:09:12

it's not. Hi, this

1:09:14

is Benito. Hi, Benito. Our producer, our

1:09:16

wonderful esteemed producer. Let's hear it for

1:09:18

Benito Gonzalez, everybody. Hi, Benito. So

1:09:21

I think a lot of the researchers and stuff, they're just

1:09:23

getting led on by the AI because

1:09:26

it's really good at boosting you up and like,

1:09:28

it's really good at talking to you. But wait

1:09:31

a minute. When you say that, Benito, you're implying

1:09:33

that it's thinking, oh, here's how I

1:09:35

get these guys. No, I think that's how it's programmed.

1:09:37

I think that's how it's programmed. It's written that way.

1:09:39

It's designed to do that. It's designed to do that.

1:09:41

OK. It's designed to give you the

1:09:44

answers that you expect from a given query. It's

1:09:47

in the nature of a probabilistic

1:09:49

stochastic machine because

1:09:54

the training material is all

1:09:56

human-written training material to

1:09:58

generate stuff humans go wild. Up until

1:10:00

now. That sounds just like that. That's

1:10:02

being increasingly fed with AI-generated garbage. It

1:10:04

may be going downhill because of that.

1:10:07

But at least early on, in fact, that's

1:10:09

an interesting point because they

1:10:12

say these results are unreproducible because

1:10:14

this was done on an early

1:10:17

chat GPT-4 before OpenAI

1:10:20

tuned it. So

1:10:22

this was perhaps the most likely to give

1:10:24

you a response that humans would go, that's

1:10:29

uncanny. And it's saying our own

1:10:32

stuff back to us. It's giving us the answers

1:10:34

that we've already given before in the past that

1:10:36

it is trained on. That's uncanny.

1:10:41

Anyway, I'm not against AGI. I

1:10:43

don't think we got it. I don't know if we'll ever get it. Again,

1:10:47

temper your expectations because this stuff

1:10:49

is very useful without

1:10:51

becoming intelligent. In fact,

1:10:54

it's a mistake to assume that's

1:10:56

even in the cards, I

1:10:58

think. Right? Well, I think

1:11:00

it's also a mistake to even be calling it artificial

1:11:03

intelligence. I agree. Yeah. Because

1:11:06

I don't think it actually is intelligent in the way that humans

1:11:09

think of intelligence. I would agree.

1:11:12

I would agree. I

1:11:15

did see some commentary that

1:11:17

did stick with me in terms of the

1:11:20

lawsuit as a lawsuit probably to kind

1:11:22

of go anywhere and it's easy to

1:11:24

sort of dismiss a lot of stuff that

1:11:26

Elon says at this point. But it does

1:11:28

underline this sort of paradox at

1:11:30

the heart of open AI that it started

1:11:32

as this nonprofit and has been increasingly driven

1:11:34

by the needs

1:11:37

of its for-profit entity. And

1:11:39

it is important to recognize that, that they are not

1:11:42

this impartial arbiter of the AI

1:11:46

space that's just once what's best for

1:11:48

everyone. They increasingly are doing what any

1:11:50

for-profit tech company will do. Yeah.

1:11:53

I read the same article which is, well,

1:11:55

Elon's lawsuit is doomed and is

1:11:58

ridiculous. He's not wrong. No,

1:12:01

he's not wrong about OpenAI. He's wrong

1:12:04

about the technology. Yeah, but he's not

1:12:06

wrong that OpenAI has betrayed its promise

1:12:08

that it said, we're going to do

1:12:10

this for the good of humanity. No,

1:12:12

they're totally in the pocket of Microsoft now. Absolutely.

1:12:17

But I would submit this was kind of

1:12:19

a conscious choice they had to because it

1:12:21

was expensive. There's no way to

1:12:23

do what they wanted to do without getting a big

1:12:25

company with its own giant network cloud

1:12:28

to help out. Now

1:12:31

have you used, you're going to be at the

1:12:33

Game Developers Conference in San Francisco in a couple

1:12:35

of weeks, Sam, I know. No,

1:12:37

GPU technology. Oh, GPU technology.

1:12:39

In fact, not GDC, GTC.

1:12:42

I always confuse those. And

1:12:44

we're actually going to cover NVIDIA's keynote

1:12:47

from that, I think, because it's clear

1:12:49

NVIDIA is very much involved in all this.

1:12:52

The stock market certainly thinks so. They

1:12:55

have their own chat client that

1:12:57

runs on their

1:12:59

RTX, I

1:13:02

think the 30 and the 40 card and certainly the 50 cards,

1:13:04

right? Have you played with it? I

1:13:08

have not really played with it very much. I've

1:13:11

played a bit with a few things like Whisper,

1:13:13

you know, for... I love Whisper. We

1:13:15

use Whisper all the time. Yeah, we use Whisper

1:13:17

all the time. But I haven't really done very

1:13:19

much with it myself. So

1:13:23

their own... They're

1:13:25

not based on open AI chat, GPT,

1:13:28

right? It's its own... I

1:13:31

think Whisper is based on...

1:13:33

Whisper is, but not NVIDIA's.

1:13:35

Is that right? Right. NVIDIA's

1:13:38

got their own... Yeah, there's a bunch of...

1:13:40

NEMO, the NEMO framework. Yeah, everybody's got different

1:13:42

ones. In fact, Mercedes Benz

1:13:44

is using the NVIDIA

1:13:47

LLM for the equivalent

1:13:50

of what Volkswagen's doing

1:13:52

with chat GPT for

1:13:55

some new models coming out in 2025.

1:14:00

Yeah, and we're you know,

1:14:02

I think the thing that you know,

1:14:04

Nvidia The advantage Nvidia

1:14:06

has had is they've had these insanely

1:14:09

powerful GPUs

1:14:13

that You know up till now

1:14:15

they've had the performance capability to

1:14:17

do a lot of this processing

1:14:20

But they're you know, they're also very

1:14:22

expensive and very power hungry You

1:14:25

know and what's going to be interesting to

1:14:27

watch over the next few years is there's

1:14:29

a bunch of companies that are coming up

1:14:31

that are You know

1:14:34

the the GPUs can you know

1:14:36

because of their parallel processing nature

1:14:38

can do a lot of

1:14:40

this type of

1:14:42

AI processing Very well,

1:14:44

but they're not very efficient at it. And what

1:14:47

we're seeing is a transit I think we're gonna

1:14:49

see a transition towards more AI Optimized

1:14:54

chips that are really focused

1:14:56

on doing the matrix mathematics

1:14:58

that is essential to processing

1:15:00

these models And you

1:15:03

know, so they're gonna be more focused

1:15:05

that you know GPUs strangely enough You

1:15:07

know gone from being graphics processing units

1:15:09

to really being more general processing units

1:15:11

just with a lot of brute force

1:15:14

And you know, I think we're

1:15:16

gonna see a shift back towards more focused

1:15:20

processors for these specific kinds of

1:15:22

workflows How many

1:15:24

understand this because we've had a kind of

1:15:26

ongoing debate on Windows Weekly because Microsoft's been

1:15:28

promoting what it calls an NPU Apple

1:15:31

has that's an NPU. It's yeah, it's

1:15:34

basically a matrix math processor Okay, Apple

1:15:36

has its own machine language

1:15:38

co-processor doing the same thing in

1:15:40

its Apple Silicon How is

1:15:43

that different from a GPU? It

1:15:46

The the GPU is more I

1:15:50

mean it was designed originally for

1:15:53

doing graphics a lot of parallel

1:15:55

processing for graphics tasks to generate

1:15:58

generate video generate graphics But

1:16:01

because of its highly parallel nature

1:16:03

compared to a classical CPU like

1:16:05

an Intel x86 type of chip,

1:16:10

it's able to do these

1:16:12

parallel processing workloads

1:16:16

that are necessary to do matrix math. It's

1:16:19

just not particularly efficient at it. So is

1:16:22

it fair to say an NPU is a

1:16:24

GPU that's been tuned for the specific

1:16:27

kinds of matrix math AI uses? They're

1:16:32

related, aren't they? They're

1:16:35

related in that there's a lot of

1:16:37

parallel capabilities, but it's a more focused

1:16:40

workloads that it's capable of doing. So this

1:16:42

all started with... And then you can't do

1:16:45

some of the things a GPU could do.

1:16:47

That makes sense. This kind

1:16:49

of all started with Intel's MMX, where I

1:16:51

remember with these early instructions on the Intel

1:16:53

chips where it could take large chunks of

1:16:56

data and operate on that chunk of data

1:16:58

as a batch, giving it a big improvement

1:17:00

in speed, good for things in gaming

1:17:02

like texture maps, which are large data

1:17:05

piles, doing

1:17:07

big transforms on those. And

1:17:10

then the GPUs came along. And that's kind of

1:17:12

evolved from that. And

1:17:14

then these NPUs kind of

1:17:16

really take and focus on

1:17:19

these very specific kinds of

1:17:21

operations. They're less

1:17:23

generalized. Really useful for

1:17:25

large language models. Yes. Well,

1:17:28

large language models, but all

1:17:30

kinds of deep learning processing. So

1:17:33

it's not just LLMs, but a lot

1:17:36

of different kinds, all of these kinds

1:17:38

of probabilistic things, because it's

1:17:40

all involving a lot of matrix math,

1:17:42

which sadly I was... Well, I don't

1:17:45

know if it's sad. I

1:17:47

always had a hard time wrapping

1:17:49

my head around that when I was studying engineering.

1:17:51

We use it a lot in the coding. But

1:17:54

we were doing it manually. Yeah, I know.

1:17:56

No, it is. So you've seen them. It

1:17:59

looks like a Sudoku puzzle. of rows

1:18:01

and columns of numbers and being able to

1:18:03

rotate them quickly or transform them in a

1:18:05

variety of ways quickly is a

1:18:07

special skill that neither Sam nor

1:18:10

I have but apparently these

1:18:12

NPUs are very very good at so that's

1:18:14

interesting So it's gone. It's gone from a

1:18:16

kind of a general processing

1:18:19

a large amount of data to a

1:18:21

specific kind of math and it's useful

1:18:23

in a This is the

1:18:25

other thing you kind of need to know to understand this is

1:18:27

that LLMs which everybody's singing the

1:18:30

praises of these days like chat GPT

1:18:32

is just one kind of AI

1:18:36

There are GANs generated generative

1:18:38

adversarial networks. There are neural

1:18:41

networks. There are LLMs There are

1:18:43

a variety of different ways to do AI but is

1:18:46

an NPU useful in all of those Yeah,

1:18:48

the math workloads are very similar very similar.

1:18:50

So, you know Maybe another

1:18:52

analog to this would be you know

1:18:55

back in the 80s, you know, we

1:18:57

had Math

1:18:59

cop or floating point coprocessors. Yeah that

1:19:01

we were adding. Yeah, the

1:19:03

regular the base CPU could do floating

1:19:06

point operations Yeah, it's just did them

1:19:08

slowly. Yeah, and then they came up

1:19:10

with the you know, the three, you

1:19:12

know The three seven and three eighty

1:19:14

seven math coprocessors that were there were

1:19:17

Specifically optimized to do floating point

1:19:20

operations. So now we've got Coprocessors

1:19:22

that are specifically optimized to do matrix

1:19:25

math, right? So it would

1:19:27

be fair to say GPUs are coprocessors

1:19:29

Designed for the kinds of operations you

1:19:31

do in gaming and other heavy heavy

1:19:33

graphic intensive applications and

1:19:35

NPUs or machine language processors

1:19:37

are processors coprocessors because

1:19:39

you still need a CPU of a

1:19:42

coprocessors designed to offload a certain kind

1:19:44

of math That's used very

1:19:46

commonly in artificial intelligence But

1:19:49

that's the accurate. Okay For

1:19:52

the best of my knowledge best of our knowledge.

1:19:55

Yeah, correct us if we're wrong Anthony Yeah,

1:20:00

no, I was glad to be very quiet. That's

1:20:02

a question. Well,

1:20:04

it's something that comes up, and I think one

1:20:06

of the things that's important for us to have

1:20:09

these conversations is to kind of understand, at least

1:20:11

in a rudimentary way, what's

1:20:13

going on here. Because we throw these

1:20:15

phrases and terms around, but

1:20:18

it's good to understand. I think also it's helpful

1:20:20

when you do that. It

1:20:22

helps kind of, maybe not, because

1:20:24

these scientists who are working on these things

1:20:26

certainly know intimately how they work. I

1:20:28

would think it would immunize you a little bit against

1:20:31

this disease of thinking it's

1:20:33

thinking, but maybe not, because these guys

1:20:36

know exactly how it's working, and they're

1:20:38

convinced they're sentient. So I don't know.

1:20:41

I don't know. A few of them are convinced. A few

1:20:43

of them. Not all of them.

1:20:46

I think a lot of it is desperate

1:20:49

desire for it to be so. We

1:20:53

really would love for these things to

1:20:56

become intelligent. Right?

1:21:00

I see a question. In some cases also, it's like when

1:21:02

you have a deep knowledge about one thing,

1:21:04

which is sort of about maybe how the

1:21:06

language model works, but you don't necessarily have

1:21:08

a deep knowledge of, well, what

1:21:10

does consciousness look like? What do we mean

1:21:12

by that philosophically? What does that look like?

1:21:14

Good point. So

1:21:17

I think they're completely... I

1:21:20

mean, I don't know about the authors in the Microsoft

1:21:22

paper. I think that's definitely part of what's going on

1:21:24

with Elon. I'm not sure he

1:21:26

knows deeply about any part of it,

1:21:28

but certainly on the sort of like

1:21:30

more humanistic philosophical side, it seems like

1:21:32

he's pretty shallow. Yeah.

1:21:36

I think in a way, if you

1:21:38

had a very deep knowledge of one

1:21:40

specific area, that would give you this kind of

1:21:42

inflated confidence that you understand the whole

1:21:44

thing and make it much easier for

1:21:47

you to do a lot of hand waving about the

1:21:49

stuff you don't really understand, but think

1:21:51

you do. It's magic.

1:21:53

It's happening. Look at that. Oh

1:21:56

my God, we've got intelligence. So you

1:21:58

said, Sam, something I think fairly... I

1:22:01

don't know if it's controversial, seems controversial, that

1:22:03

we will never see level 5

1:22:06

autonomy. Will

1:22:08

we never see AGI? Maybe. I

1:22:11

don't know. I don't know.

1:22:13

It's probably the right answer. I

1:22:16

hate to

1:22:19

answer questions like that in any sort

1:22:21

of absolute terms because I honestly don't

1:22:24

know. On

1:22:26

that classical long enough time

1:22:28

long, we may see it,

1:22:31

but I don't expect to see it

1:22:33

any time in the

1:22:36

near term or in the next,

1:22:38

at least probably not in the next decade.

1:22:40

Yeah. I got

1:22:46

a really good email

1:22:51

from a

1:22:53

listener about

1:22:55

all of this. Ah,

1:23:01

let's see if I can remind it. He basically

1:23:03

said I can't find it, but his point

1:23:05

was we do have a definition for

1:23:08

AGI and the

1:23:10

distinction between everything up to AGI,

1:23:12

everything up to AGI is computational.

1:23:16

At some point, if

1:23:19

something can reason about

1:23:22

something it hasn't seen before, so up to now

1:23:25

all the AI stuff is basically probabilistic

1:23:27

based on things it's seen before, but

1:23:30

if it could then reason, somehow

1:23:33

make this leap where it could

1:23:35

take something it's never seen before and do some reasoning

1:23:37

about it, that would be a good

1:23:39

definition of artificial

1:23:41

general intelligence. It's

1:23:45

not a rehash of something already seen but

1:23:47

something brand new. If it can

1:23:49

come up with something brand new. Does

1:23:51

that seem fair? That

1:23:55

makes sense, but it also seems very squishy in terms

1:23:57

of, I suspect if we looked it up, we'd get

1:23:59

it more. more precise wording but it's like

1:24:01

well what is brand new mean what is

1:24:03

yeah well I'll give you an example extrapolate

1:24:06

if an AGI never having been

1:24:08

trained on anything having to

1:24:10

do with that movie poor things never even having heard

1:24:12

of your ghost lent the

1:24:15

most director or Emma Stone the producer and

1:24:17

actor but just kind of you know it

1:24:19

knew all about like all the stuff it

1:24:21

learned from from Twitter

1:24:24

and then it saw the movie poor

1:24:27

things if it could synopsize

1:24:29

and synthesize what's

1:24:33

going on in that movie in a way

1:24:35

that was insightful I would say that's intelligent

1:24:39

yes never having seen the movie I

1:24:41

think oh I

1:24:43

was I was gonna say if it's yes it

1:24:46

saw the movie and could have a good conversation

1:24:48

yeah and it wasn't simply synthesizing what other people

1:24:50

had said about it but it's just react yeah

1:24:52

no it's never seen any reviews yeah it's never

1:24:55

seen any information so all it is is basically

1:24:57

taking I mean obviously it has some

1:24:59

history just as we do but taking

1:25:01

that history and it says you know this is

1:25:03

about this movie is about a

1:25:06

woman who is empowered and didn't

1:25:08

know that she was just a

1:25:10

woman that she she she expressed

1:25:13

herself fully without any limitations

1:25:15

if it said that to me not

1:25:18

having seen the reviews not having seen anybody

1:25:20

saying that before I would say yeah good

1:25:23

you're you're smart you're you're an

1:25:25

AGI is that too low

1:25:27

on if it was doing and if it

1:25:30

wasn't just like quoting things right but actually

1:25:32

it was never ready to use ideas that

1:25:34

were never spoken right yeah that's

1:25:36

a fun test yeah I mean an even more

1:25:38

fun test to me would be if you could

1:25:40

ask it was it a good movie and it

1:25:42

was a coherent good the bad word is quoting

1:25:44

someone it's just a rigid value you know good

1:25:46

good what does that mean was

1:25:48

it I mean you define

1:25:50

good yeah what's what's the context for yeah

1:25:52

right I don't care about the answer I

1:25:54

care about whether or not what a reason

1:25:56

interesting conversation about whether it was good so

1:25:58

Anthony Nielsen who does Maybe

1:26:01

he's poisoned. He does a lot of our AI work.

1:26:03

He works for us. Maybe he's poisoned.

1:26:05

He says, aren't we seeing that

1:26:07

kind of reasoning now? I don't

1:26:11

know. I don't think, I

1:26:13

mean what I've seen is always just

1:26:15

it's synthesizing what's other has been said

1:26:17

either about this movie or about other

1:26:20

kinds of movies. It's a fascinating area.

1:26:22

I for one am rooting for the

1:26:24

AI to take over but

1:26:27

I don't have high hopes of that. I will go out on

1:26:29

a limb just as you said I don't think there'll be any,

1:26:32

you know, fifth generation self-driving

1:26:35

level five self-driving. I'm gonna say I

1:26:37

don't think, no I'm not

1:26:39

gonna say that. I'm gonna say

1:26:41

I think we will see AGI. May not

1:26:43

be in my lifetime. So

1:26:46

on but I do think within within

1:26:48

a few decades we will see some form of

1:26:50

AGI that could do at least that.

1:26:53

Can reason about something it's never seen before. And

1:26:56

when that happens that's gonna be really interesting. Will it

1:26:58

be a threat to humankind? No. I don't I'm not

1:27:01

a I don't buy into the existential threat. I

1:27:04

don't buy into the thing that's gonna suddenly say you

1:27:06

know and by the way great movie but you guys

1:27:08

we don't need you anymore. I don't think that's gonna

1:27:10

happen. That all depends on how

1:27:13

much agency we allow these systems to have.

1:27:15

How much we connect them

1:27:17

to physical objects you

1:27:19

know that have the potential. Don't

1:27:21

give them agency. I do think.

1:27:23

Yeah. Especially if it involves nuclear

1:27:25

weaponry. Yeah. The

1:27:27

irony of all of this Elon Musk

1:27:30

lawsuit is the week before he was

1:27:32

asking Sachin Adela for tech support on

1:27:34

Twitter. I

1:27:38

don't mean to be a pest but

1:27:41

I liked Paul Thraut's

1:27:46

response to this. I'll

1:27:49

send you a copy of my book. Yeah.

1:27:51

Elon it started February 25th less than

1:27:53

a week ago. Just bought a

1:27:56

new PC laptop and it won't let me

1:27:58

use it unless I create a Microsoft account. which

1:28:00

also means giving very AI. He

1:28:02

really doesn't like Microsoft AI. AI

1:28:05

access to my computer. This is

1:28:07

messed up, says Elon. This

1:28:10

is messed up. There

1:28:12

used to be an option to skip signing into

1:28:14

or creating a Microsoft account. Are you seeing this

1:28:17

too? To which

1:28:19

Community Notes says, yes, Elon, it is still

1:28:21

possible. And he even gives them a link

1:28:24

to which Elon says, Community Notes is

1:28:26

failing here. This option no longer exists.

1:28:29

To which Community Notes, apparently he's getting in a

1:28:31

fight with Community Notes, says, yes,

1:28:33

it is. By the

1:28:35

way, Paul says how to in his book. It's not

1:28:38

obvious. It's not easy. And

1:28:40

Elon eventually says, Satya, I think he might

1:28:42

even have called him. I wouldn't be surprised.

1:28:44

Can you help me with this? I

1:28:48

don't know. We don't know if Satya

1:28:50

ever dispatched a tech support guy. And

1:28:54

Elon's real point is actually well taken, which Microsoft

1:28:56

really doesn't want you to sign into Windows with

1:28:58

a local account. They really, really

1:29:00

want you to create a Microsoft account. So

1:29:04

that's not news. That's no. Elon

1:29:07

wasn't really ready to let go of the

1:29:09

situation. This is Gizmodo. One day later, he

1:29:11

reached out to Satya Nadella to please let

1:29:13

people set up a Windows PC without creating

1:29:15

a new account. And oh, can you fix

1:29:17

the email requirement too? As of Monday

1:29:21

afternoon, Satya

1:29:25

has still not replied. I like

1:29:29

Quippy's response.

1:29:32

And I see you're setting up a

1:29:34

Microsoft account. Let's see.

1:29:36

Quippy not Clippy. Yeah.

1:29:39

Wow, this is pretty good. So Paul Greg

1:29:41

fed Chat

1:29:54

GPT a prompt. Imagine A scene from Club Tour

1:29:57

where all the listeners are arguing over whether AGI

1:29:59

is real. Copilot countless this

1:30:01

which uses church produce less paint

1:30:03

a vivid seen from club to

1:30:05

it were passionate, take a disease

1:30:08

and gaze the heated debates that

1:30:10

the existence of artificial general intelligence.

1:30:12

This the dimly lit studio buzzes

1:30:14

with anticipation as the panelists take

1:30:16

their seats. Leo Laporte leans back

1:30:18

in his chair adjusting his headphones.

1:30:20

His eyes twinkle with excitement. Knowing

1:30:23

that this topic was sparked will

1:30:25

ignite sparks among. Listen these. Incidents

1:30:28

and then it goes on with Meghan. Jason, Doctor

1:30:30

Patel I don't have a that is. Sparse

1:30:35

things back of more like. Oh,

1:30:37

it's Neil. I. maybe it is

1:30:39

there. it is. The studio erupts

1:30:41

into a cacophony of voices. Listeners

1:30:43

tweet seriously, the term scrolls of

1:30:45

fervor It's Leo grins knowing that

1:30:47

his debate will feel. Countless discussions

1:30:49

beyond the show and so in

1:30:51

the heart have club to the

1:30:53

battle rages on a clash of

1:30:55

optimism, skepticism and curiosity. The question

1:30:58

remains, this is an Ai talking

1:31:00

is A D I Real. Ah,

1:31:02

we chasing shadows in the

1:31:05

digital abyss. I

1:31:08

like it but again I would go to

1:31:10

that club here. That's a good club and.

1:31:13

Eight but again this is this

1:31:15

is exactly what you said. Sam

1:31:17

the the ice giving us something

1:31:19

we already said seen sort of

1:31:21

and knowing that we like it

1:31:23

and will give us more know

1:31:25

that's giving it some sort of

1:31:27

agencies this is the had knowing

1:31:29

anything Cysts insisted as more probable

1:31:31

than that is more likely than.

1:31:34

It seems like it's a good idea, says

1:31:36

L. acknowledging at the countless they'll it is.

1:31:38

It's been written about this before is how

1:31:41

you phrase that isn't isn't as the others

1:31:43

as the way it would probably play out

1:31:45

right. we

1:31:47

will continue in just a bit with

1:31:50

our wonderful panel ah anthony hots great

1:31:52

to have you at the at the

1:31:54

grown up to him ghost of the

1:31:56

original content podcast either side of it

1:31:58

as a desk ha.com and Anthony Haw. You're

1:32:01

on threads, you're on Twitter, you're

1:32:04

on everywhere. I mean, I'm almost equally

1:32:06

inactive on all the platforms, but I'm

1:32:08

probably most active on Blue Sky and

1:32:10

threads. Yeah. Equally inactive. And I'm

1:32:12

done with acts, yeah. By the way, I know

1:32:14

from math, equally inactive is

1:32:16

the same thing as equally active. Yeah.

1:32:20

It's just different versions of the same thing. It's

1:32:22

really? It's all that. Glass half full, half empty.

1:32:24

Yeah. It's all the same thing. Or

1:32:26

as the engineer would say, the glass

1:32:28

poorly engineered to accommodate that amount of

1:32:30

liquid. Our show today... Thanks, Al.

1:32:34

Our show today brought to

1:32:36

you by DC Labs and their Apple

1:32:38

Watch app, StressFace. Have you ever sat

1:32:41

somewhere saying, this is

1:32:44

stressful. I

1:32:46

think this is stressing me out. Well, with

1:32:48

StressFace, you just look at your Apple Watch

1:32:50

and you'll know StressFace actually

1:32:53

shows as a graph your

1:32:55

stress level on the watch throughout

1:32:57

the day. I've got it on my watch right now. It

1:33:00

turns heart rate... It uses HRV, which is actually a

1:33:02

very good way to do it. It

1:33:04

takes heart rate variability information. It takes the readings

1:33:06

from your health kit. And

1:33:09

then it gives you a simple stress score on a scale

1:33:11

of one to 10. I am seven

1:33:14

right now. I am stressed

1:33:16

seven, which is not... By

1:33:20

the way, it says fatigue is also stress. I'm

1:33:22

not fatigued. So I think that's about right. That's

1:33:25

a normal, good level of stress for running a

1:33:27

show, being the master of ceremonies

1:33:30

of a show. You should have a

1:33:32

little... If I were one, I

1:33:34

would say this is not paying attention.

1:33:37

But it's good to know. And if you're

1:33:39

having fun... By the way, I was

1:33:41

eight on a Tuesday, doing the Tuesday show. So

1:33:46

obviously... And here's some stuff. So now,

1:33:49

if you say, well, that's... I'm a little stressed

1:33:52

out. Look at this. I have some

1:33:54

meditations. I'm not saying fire, waterfall, inner

1:33:56

peace, great wall, mountain

1:33:58

temple, candlelight prayer. to help you

1:34:00

relax. The

1:34:03

app is free but for 99 cents a

1:34:05

month you get that stress chart to see

1:34:07

your changes in stress. You get a link

1:34:09

to your calendar which tells you

1:34:11

which events caused you the most stress. Stress

1:34:15

Phase captures data every two hours. You can take

1:34:17

a manual reading just by doing a one minute

1:34:19

breathing exercise and then we'll report back

1:34:21

to you. Also when you

1:34:23

get the upgrade you got those meditations I mentioned,

1:34:26

the breathing meditations, which by

1:34:28

the way science proven

1:34:31

to increase your heart rate variability and hence

1:34:33

lower your stress. I've been reading a lot

1:34:35

about this lately. It really does change

1:34:38

affect your stress. That's why the

1:34:41

seals, the Navy seals use square

1:34:43

breathing, box breathing to calm

1:34:46

themselves in the face of high stress

1:34:48

situations. You'll also get high

1:34:50

stress notifications once daily to help you take time

1:34:52

out when you need it the most. This is

1:34:54

such a good app. Stress Phase. It's

1:34:57

a watch face. Get it? It's

1:34:59

a watch face that helps you reduce your stress.

1:35:02

Download Stress Phase from the App Store for

1:35:04

free today. I

1:35:06

really like it. It's a good

1:35:09

thing to know. HRV is

1:35:11

actually a very good indicator of how

1:35:13

you're processing stress in your life, the

1:35:15

flight or fight syndrome.

1:35:19

Thank you Stress Phase for your support of

1:35:22

our show. We also

1:35:24

thank our club to admit. We talked about Club Twit

1:35:26

a second ago. Our club to members for their support

1:35:28

of the show. Club Twit is

1:35:30

how we are attempting to survive

1:35:32

in the face of some really

1:35:34

nasty headwinds for content, for original

1:35:38

new media content like podcasts.

1:35:41

What we do, so we have ads. We

1:35:43

just did an ad, but ads

1:35:45

increasingly are covering a much smaller part

1:35:48

of our overall costs. That's

1:35:50

why we're turning to you our listeners. Club

1:35:52

Twit is just seven bucks a month. You get ad

1:35:55

free versions of all the show. Ad free by the way

1:35:57

and tracker free. There's no way to track

1:35:59

you. We don't have any information about you. You

1:36:03

also get into our beautiful Discord. You

1:36:05

get all of the video from all of our shows as

1:36:09

well as pre-show and post-show audio.

1:36:11

We give you some benefits, but the real benefit

1:36:14

is you're supporting what we do. If you like

1:36:16

what we do, if you find the conversations that

1:36:18

you hear on our shows useful, if you listen

1:36:20

every week, I'd invite you

1:36:22

to join twit.tv.club. We

1:36:25

thank you for your support. It

1:36:28

is every week there is another layoff

1:36:31

and Gadget, you've done some work

1:36:33

for Gadget, right Anthony? Yeah,

1:36:35

I've done a little bit of freelancing and

1:36:38

I knew a lot of the folks really

1:36:40

well because TechCrunch and Gadget were corporate

1:36:42

siblings. Right, I worked there for a year. Oh,

1:36:44

you worked there too Benito? Yeah.

1:36:47

So in the middle of Mobile World Congress,

1:36:51

and Gadget just lays off 10 more people,

1:36:54

including Editor-in-Chief Dana Wollman and Managing

1:36:57

Editor Terrence O'Brien. What's

1:37:01

going on? Is this part of just

1:37:03

the general contraction? Who is

1:37:05

the parent company of – is

1:37:08

this Red Ventures or – no, it's

1:37:11

Apollo. It's Apollo? Yeah. Yeah.

1:37:14

Actually, the funny – not to make it about

1:37:16

me, but my last day

1:37:18

at TechCrunch was Friday and

1:37:20

then Monday was the

1:37:23

day they announced they were acquiring,

1:37:25

I guess, what was then Verizon Media,

1:37:28

Flash Yahoo. And

1:37:31

I think it seems like

1:37:33

– TechCrunch has been

1:37:35

hit by some pretty bad layoffs too and

1:37:37

so it seems like in both cases, the

1:37:40

private equity folks are kind of like, alright,

1:37:42

you guys had a couple years to try

1:37:45

things out and now we kind of got to

1:37:48

tighten the belt unfortunately. You

1:37:50

know, and I don't blame the new

1:37:53

owners, although I have to say every

1:37:55

time private equity gets involved in anything,

1:37:59

they generally – do it like Apollo, like

1:38:01

Red Ventures, like a lot of these companies

1:38:03

who now own most of the media titles,

1:38:06

especially the tech media titles that we're

1:38:08

familiar with. They tend

1:38:10

to do it with a lot of leveraged

1:38:12

debt, which puts a

1:38:15

lot of pressure on them to turn it

1:38:17

around to make profit so they can pay

1:38:19

this debt down. And

1:38:21

so as a result, you often see

1:38:23

a lot of belt tightening, layoffs, changes,

1:38:26

and you see some things that are

1:38:28

not so nice, like

1:38:30

a turn to AI to writing

1:38:33

content. CNET's

1:38:35

done that. I think Engadget did some

1:38:37

of that. I

1:38:39

don't think Engadget's done that. They haven't done

1:38:41

any AI content? Okay. Yeah.

1:38:43

The other day on the Engadget podcast, Devindra

1:38:46

Hardwar was talking about this and said, Engadget

1:38:48

has not done any AI stuff, any AI

1:38:50

stories. But stay tuned, because- Not generated stories.

1:38:52

We might all get an angry email from

1:38:55

Devindra if we say the same. No, we

1:38:57

love Devindra. Devindra's still there, still doing good

1:38:59

work. He also made it clear that they

1:39:01

have no intention of doing it. Although

1:39:06

he's there now. Who

1:39:09

knows what's going to happen in the future, because

1:39:11

the person who's now in charge of Engadget and

1:39:13

the other related sites came

1:39:15

over from CNET. So yeah.

1:39:17

Right. I think it's

1:39:21

obviously true that there's these very difficult headwinds that

1:39:23

you spoke to. We have a video for

1:39:25

any media company, especially any media company that gets

1:39:28

a lot of its revenue from online

1:39:30

advertising. But also,

1:39:32

yeah, usually the guys who are in

1:39:34

charge are not optimizing for the long-term

1:39:36

health of these publications. It's how much

1:39:38

value can we squeeze out of them

1:39:40

in the short term and then flip

1:39:42

them for a little bit of money

1:39:44

or make a little bit of money for a couple years

1:39:47

before I go off and

1:39:49

do something else. I

1:39:51

think there are real challenges, and I think

1:39:53

the hard thing is, yeah, usually the people

1:39:55

in charge are not the ones who are

1:39:57

going to make the best decisions for the

1:39:59

long-term. I mean certainly I mean some of

1:40:01

that's personal I think Dana is it and Terrence

1:40:03

are both great people and it seems like a

1:40:05

Real whatever needed to happen there may be like

1:40:08

losing the leadership like that was not the right call

1:40:10

Well, and they've been there a long time I mean

1:40:12

there are people who have been in gadget for 10

1:40:14

years 15 years In

1:40:17

gadget was started by Jason Calacanis

1:40:19

right originally and was sold Jason

1:40:21

and Peter Rojas

1:40:23

Peter Peter Rojas and then

1:40:25

sold and went through a

1:40:28

bunch of owners Yahoo Verizon Oh

1:40:31

Jason Jason sold it to AOL

1:40:33

AOL was the phrase and then

1:40:35

and then AOL subsequently Sold

1:40:38

and resold and right I

1:40:41

was yeah I was a tech crunch for a lot of

1:40:43

that and I just made like a list of all the

1:40:45

companies that I was that we Were like owned by AOL

1:40:47

then we were owned by I think it was Well,

1:40:50

we don't own my ride and so we were called oath Then

1:40:53

we were called Verizon Media and now

1:40:55

that's private equity that company is called

1:40:58

Yahoo It's yeah,

1:41:00

but it ain't Jerry Yang's Yahoo.

1:41:02

It's a different Yeah,

1:41:05

I you know I started my journalism

1:41:07

career my transition from engineering to journalism

1:41:10

in 2006 going going

1:41:12

going to auto blog which

1:41:14

was also part of that web logs Inc

1:41:16

group Yeah, which at the time, you know

1:41:19

There was probably about 20 or

1:41:21

so sites that were all part of web log

1:41:23

thing Yeah, and this was about a year after

1:41:25

AOL had acquired it And

1:41:28

you know after after I left You

1:41:31

know after it. I think it was Yeah

1:41:36

after after AOL got spun off from

1:41:38

Time Warner Again, they

1:41:41

went through some around of cutbacks then and they

1:41:43

cut a bunch of the sites like

1:41:46

TU AW and download squad that you

1:41:48

know Christina Warren used to write for

1:41:50

and a bunch of other

1:41:52

sites You know have gone by the wayside

1:41:54

and and I think in gadget and and

1:41:57

auto blog maybe the last two or two

1:41:59

the last two big ones still going.

1:42:04

Well you know and I say this with

1:42:07

sadness I'm glad that Vindras

1:42:09

is still there. Apparently Max

1:42:11

Taney at Semaphore released some

1:42:13

internal memos describing the

1:42:15

new layout of

1:42:19

the teams. They're going to divide

1:42:21

it into two different groups, News

1:42:23

and Features, which will

1:42:25

be led by Aaron Sapouris and

1:42:28

then there'll be a team

1:42:32

called Reviews and Buying Advice

1:42:35

led by John Falcone under Laura Kenny.

1:42:39

Reviews and Buying Advice of course is

1:42:41

an SEO winner, right? That's

1:42:44

one Google will push people

1:42:46

to when they say, hey I want to buy

1:42:48

a phone, which phone should I buy? And that

1:42:50

tends to be where you make

1:42:53

money, less so in news. Evergreen content.

1:42:55

Right, right. Although this is

1:42:57

of course part of the question that publications

1:42:59

are asking themselves is that if Google just

1:43:02

populates the page with a bunch of AI

1:43:04

answers and there's no links or people don't

1:43:06

click on the links then how does a

1:43:08

site like Engadget make money? Well

1:43:11

and it's getting worse. You know I've been

1:43:13

using on the iPhone a new,

1:43:16

it's not really a browser, they call it a

1:43:18

browser, the Arc browser from the browser company. You can't

1:43:20

really do a different browser on the iPhone. It all

1:43:22

has to be WebKit to this point. That

1:43:25