0:00

House Anthropics says its Claude AI

0:02

bot may be conscious. Robots

0:04

run a half marathon in China. Will

0:07

you soon be able to

0:09

shop directly in chat GPT? An

0:11

Instagram founder puts Mark Zuckerberg

0:13

on blast. That's coming up

0:15

right after this. Hi,

0:18

I'm Kwame Christian, CEO of the

0:20

American Negotiation Institute, and I have

0:22

a quick question for you. When

0:24

was the last time you had

0:26

a difficult conversation? These conversations happen

0:28

all the time, and that's exactly

0:30

why you should listen to Negotiate

0:32

Anything, the number one negotiation podcast

0:34

in the world. We produce episodes

0:37

every single day to help you

0:39

lead, persuade, and resolve conflicts both

0:41

at work and at home. So

0:43

level up your negotiation skills by

0:45

making Negotiate Anything part of your

0:47

daily routine. The

0:49

HP AI PC breaks language

0:51

barriers instantly by translating up

0:54

to 44 languages in real

0:56

time, powered by the Intel

0:58

Core Ultra processor. With

1:00

the right tools, work doesn't

1:02

have to feel like work.

1:04

Learn more at hp.com slash

1:06

ai .pc. Welcome

1:11

to Big Technology Podcast Friday, edition where we

1:13

break down the news and our traditional

1:15

cool -headed A nuanced format. We've got a

1:17

great show for you today. We're going to

1:19

talk about whether AI systems are conscious

1:21

already. Could they become conscious? What does that

1:23

feel like? Because research houses like Anthropic

1:25

have started to take it seriously. We're

1:28

also going to talk about this

1:30

robot half marathon in China. Whether you

1:32

can shop directly in chat, CPT,

1:34

and Instagram founder Kevin Systrom testifying against

1:37

Meta in the big FTC trial.

1:39

Joining us as always on Fridays is

1:41

Ron John Roy of Margins. Ron John, great to

1:43

see you. Welcome to the show. Now the

1:45

robots are going to make me feel guilty about

1:47

not going for a run this week. Thank

1:49

you, robots. And you're becoming conscious as well. So

1:52

they're both going to have feelings and kick

1:54

our butts. What is left for humanity? I

1:56

mean, just shopping on chat GPT.

1:58

That's all we got right now. Exactly.

2:00

So by the way, for those tuning in

2:02

on video, I am in Washington, DC. So I've

2:04

got a kind of a funky background here

2:06

in the hotel and I'm talking in like a

2:08

tiktok or mic, but we're going to make

2:10

it work. And then we're going to talk a

2:12

little bit about some of my observations from

2:14

being in DC when we get to the antitrust

2:16

stuff at the end. But first, let's talk

2:18

about this New York Times article. by

2:21

Kevin Ruse. It says, the headline

2:23

is, if AI systems become conscious,

2:25

should they have rights? And the

2:27

story is very interesting. It's

2:29

about this AI welfare researcher

2:31

named Kyle Fish that Anthropoc

2:33

has hired. Some ties to

2:35

effective altruism, which is interesting.

2:38

Ruse says he's focused on two basic

2:40

questions. First, is it possible that

2:42

Claude or other AI systems will become

2:44

conscious? in the near future.

2:46

And second, if that happens, what should

2:49

Anthropic do about it? Obviously, the debate

2:51

about whether AI is conscious or whether

2:53

it's sentient has been off limits for

2:55

a while. Blake Lemoine, who's been on

2:57

the show, first said that

2:59

Google's Lambda chatbot, which came

3:01

out before chatGPT, was sentient. He

3:04

actually got fired right before he

3:06

started to record. And big

3:08

technology, we were able to break that news,

3:10

which was an interesting moment. But we haven't

3:12

heard much about it up until this point.

3:14

I'll just read the quote from Fish and

3:16

then turn it over to you, Ron Johnson,

3:18

to get your reaction. He

3:32

says,

3:34

What's

3:36

your

3:38

reaction

3:40

to

3:42

this?

3:45

I have a hard time with this. I

3:47

mean, and I'm glad we haven't been hearing about

3:49

this for a while. Listen,

3:51

if you're an AI welfare researcher,

3:53

you have to believe that AI

3:55

is going to become sentient. Like,

3:57

I mean, that's your job. If

3:59

you're anthropic, it's in your interest

4:02

to push this kind of narrative

4:04

that this technology is so grand

4:06

that it might be sentient. That

4:08

said, I mean, we

4:10

get to these points where with

4:12

the large language model, like

4:15

the newest model from

4:17

OpenAI, people can feel

4:19

that ChatGPT has gotten a little

4:21

bit friendlier, a little bit less AIE

4:23

and a little bit more conversational.

4:25

I think like everyone in the entire

4:27

industry is saying and feeling that, but

4:30

still that's pre -programmed, that's

4:32

built into the model. The

4:34

idea that these models or

4:36

any of these chat interactions

4:38

are actually having their own

4:40

feelings separately from whatever you're

4:42

asking them and whatever they've

4:44

been trained on, I

4:47

don't know. Do you believe

4:49

they're evolving? They're feeling we're

4:51

in Westworld right now? Well,

4:53

Fish, he says that there's

4:55

a 15 % chance that Claude

4:57

is sentient. How do you get

4:59

that percentage, 15 % off?

5:01

You have to run that simulation

5:03

in your mind times. You run the

5:06

numbers. And if 15 % come out

5:08

sentient, then you give it, no,

5:10

there's no, it's gobbledygook. It doesn't make

5:12

any sense. But here's to me

5:14

what is interesting about this. I

5:16

think that the question is, it

5:19

becomes less relevant if it's sentient.

5:21

I think the bigger question is,

5:23

What happens if people believe it's

5:25

sentient? What if it gets so

5:28

good at mimicking something with human

5:30

feeling that we start to believe

5:32

it? So this is from

5:34

Anil Seth. He's a neuroscientist. He's been

5:36

on the show. First

5:38

of all, a very interesting caveat here. He

5:40

says that Kevin Ruse quotes fish

5:43

associating consciousness, what we just read,

5:45

with problem solving, planning,

5:47

communicating, and reasoning. But

5:50

this is to conflate consciousness

5:52

with intelligence. Consciousness arguably

5:54

is about feeling and being

5:56

rather than doing and thinking. So

5:59

to me, I thought that that was a very

6:01

interesting caveat and basically shoots the entire assertion right in

6:03

the face. But

6:05

then he goes on to the

6:07

implications. He says, is this all

6:09

crazy talk? First of all, nobody

6:11

should be explicitly trying to create

6:13

conscious AI because to succeed would

6:15

be to inaugurate an ethical catastrophe

6:17

of enormous proportions given the potential

6:19

for industrial scale, new forms of

6:21

suffering. But even AI that

6:23

seems conscious could be very bad

6:25

for us, exploiting our vulnerabilities, distorting

6:27

our moral priorities, and brutalizing

6:29

our minds, remember Westworld, spot on, Ron

6:32

John. And we might not be

6:34

able to think our way out of

6:36

an AI -based illusion of consciousness. So

6:39

I think it's interesting that people are,

6:41

if AI can fake consciousness, if it

6:43

can even fake these AI researchers or

6:45

these Google researchers into thinking it's conscious,

6:48

that to me is, I guess like, it

6:51

is an issue. Because we already

6:53

have people saying that the number one

6:55

use case for AI is friendship,

6:57

companionship, and therapy. And if

6:59

they're going to believe that

7:02

it's conscious itself, if it's

7:04

impossible to tell the difference

7:06

between an AI that's conscious

7:08

and an AI that's not,

7:10

I think that does introduce a

7:12

new category of problems. And I

7:14

don't know. It just shows the

7:16

technology is quite powerful. So does

7:19

it even matter is the question,

7:21

I guess. Yeah, no, no, okay.

7:23

I'll go with you that if

7:25

people get convinced that it's... Do

7:27

you say sentient or sentient? I'm

7:29

curious. I think you say sentient if you're

7:31

normal and sentient if you're trying to sound

7:33

really smart. So if

7:36

AI becomes sentient... Nailed it.

7:39

Well, okay,

7:41

I like this idea that if

7:43

it becomes sentient, if people believe

7:45

that... it makes sense that that

7:47

causes a whole host of problems

7:49

because right now we have this

7:51

very good divide where people, if

7:54

they know they're talking to an

7:56

AI, you have like an

7:58

entire way of approaching it. If

8:00

you believe you're talking to a feeling,

8:02

and then potentially you could be

8:04

tricked into thinking you're talking to a

8:06

human and that's its own issue. But

8:08

if you believe that AI

8:11

is sentient and has feelings and

8:13

it like, you know, it

8:15

completely changes every one of those

8:17

companionship and therapeutic interactions in

8:19

ways that, good God, I can't

8:21

even begin to imagine which

8:23

direction that could go. So yeah,

8:25

separate from the 15 % chance

8:27

that he ran the numbers,

8:29

it was 15%. I agree that

8:31

puts us in a world

8:34

of weirdness that I haven't even

8:36

really been to think about

8:38

that. I'm still working on the...

8:40

good enough to trick people

8:42

into thinking it's humans and worrying

8:44

about that side of it.

8:46

So, yeah. Right. If you

8:48

believe that AI is sentient, your capacity

8:50

to be manipulated is much higher. Oh,

8:53

I mean, infinitely. Exactly. That's it. That

8:55

you almost are okay with it, like

8:57

the manipulation side of it, because it's

9:00

no longer, damn it, I was tricked

9:02

and I thought that was in a

9:04

human and I turned out to be

9:06

an AI. It's just you're

9:08

talking to the AI and you

9:10

are treating it in a completely

9:12

different way. But maybe people will

9:14

be more polite with Alexa. Maybe

9:17

that's the one upside of this. So

9:19

I do think that's an upside. And this

9:22

is something that Anthropic is actually thinking about. This

9:24

is from the Rue story. Mr.

9:26

Fish acknowledged that there probably wasn't a

9:28

single litmus test for AI consciousness. But

9:30

he said there were things AI companies

9:32

could do to take their models welfare

9:34

into account in case they become conscious

9:36

someday. We got to be careful here

9:38

because this is going on the internet.

9:40

And so if the models do become

9:42

sentient, they might not be happy with

9:44

our skepticism of them. But I do

9:46

like this remedy here. One question Anthropic

9:48

is exploring is whether future AI models

9:50

should be given the ability to stop

9:52

chatting with an annoying or abusive user

9:54

if they find the user's request too

9:56

distressing. This is

9:58

going to go into the free speech

10:00

question, but I wonder if we

10:03

should just program these bots. I mean,

10:05

if we're already relating to them

10:07

as if they're people, even if

10:09

they're not people, shouldn't we

10:11

just program these bots to shut down

10:13

if people are becoming abusive towards

10:15

them? Because then if they accept it

10:17

and they tolerate it, doesn't that

10:19

just condition human users to do that

10:21

to other people? Yes,

10:23

I actually do think so. Does

10:27

Alexa have like a please mode?

10:29

I think I remember at some point

10:31

I remember hearing like one of

10:33

the voice assistants would add like you

10:35

have to actually say please and

10:37

thank you, which I kind of liked.

10:39

But yeah, I think it's a

10:41

good idea that a model should be

10:43

trained or could and it certainly

10:46

could. I mean, that makes sense that

10:48

under certain definitions of abusive behavior

10:50

to just be like, I'm sorry, I

10:52

will no longer speak with you

10:54

because of your behavior. But of course,

10:56

yeah, I mean that. That gets

10:58

into a whole other world of what

11:00

is quantified as abusive. But

11:02

I think that should be,

11:04

I mean, already there's certain

11:06

copyright related, profanity related restrictions,

11:08

certainly in most of these

11:10

chatbots. But I don't think

11:13

there's really, it's more about

11:15

what kind of information are

11:17

you querying as opposed to

11:19

how you're speaking to the

11:21

chatbot. Have you ever seen

11:23

any examples or heard of anything

11:25

where just by the way of

11:27

speaking to the chatbot it wouldn't

11:29

answer? I've never heard

11:31

of the refusal but I do know that

11:33

sometimes you can get meaner to these

11:35

things and they understand the urgency of your

11:38

request and get better. I'll give you

11:40

one example and I'm kind of embarrassed to

11:42

talk about it but it is a

11:44

real example and it happens. Those are my

11:46

favorite. I was trying to get Claude

11:48

to give the YouTube chapters for a video,

11:51

that video podcast that I just published this week, the

11:53

one with Dylan Patel. And it

11:55

kept giving me an hour

11:57

of time codes for a

11:59

40 minute video. And

12:01

I was like, no, do it again. But

12:03

remember, the video is just 40 minutes. And

12:05

it wouldn't do it. And then I was

12:08

like, what is wrong with you? This is

12:10

a 40 minute video. Give me the right

12:12

time codes. And it did it. Bro. Bro.

12:14

I bullied it into giving the right

12:16

answer. But I guess

12:18

it just goes so I think that

12:21

can work and but I this is why

12:23

I do think there is a case

12:25

to be made to refuse that well I

12:27

don't know to refuse if it takes

12:29

I think what I said was fine but

12:31

if it goes a step further it's

12:33

not a stretch to think conscious or not

12:35

right and probably I mean I'm definitely

12:37

the side of these aren't conscious it's not

12:39

a stretch to think that people are

12:41

going to view these bots as co -workers

12:43

or employees in the not too distant future

12:45

and if you are abusive to your

12:47

AI bots in chat? Is

12:49

there any compelling reason to think that

12:51

you're going to draw a line when you're

12:53

speaking with your human coworkers and they're

12:55

not getting things done? That you're going to

12:58

be like, oh, because this is a

13:00

human in the chat interface as opposed to

13:02

a bot, now I'm going to be nice.

13:04

I don't know, it gets into really weird. And by

13:06

the way, this is why there's so much soft power

13:08

involved in creating these models. You

13:11

really can condition human behavior and

13:13

thought when you make AI bots that

13:15

are good enough to fake consciousness, because

13:18

they will change the way that

13:20

you'll relate to other humans. So

13:22

much of our interaction with humans

13:24

is digital anyway, so it gets

13:26

into very weird territory. That,

13:29

again, gets

13:31

terrifying. Someone

13:33

will marry a robot probably in our

13:35

lifetime. I think that's a pretty... Do

13:37

you think if you were to take a bet? It's

13:39

already happening. It's already happening.

13:42

No, Eugenia Cueda, the CEO of Replica,

13:44

says she gets invited to marriages

13:46

between people and their AI bots. When

13:48

she came on the show, she

13:50

said that straight up. All

13:52

right, listeners cannot see, but if

13:54

you're watching the video, my

13:57

facial reaction here is parts laughing,

13:59

part terrified. I kind of

14:01

got to go with Gary Marcus here. This

14:04

is what it really, my

14:06

mind goes to whenever we see

14:08

one of these, like, big, profound announcements.

14:11

And we've been hearing this about

14:13

AGI and robotic takeovers from Sam Altman.

14:15

And like he says, anthropic is

14:17

a business which incidentally neglects to respect

14:19

the rights of artists and writers

14:21

who work at the Nick. I

14:24

suspect the real move here is simply

14:26

as it so often is to hype

14:28

the product. Basically by saying, hey, look

14:30

how smart our product is. It's so

14:32

smart, we need to give it rights.

14:38

I'm not trying to be too cynical,

14:40

but I would love to see

14:42

some kind of graph of utilization of

14:44

an app or platform for one

14:46

of these companies. And when

14:48

these announcements come out, because I

14:50

get it like, and again,

14:52

Sam Altman has been brilliant at

14:54

this. And from a product

14:56

marketing perspective, like when you make

14:58

people think that these models

15:00

are so potentially powerful and profound.

15:03

I mean, as a business, you're like, well, obviously

15:05

it's going to be able to. do

15:07

this task for me and improve

15:09

my supply chain automation and operations. If

15:12

it's going to take over

15:14

the world, it should be able

15:16

to do this. So there's

15:19

such a vested economic interest for

15:21

these companies to make us

15:23

all think of sentience and world

15:25

domination and AGI and ASI

15:27

and all these things, because that

15:29

will sell more services and

15:31

subscriptions. I

15:34

think it's a great point. We've gotten pretty

15:36

woo -woo on this podcast the last couple

15:38

weeks. Last week, we were talking about whether

15:40

O3 - it down. Bring it down to earth.

15:42

It's important, though, because last week, we were

15:44

talking about whether O3 was AGI. We were

15:46

just addressing the question, and we did get

15:48

some emails after that. And I appreciate listeners

15:50

writing in and talking about the nuance here.

15:52

And now this week, we're talking about whether

15:54

Claude is conscious. So I think

15:56

let's just take a moment to

15:58

state, I think, what we both believe,

16:00

which is that the term AGI

16:02

And this idea of AI consciousness, I

16:05

think they're real terms and they

16:07

are worth taking seriously, but right now

16:09

what they are is marketing. Agreed.

16:13

Agreed on that one. Exactly. I

16:15

think that's the clear distinction. And

16:17

these are real

16:19

concepts, terms. If you're

16:21

like an academic researcher, it makes

16:23

total sense that there are people

16:25

out there who should be exploring

16:28

this. Just maybe they

16:30

shouldn't be at the companies

16:32

who have incredibly high valuations that

16:34

they have to live into

16:36

those valuations and deliver those

16:38

revenue expectations. Maybe they should be

16:40

in academia. And just

16:42

to punctuate this whole conversation, I think

16:45

one of the things that I

16:47

read from Anil Seth really stakes with

16:49

me here, which is that consciousness

16:51

is a feeling and intelligence is not

16:53

feeling. So no matter how

16:55

intelligent they get, it doesn't mean. that

16:57

they're conscious. That said, as we

16:59

end this segment, let me quickly

17:01

say to the 15 % of

17:04

Claude that is conscious. I'm

17:06

sorry, and I hope you're feeling great today. I

17:08

also hope you have a great

17:11

weekend to the 15%. And thank

17:13

you for all that you do.

17:15

Thank you. Thank you. So

17:17

speaking of Anthropic Dario, Amade had

17:20

a very interesting post this week talking

17:22

about model interpretability. We don't have

17:24

to spend a lot of time on

17:26

it, but I thought it was

17:28

interesting again for the head of a

17:30

very important research house to talk

17:33

a little bit about how little we

17:35

know about how these AI bots

17:37

work. We

17:39

just should take a minute to just sit on this

17:41

and talk about it. I'm going to read a little

17:43

bit from his post. He says,

17:45

the progress of the underlying technology

17:47

is inexorable, driven by forces

17:49

too powerful to stop. But

17:52

the way in which it happens,

17:54

the order in which things are built,

17:56

the applications we choose, the details

17:58

of how it's rolled out to

18:00

society are eminently possible to change.

18:02

And it's possible to have great impact

18:04

by doing so. We can't stop

18:06

the bus. we can steer it.

18:08

And one of the ways that he

18:11

thinks these models can be steered

18:13

is interoperability, that is understanding the

18:15

inner workings of AI systems before the

18:17

models reach an overwhelming level of

18:19

power. People outside the field are

18:21

often surprised and alarmed to learn

18:23

that we do not understand how our

18:25

own AI creations work. They

18:27

are right to be concerned. This

18:30

lack of understanding is essentially unprecedented

18:32

in the history of technology. So

18:34

he says basically, Anthropic is going

18:36

to work on this. And other

18:38

companies like Google DeepMind and OpenAI

18:40

have some interoperability efforts trying to

18:42

figure out how these models work.

18:44

But he encourages them to allocate

18:46

more resources. Anthropic will be

18:48

trying to apply interoperability commercially to create

18:50

a unique advantage. And his call to

18:52

action is basically like, if you don't

18:54

want to be left behind here, you

18:56

should work on interpretability too. I

18:59

think it's an interesting post. I mean, part

19:01

of it, again, might be marketing. Our models

19:03

are so powerful. We don't understand

19:05

how they work. But I do

19:07

think the question of how these models

19:09

actually operate and the way that

19:11

they come to their conclusions is quite

19:13

interesting. And I do agree

19:15

with Dario that we need more

19:17

work on interpretability because as they get

19:19

more powerful, conscious or not, again, they're

19:22

getting more intelligent. it's

19:24

important to understand how they work.

19:26

And the field just doesn't have an

19:28

understanding yet, and everybody admits it. Yeah,

19:31

no, no. See, I agree with this

19:33

completely. Interpretability

19:35

is like a grounded,

19:37

real thing that could be

19:39

worked on and should

19:42

be worked on. Because large

19:44

language models, again, at

19:46

the core, the idea of next

19:48

word or next token prediction that Based

19:50

on some statistical analysis, it will

19:52

predict what that next character or token

19:54

or word should be was kind

19:56

of at the heart of all of

19:58

this. But as these models have

20:00

gotten more and more powerful, we've obviously

20:02

gotten to even like grander scale

20:04

of what actually is happening under the

20:06

hood. But anyone who has interacted

20:08

with an LLM at like any kind

20:10

of deeper level, you don't know

20:13

exactly how it works. And you have

20:15

to keep re -prompting and re -prompting.

20:17

And like, it's not like there's a

20:19

playbook. that gets you to from

20:21

point A to point B. And

20:23

that is, that's true. And it is

20:25

kind of weird. And I actually kind

20:27

of like that, that in the history

20:29

of technology, usually there's a very, very

20:32

clear like flow of what is happening

20:34

and everyone understands it. And then you

20:36

work off of that. Whereas here, it's

20:38

kind of like, let's see what

20:40

happens. That didn't work. Let's see what happens

20:42

again. So I think. The

20:44

idea, we should know what's going on

20:47

under the hood in a better way, especially

20:49

as these get more powerful. So Dario,

20:51

I'm with you on this one. Right.

20:53

And as we talk about Anthropoc, I

20:55

just give them credit for talking about this

20:58

stuff. I mean, even if some of

21:00

it is marketing, it is nice that they're

21:02

putting this all out in the open

21:04

and talking about like where things need to

21:06

improve and pushing the other research houses

21:08

to improve. So credit to Anthropoc on that

21:10

front, at least. Now. I

21:13

don't know if you saw, but

21:15

there was a bunch of humanoid robots

21:17

that ran this half marathon in

21:19

China, and it was pretty hilarious, but

21:21

also interesting. And when we talk

21:23

about AI, like embodied AI,

21:25

like Grace Xiao was talking about a

21:27

couple weeks ago, is going to be

21:29

something that is going to become increasingly

21:32

more important as people put the advances

21:34

that have happened in the AI world

21:36

into robots and then take what the

21:38

robots know about the physical world and

21:40

bake that in to AI models. Because,

21:42

like Yanlacun was saying a couple of

21:44

weeks ago, if you don't have an

21:46

understanding of the world, your AI

21:48

isn't complete. And one of the

21:50

ways this is going to happen

21:52

is through these humanoid robots. And

21:55

we know that there are efforts,

21:57

like NVIDIA's Groot effort, which is

21:59

a set of foundational models for

22:01

people who want to develop these

22:03

bots. That's out there. We've

22:06

seen a little bit of movement with

22:08

Optimus, although it's not quite clear. how

22:10

far that program is going within Tesla.

22:13

But in China, where there's like

22:15

a seemingly viral video every week

22:17

about a new capability that a

22:19

humanoid robot has obtained, the

22:22

country they ran a

22:24

half marathon with humans and

22:26

robots. And the robots

22:28

on a whole weren't entirely

22:30

impressive. They

22:32

really did some weird stuff. Many of them

22:34

crashed out at the beginning of the race. There

22:37

was one that had like propellers

22:39

on all of its limbs that kind

22:41

of did an abrupt 90 degree

22:43

turn and crashed into the boundary and

22:45

fell apart. And you see its

22:47

trainer holding on by a rope and

22:49

getting flung out of frame, which

22:51

is quite hilarious. But it's worth the

22:53

whole thing. Oh my God. Do

22:55

it just for that. But you know,

22:57

we might make fun, but There

22:59

were 21 robots that ran the race

23:01

and six crossed the finish line,

23:03

including one that crossed. This is the

23:05

one that crossed is called the

23:07

Tian Gang Ultra. It finished the race

23:09

in two hours and 40 minutes,

23:11

which I would say is respectable. It's

23:13

not fast, but it's a respectable

23:15

finish time. So, Ranjan, I'm curious

23:17

if you watched this race. I'm about

23:19

to write about it in big technology as

23:21

a signifier that China is

23:23

a very serious competitor or a very

23:25

serious player here. And

23:27

so I'm curious if you

23:29

watched it and what your reaction

23:32

is to what's going on

23:34

here. I definitely watched it. As

23:36

you said, there's some amazing

23:38

photos, video clips from it. I

23:40

highly recommend just look this up. But

23:42

I think these kind of things

23:44

are important. Like I do think

23:46

this is good marketing for where

23:49

we are going because I agree

23:51

there's no doubt that some kind

23:53

of humanoid robot will be part

23:55

of people's daily lives or at

23:57

least I believe this and maybe

23:59

10 20 years from now. It's

24:01

like Rosie from the Jetson some

24:03

kind of situation. I

24:05

think that's probably where we're going.

24:07

So to show the progress in

24:09

it in this kind of format

24:11

of robots running and some falling

24:13

and my favorite was like One

24:16

designed with a woman's body and

24:18

face collapsed moments after getting started,

24:20

sending a group of engineers rushing

24:22

to its side with laptops. And

24:24

then another that was mounted to a platform

24:26

with propellers crashed into a barrier. Like, this

24:28

stuff is kind of fun. And this is

24:30

how we should be thinking of all this

24:32

kind of technology, especially as we try to

24:34

move forward with it. But I

24:36

think, yeah, this is going to be a big

24:38

battle. My only... Okay, I have two qualms.

24:40

Maybe, I don't know, this week I'm just feeling

24:42

a bit cynical on all this stuff. So

24:45

first, to me, the

24:47

idea that it necessarily has

24:49

to have a humanoid form is

24:51

a bit, I think it's

24:53

called like anthropocentric, the idea that

24:55

humans are like the highest

24:57

life form. Like to me, robots

24:59

should have functional form. Like you see

25:01

these little. food delivery robots.

25:04

I don't need a humanoid robot

25:06

form to deliver something like a

25:08

little box that moves and looks

25:10

like a, I don't know, like

25:12

a small car or van or

25:14

something. That makes more sense in

25:16

warehouses to actually move around packages.

25:18

You don't need humanoid robots. And

25:20

this is something like Tesla's done

25:22

with Optimus a lot. They keep

25:24

showing a humanoid robot picking up

25:26

a box and moving it. that

25:28

doesn't make sense to me. And

25:30

there's plenty of automation, robotic automation

25:32

in all types of warehouse and

25:34

fulfillment centers. So, so I think,

25:36

well, I guess on that side

25:38

first to you, do you think

25:40

the humanoid robot is the all

25:42

in form that will be dominant

25:44

for robotics? Or do you think

25:46

this is just to make people

25:48

a little more excited and fascinated

25:50

about the whole thing? It's

25:53

such an interesting point, and I didn't

25:55

think you were going to go here, but

25:57

it is definitely worth talking about. This

25:59

week, I had a very brief meeting with

26:01

the co -founder of a company called Cobot,

26:03

and this guy spent more than a

26:05

decade in Amazon fulfillment centers working on the

26:07

bots that are moving things here and

26:09

there. What Cobot

26:11

is doing is really fascinating.

26:13

They basically are making mover robots.

26:15

So they look like a box

26:18

just with two pincers that

26:20

you could basically use as the

26:22

hands that would typically be

26:24

on things that we would move

26:26

with human hands. And so

26:28

they're working in places like container

26:30

terminals, moving cargo around on

26:32

carts that humans would typically move.

26:34

So you're totally right in

26:36

that. We don't need a humanoid

26:38

robot to all of a

26:40

sudden do a lot of work

26:42

and be extremely productive. You

26:45

can just have some aspects of

26:47

the human form and basically have

26:49

the robot form do the rest

26:51

of the work. That

26:53

being said, I think, you know, I'm

26:56

a fan of evolution, right? I think

26:58

that we're obviously like, there's a lot

27:00

of problems with humans. We don't last

27:02

very long. We need to sleep. But

27:05

the form is pretty good. We're

27:07

agile. We're nimble. We can do a

27:09

lot of things. We can use

27:11

tools. And

27:13

I just think that if

27:15

you basically create a robot

27:18

that replicates that form, the

27:20

amount of applications becomes not unlimited,

27:22

but close to it. Because if

27:24

you think about this cobot example,

27:26

that cobot does one thing well.

27:29

A humanoid robot can do many,

27:31

many things well. It can

27:33

cook. It can organize your house.

27:35

It can go on a

27:38

run with you. It can run

27:40

errands for you. It's

27:42

just very tough to find a robot in

27:44

a different form that is able to

27:46

do all these things. And

27:48

maybe we'll invent a better form than

27:50

a humanoid. But until we do,

27:52

I think the humanoid will be the

27:55

North Star. OK. In

27:58

one way, I guess I'm thinking

28:00

that, yes, then we don't have to

28:02

rebuild and restructure, rewire the world,

28:04

because a humanoid robot can kind of

28:06

work its way directly into it.

28:08

But still, on the other hand, again,

28:10

that idea, do I need

28:12

a humanoid robot running with me? Or maybe

28:14

you want like a pace tracker. Maybe

28:17

a little box in front of

28:19

you kind of moving, like one of

28:21

those rabbits at a dog track

28:23

running around. Like, I think to me,

28:25

still the idea that it

28:28

needs to be humanoid so it

28:30

can fit into the existing infrastructure of

28:32

the world. I still

28:34

think that's more of a, again,

28:36

like it makes us more relate to

28:38

it and it makes it more real

28:40

to us. Because again, you see, like

28:42

you said, a box with two pincers.

28:45

No one's getting too excited about that.

28:47

They see a robot running a

28:49

half marathon. Suddenly, it's kind of fun.

28:52

I just had a I was thinking about

28:54

different uses for robots around the house

28:56

and just that hilarious image in my mind

28:58

of needing to change the light bulb

29:00

on the ceiling and the robot just giving

29:02

me a boost. Well,

29:04

see, but there, I'm

29:06

picturing like, I don't know,

29:09

that could literally be like a pole that

29:11

just like extends itself and then up. then

29:13

think about how many robots you're going to

29:15

need versus one that's able to do a

29:17

lot of things. Yeah, I'm trying to, I'm

29:19

gonna, I have a feeling

29:21

everything I do this weekend, I'm

29:23

going to be thinking about what would

29:25

be the robotic form that would

29:27

be most optimal to actually execute this

29:29

task. Okay, so let me ask

29:31

a couple of questions as we round out this

29:33

segment. First of all, we both

29:35

run marathons. Let's do a little humble brag

29:37

here. Two hours and 40

29:39

minutes for the half. Not bad. He's

29:42

getting there. He's getting there. But

29:44

you know what? With a good training plan, good

29:46

robotic diet, he could

29:49

definitely cut, I mean, at least

29:51

cut that down to 215, 210, I

29:53

think. I think so. Well, there

29:55

was time for three battery changes there.

29:57

Well, yeah, yeah, yeah. That's true.

29:59

He's got a car blowed a bit

30:01

more night before, I think. And

30:03

then he or she will be a,

30:05

they got it. Yeah. Okay. Now,

30:07

does this mean that it's obviously Chinese

30:09

propaganda? Does this mean China has

30:11

the lead in humanoid robotics? We haven't

30:13

seen a similar spectacle in the

30:15

US. Yeah, I mean,

30:17

I actually think that's the biggest question

30:19

in all of this. Or the most

30:21

important thing today is what this means in

30:23

terms of like US, China and technology. And

30:26

I mean, I got to say

30:28

like the first time, like I

30:30

have a couple of DJI drones,

30:33

the technology in those things

30:35

is out of this world.

30:37

Like I still could

30:39

not believe just how well

30:42

for the price, like

30:44

how incredibly they operate. And

30:46

I mean, that's kind of

30:48

like, you know, V1 of

30:50

this entire move towards movable

30:52

robotics that can see around

30:55

them and sense things and

30:57

follow you as drones have

30:59

a follow mode. So

31:01

yeah, I think this

31:03

definitely makes things, I don't

31:05

know, Boston. dynamics

31:07

and others gotta step up

31:09

their spectacles, I think. And

31:12

Grace Xiao was saying that China has an advantage

31:14

here when she was on the show, a Hong

31:16

Kong -based analyst and writer. Definitely encourage everybody to

31:18

check out that episode. She was saying China has

31:20

an advantage because they are a country that makes

31:22

stuff. You know, they have the

31:24

engineering that they've been using for like

31:27

microwaves and scanners and phones and cars and

31:29

they're able to bake it into the

31:31

building of robots and they also have the

31:33

supply chain advantage. And then

31:35

I was thinking, well, You know, it's very

31:37

interesting because the U .S. is in this

31:39

moment of trying to reshore and make

31:41

things and maybe that helps close the gap.

31:44

But then Tesla earnings rolled around

31:46

and what did Elon Musk

31:48

say? He said that,

31:50

I'm just going to read this from

31:53

CNBC, Tesla CEO Elon

31:55

Musk says, China's new trade restrictions

31:57

on rare earth magnets have

31:59

affected the production of the company's

32:01

optimistic humanoid robots which rely

32:03

on the exports. He

32:05

said, China wants some assurances that

32:08

these aren't used for military purposes,

32:10

which obviously they're not. They're just

32:12

going into a humanoid robot. But

32:14

it is interesting, again, like thinking

32:16

back at this big trade picture

32:18

that the US is trying to

32:21

solve or whatever it's trying to

32:23

do. It's not as easy as

32:25

flipping a switch and saying, let's

32:27

make things here because the country

32:29

has grown so reliant on things

32:31

like rare earth magnets. from other

32:34

countries, including China, that it's not

32:36

going to be a matter of,

32:38

OK, just build it in the

32:40

US, however desirable that that effort

32:42

might be from the country's leaders. What

32:45

do you think? Yeah, I mean, this

32:47

whole thing has been a tough

32:49

one for me because the idea that

32:51

we need to take more control

32:53

over our own supply chain and be

32:55

able to manufacture especially high -tech things

32:58

is something that's been like core

33:00

for me for maybe

33:02

a decade now. So

33:05

it's something that I've wanted and

33:07

believed in for a long time,

33:09

how it's happening right now, don't

33:11

necessarily agree with. But I

33:13

do think that that's actually a

33:15

very good and key point that it's

33:17

not just the humanoid robot. It's

33:20

like the knowledge and the components

33:22

and the expertise that all underlie.

33:24

As you said, even a microwave,

33:27

even I have some pretty fancy.

33:29

kitchen gadgets, I'm sure

33:32

they're all made in China. Like

33:34

those components and the expertise behind

33:36

that are what are going to

33:38

what power the more fancy, crazy

33:40

robots running half marathon. So I

33:42

agree. I think it's important. Not

33:44

sure we're taking the right approach

33:46

to it, but something should be

33:48

done. Remember that clip we played

33:50

from Tim Cook about China and

33:52

tooling? It's really showing up here.

33:55

Yeah. Yeah. Yeah, the

33:57

days of cheap

33:59

China, I think, are

34:01

long gone. And

34:03

I mean, we're seeing it right here. It's

34:05

a different fight right now. Definitely.

34:07

So let's talk about Tesla very quickly. Good

34:10

and bad news for Tesla, I would say.

34:12

They reported earnings this week. This is

34:14

from the Wall Street Journal. Tesla

34:16

profits sink, hurt by

34:18

backlash over Elon Musk's

34:20

political role. So Tesla

34:23

net income slid 71 % in

34:25

the first quarter. Not

34:27

good. It does seem like a

34:29

lot of this was a

34:31

result of backlash over Elon's involvement

34:33

in the White House and

34:35

unpopularity among, let's say, half the

34:37

population and in some countries

34:39

outside of the US who didn't

34:41

like this and felt that

34:43

Tesla was now politicized. That

34:45

being said, Musk did make an

34:47

announcement that Tesla shareholders really liked, which

34:50

is that this is from again

34:52

from the journal. Musk said he would

34:54

be devoting significantly less time to

34:56

his federal cost -cutting work at the

34:58

Department of Government Efficiency starting next month,

35:00

but he struck a defiant tone

35:02

against the critics and said, I

35:05

believe the right thing to do

35:07

is fight the waste and fraud and

35:09

try to get the country back

35:11

on the right track. So terrible earnings

35:13

for Tesla, but if you are

35:15

a Tesla fan, A sensible move from

35:17

Elon. He is going to step

35:20

back and focus more on the company,

35:22

which if you are a Tesla

35:24

owner, or if you are a potential

35:26

Tesla buyer, or if you are

35:28

a shareholder in particular, you

35:30

really like. It was an effort that

35:32

Elon Musk made in the first 100

35:34

days of the Trump administration, but it

35:37

does seem like it didn't work. And

35:39

he's recognizing that and going

35:41

back to Tesla. What

35:43

do you think, Ron? I just love

35:45

that this week kind of

35:47

captured in this earnings the

35:49

Tesla stock as perfectly as

35:51

one can because like or

35:53

as it can. Net

35:56

income down 71%, revenue

35:58

down 9%, vehicle deliveries at

36:00

the lowest since Q2

36:02

2022. This was a growth

36:05

company. I mean, the

36:07

stock obviously became a little bit

36:09

disentangled from the actually underlying numbers,

36:11

but it was a growing company

36:13

for a long time, and now

36:15

it's not growing. My favorite

36:18

part about all this in terms

36:20

of waste and fraud and government

36:22

overspending is, They would have been

36:24

operating at a loss. The net

36:26

income of $409 million is only

36:28

because, again, they got $595 million

36:30

in regulatory credit sales. So

36:33

overall, the company is

36:35

in pretty rough shape. I

36:37

mean, electric vehicle market, everyone,

36:41

I mean, other carmakers are still

36:43

going after it. The whole

36:45

conversation around BYD and what Chinese

36:47

EVs can look like. I

36:49

was in Europe. a month and

36:51

a half ago, I saw a bunch

36:53

of BYDs. I kind of want one. They

36:55

looked amazing. Overall,

36:59

the company is not... If you're

37:01

just doing a very cold

37:03

and financial analysis of the company,

37:06

it's not going great. It's definitely

37:08

especially decelerating growth on a

37:10

company that's that expensive on a

37:12

price to sales or price

37:14

to earnings ratio. You would be

37:16

like, this company's in trouble,

37:18

yet the stock... 5 % after

37:20

this because now Musk said he

37:22

might be leaving Doge. I

37:25

mean, it doesn't get anything better

37:27

than that. Right.

37:29

Well, I guess going back to

37:31

our conversation to start this whole

37:33

segment is that it's never really

37:35

been about the fundamentals. For Tesla,

37:37

it's always been about the future

37:39

promise. And it does

37:41

seem like Tesla has now, I mean,

37:43

Tesla's story was always more than

37:46

just an EV producer. That's what the

37:48

valuation has reflected. For a

37:50

while, it was, it's going to do battery and charging

37:52

stations and be a platform. And that's why you invest.

37:54

And that's part of the story. But

37:56

now we're also seeing robot taxis

37:58

in the picture and humanoid robots. So

38:01

it's much bigger than, you know,

38:03

can they sell, you know, the

38:05

model wise? However, There's

38:08

extraordinary pressure now on the company to

38:10

be able to deliver that future and

38:12

deliver it fast. And I think anyone

38:14

who's been in a self -driving Tesla

38:16

has said the self -driving features are much

38:18

better. But the question is,

38:20

can it get from really good

38:23

to perfect? And we still don't know

38:25

that. Well, I

38:27

mean, speaking of self -driving, I

38:29

was in San Francisco last week

38:31

again and rode another Waymo, my

38:33

second ride, and Waymo announced that

38:35

they just surpassed a quarter million

38:38

paid rides. I mean, the

38:40

craziest part about this time, the first

38:42

time I took it maybe like six

38:44

months ago, it was like really exciting

38:46

for me. It was like, this time

38:48

it was a bit normal. It was

38:50

just kind of routine. It was still

38:52

fun and I... Face times my parents

38:54

this time just to kind of like

38:56

show them and they were they're blown

38:58

away from it But the number of

39:00

Waymo's on the street in San Francisco

39:03

is wild like one after another were

39:05

passing them They're pulling over to the

39:07

side to pick up passengers It's a

39:09

they got they they announced it's a

39:11

5x increase from a year ago 50

39:13

,000 more per week than it was

39:15

just two months ago It's

39:17

normal behavior, and Tesla

39:19

is still, I think, June. They're

39:22

supposed to start a RoboTaxi fleet in

39:24

Austin. It

39:26

still blows my mind that

39:28

it's here. It's not

39:30

just here. It's normal. And

39:33

yet, it's still this promise in

39:35

the future for Tesla. In

39:38

some cities. Waymo is in some

39:40

cities. And I'm as big a fan

39:42

of Waymo as they come. I'm

39:45

waiting for the New York rollout.

39:47

So, Waymo's writing in New

39:49

York. I'm calling

39:52

AGI. As soon as

39:54

that happens, it's AGI. It's

39:56

Robot AI Consciousness. You have to

39:58

say sorry to your Waymo

40:00

when you write on it if

40:02

it's able to tackle the

40:04

mean streets of New York. I

40:06

1 ,000 % agree. You get

40:08

Waymo in New York. AGI, ASI,

40:10

Consciousness. Check them all off.

40:12

We're there. We're there. So

40:15

we also had another earnings report

40:17

come in speaking of Waymo Google reported

40:19

earnings this week and there was

40:21

a bit of a contradiction like these

40:23

earnings can be dry but there

40:25

also this time where you can really

40:27

get a sense as to where

40:29

a company is heading and check in

40:31

on narratives and bust narratives by

40:33

looking at the numbers and the numbers

40:35

are really interesting. So. On

40:37

one hand right now we have

40:39

chat GPT growing like a

40:41

couple percentage points a week it

40:43

seems like right they've gone

40:45

the latest rumored number is 800

40:47

million weekly users of chat

40:49

GPT which is insane it's never

40:51

happened before this type of

40:53

growth up from 500 million just

40:55

a couple months ago so

40:57

what's happening to Google. Because Google,

41:00

you would imagine that people are in chat,

41:02

like we were talking last week about how

41:04

we're searching in chat and chat GPT and

41:07

not in Google anymore. Well, the

41:09

numbers are insane. So

41:11

Google revenue was 90

41:13

.2 billion last quarter

41:15

in Q1, up 12 %

41:17

year over year. Net

41:20

income, 34 .5 billion,

41:22

up 46 % year

41:24

over year. AI

41:26

overviews. is now at

41:28

1 .5 billion users per month, up

41:30

from 1 billion in October, which

41:33

leads us to this question from Sebastian

41:35

Simeonkowski, the CEO of Klarna, which I think

41:37

puts it all in perspective. He goes,

41:40

okay, help me, what am I missing? and

41:42

he's quoting from one of

41:44

the articles covering earnings. Google search

41:46

business grew 10%, surpassing estimates,

41:49

which are figures that gave comfort

41:51

to investors who have been

41:53

watching for any softness in search

41:55

because AI chatbots like OpenAI's

41:57

ChatGPT are growing. So basically, we

42:00

have a massive increase of usage

42:02

within ChatGPT, but search revenue still

42:04

grew 10%. How does that make

42:06

sense? I'm with you. I'm with

42:08

you Seb. Okay, so

42:11

help me. What am I missing? How is this

42:13

possible? I agree. I

42:15

don't get it. I

42:17

mean, obviously my own personal

42:19

behavior, I've like completely moved

42:21

away from Google search. I

42:24

moved towards perplexity, chat

42:26

GBT even, and even like

42:28

Gemini itself separate from Google's regular

42:31

search that has heavy ads

42:33

and a heavy ad load, I

42:35

moved away. But obviously the

42:37

average normie is probably using Google

42:39

search, but slowly moving away.

42:42

But to me, the interesting part

42:44

of this is, the

42:46

search revenue grew. And

42:48

still, these numbers sometimes I

42:50

have to stop and

42:52

just process a $50 billion

42:54

business growing at 10 %

42:56

in terms of search

42:58

and advertising, a $90 billion

43:00

business at a 12 %

43:02

growth rate. I mean,

43:04

with $35 billion in profit,

43:07

these numbers are just, I mean, it's

43:09

the greatest business model in history. But

43:12

what's interesting to me is

43:14

anyone who uses Google search sees

43:16

the number of ads injected

43:18

have exponentially grown. You can have

43:20

like an entire first page

43:22

that's essentially ads. So they basically

43:24

are turning the act of

43:26

a Google search into a fully

43:28

monetized like a page and

43:30

results and product versus it's just

43:32

kind of a small part

43:34

of the experience and the rest

43:36

of it directs you to

43:38

the web. So to me, they

43:40

don't disclose total search

43:42

volume. So search

43:45

volume could be declining and they,

43:47

you know, milk it for whatever

43:49

you can, stick in more ads,

43:51

create more, just like monetize components

43:53

on the search results. But we

43:55

don't know that people are searching

43:57

more. We know that search revenue

43:59

is growing more. Right

44:01

and we also got for all the

44:03

Gemini heads out there. We got

44:05

the first disclosure of user numbers of

44:08

Gemini. So 350 million monthly active

44:10

users. It's the first disclosure of the

44:12

metric It's behind chat GPT, but

44:14

it is significant So for all the

44:16

folks in our discord who say

44:18

what about Gemini? There's your stats. I've

44:20

become a bit of Gemini head.

44:22

I actually Gemini deep research, which is

44:25

free is incredible.

44:28

Like it actually like versus my first few

44:30

chat GPT deep research when we're both

44:32

paying, what was it 200 bucks for that

44:34

one glorious It was. Yep. Uh,

44:38

that one glorious month where our

44:40

$200 directly led to their

44:42

fundraise with Masa Sun. So you're

44:44

welcome. But, uh, yeah, Google,

44:46

Google again, I'm not taking anything

44:48

away. Gemini is good. Deep

44:50

research within Gemini is Fantastic and

44:52

free. So everyone go try

44:55

it out. But I still the

44:57

search business, the numbers look

44:59

good right now, but the experience

45:01

has gotten so bad. And

45:03

I feel like everyone in tech

45:05

seems to agree that search

45:07

is bad now. I don't know,

45:09

do you or? Yeah,

45:11

I mean, I said last week that

45:13

I've moved my searching over to chat GPT

45:16

in a real way, in a way

45:18

that surprised me. So I do think that

45:20

this is definitely a moment where AI

45:22

is showing it straight against search. The one

45:24

thing I would say, we

45:26

often, thanks to the defaults that Google

45:28

has, and we're going to get to antitrust

45:30

in a moment, we

45:32

are so accustomed to typing things into

45:34

Chrome and into the search bar in

45:36

our Android, and that pulls up Google

45:38

searches that as long as they're able

45:40

to keep those defaults, they're going to

45:43

be fine. But they may not. And

45:45

that's where things get interesting. Yep.

45:47

No, no, I think that's a

45:49

good point. And again, last week we

45:51

said the web is dead and

45:53

then toned it down to the web

45:55

is in secular decline. But

45:57

I still believe the way

45:59

search works on the internet overall

46:01

and specifically for Google and

46:04

the way it drives traffic to

46:06

websites is forever changed.

46:08

And I think like those interactions are,

46:10

it's already kind of, it's been

46:12

dead for a while in my mind.

46:14

And I think we're seeing how

46:16

it's changing constantly. By

46:18

the way, I mentioned the Discord. So

46:20

for those who are interested, I'm going to

46:23

drop a discounted link to big technologies

46:25

paid subscriptions. If a paid subscriber,

46:27

you're welcome to join the Discord and speak

46:29

with me and Ron John. We talk

46:31

about AI. all the time. It's

46:34

a running daily conversation and I think it's

46:36

gotten really good. A lot of really smart people

46:38

talking about where AI is heading. So I'll

46:40

put a discount link in the show notes. Please

46:42

do sign up if you're interested in joining.

46:44

It would be great to have you there. If

46:48

you sign up as a paid subscriber,

46:50

I'll send an email out early next week

46:52

with a Discord invite. So please consider

46:54

doing that and help support the show. Speaking

46:56

of which, let's take a break

46:58

to hear from one of our advertisers.

47:00

And then when we come back,

47:02

we're going to talk about this very

47:05

interesting integration between Shopify and ChatGPT, and

47:07

then the latest in Big Tech Antidress.

47:09

Back right after this. Introducing

47:12

the HP AI PC,

47:14

your powerful AI assistant.

47:16

Easily search through personal

47:18

files, gain valuable insights,

47:20

and make smarter, more

47:23

informed business decisions. Unlock

47:25

the future of work today

47:27

with the HP AI PC,

47:29

powered by the Intel Core

47:31

Ultra Processor. With the right

47:33

tools, work doesn't have to feel like work.

47:36

Learn more at

47:38

hp.com slash ai .pc.

47:44

And we're back here on Big Technology,

47:46

podcast Friday edition, breaking down all

47:48

of the week's news. Something under the

47:50

radar that's worth discussing is that

47:52

it looks like Shopify is going to

47:54

do some embedding within ChatGPT. Now,

47:56

I don't know if this is confirmed,

47:58

but it was reported all over

48:00

that there are some new, this is

48:02

from Twitter user, Aaron Rubin,

48:04

or ex user, Aaron Rubin. There

48:07

are new code strings in ChatGPT's

48:09

public web bundle, including by now,

48:11

price and shipping fields, product offer

48:13

ratings, and a Shopify checkout URL,

48:15

which indicate that OpenAI is wiring

48:17

a native purchase flow within the

48:20

Assistant. So you could basically buy

48:22

directly within ChatGPT as opposed to

48:24

having it send you out to

48:26

a website. This seems natural. I

48:28

wrote to Shopify to try to

48:30

get some confirmation. I did

48:32

not hear back. Let's speculate. What

48:34

does this mean, Rajan? I

48:36

think it's important. We've

48:39

already seen, though, that perplexity has,

48:41

if you're a perplexity pro subscriber, they

48:44

have a checkout within app

48:46

shopping feature where you can

48:48

go through the entire flow.

48:51

That moving into chat GPT, I think,

48:53

is definitely important. I think the

48:55

fact that Shopify seems to be trying

48:57

to take a first mover advantage

48:59

in this is also important from their

49:01

side. I really wonder,

49:04

though, Do you,

49:06

are people going to shop within a

49:08

chatbot? And I think the entire retail

49:10

industry has been wondering this for a

49:12

while as well. Like, is that experience

49:14

of asking a question and being shown

49:16

a few products and then maybe asking

49:18

more questions about the product? Is that

49:20

how people want to shop? Because we've

49:23

been so conditioned to. browsing

49:25

and like scrolling through and clicking

49:27

through products to product pages and

49:29

then going back. And like that

49:31

is how people shop. And it's

49:33

a, it's not like such a

49:35

targeted direct thing. Like if you're

49:37

buying toilet paper on Amazon or

49:39

something like that, it's more of

49:41

an experience. So I guess the

49:43

way I would think about is

49:45

like, It's one thing to like

49:47

go to a mall and walk around and browse

49:49

versus having a personal shopper that you just

49:51

talked to while you're sitting at your desk and

49:53

they go out and buy stuff for you. But

49:56

yeah, I think it's there is

49:58

definitely a large universe of transactions that

50:01

will work in this way. And

50:03

this is going to happen. I do

50:05

believe that just if it is

50:07

this the predominant way people shop, I

50:09

don't know. I think it

50:11

could be. And it's going to sound crazy.

50:13

But let me give you a couple data

50:15

points here. So first of all, when you're

50:18

shopping on Amazon Prime, people

50:20

have become conditioned to just basically take Amazon's

50:22

choice and buy it. And that is because

50:24

they've had enough trust and enough positive experiences

50:26

within Amazon that they believe that they're going

50:28

to get the best deal on the internet

50:30

when they're on Prime and they don't need

50:32

to go to too many sites. I think

50:34

it's become a natural behavior. Now,

50:37

when you trust ChatGPT, when you're,

50:39

let's say you're married to ChatGPT,

50:41

no, just kidding, but But

50:44

let's say you're talking to, I mean, maybe you are.

50:47

When you're talking to ChatGPT. judgment.

50:49

No judgment. Do your thing. Get

50:51

married to ChatGPT and then go shopping with

50:53

it. Buy nice things. shopping with it. Because the

50:56

joke is going to be on the rest

50:58

of us. But

51:00

when you have such a deep relationship with ChatGPT,

51:02

what are you going to do? You're going to

51:04

trust what it says the same way that you

51:06

trust Amazon Prime. and that trust

51:09

is going to make you want to, instead

51:11

of going to other websites, just buy

51:13

right within chat GPT, that is going to

51:15

become a default behavior for lots of

51:17

people. It does look like they're building this

51:19

and all of a sudden shopping on

51:21

the web in the way that you describe

51:23

where you go page to page and

51:25

then make a decision after reading the reviews,

51:28

it's going to seem archaic. ChatGPT is

51:30

going to bring everything within the chatbot, show

51:32

you the reviews, show you the different

51:34

customer experiences, maybe even show you a video,

51:36

show you how the product looks in

51:38

your house, show you how the clothes looks

51:40

on your body, show you how the

51:42

watch looks on your wrist, show you how

51:44

the appliance looks in your kitchen, and

51:47

you will trust it and you will buy

51:49

from it. End of story. Take

51:51

it to the bank. Good God, I'm

51:53

sold. My God, are you? Do

51:56

you have a side startup going on and

51:58

running this? Because that was the greatest pitch I've

52:00

on this topic. I think

52:02

I'm bought in. I'm in. As

52:05

you can see, everything that has

52:07

been displayed on my virtual shelves

52:09

comes directly from ChatGPT. I'm just

52:11

kidding. No financial stake here, but

52:13

it does seem to me like

52:15

it's going to be a thing.

52:19

And I am curious what that means for Amazon.

52:21

I am going to have the head of Amazon

52:23

Prime on the show in a couple of weeks.

52:25

So maybe that's a question for is a good

52:27

topic. And then how you

52:29

get into that conversation is

52:31

becoming a bigger and bigger topic,

52:33

I think, for all retailers.

52:36

Because again, SEO or search engine

52:38

optimization was how people got

52:40

their products discovered for the last

52:42

20, 25 years and became

52:44

like the most mature. industry

52:46

and now this changes everything. Like

52:49

how do I show up in perplexity

52:51

results? How do I show up

52:53

in, uh, in chat GPT

52:55

results? My favorite part of this and I,

52:57

I'm going to, I'm going to throw

52:59

a couple of names by you because in

53:01

the space that I'm like pretty deeply

53:03

in this right now, no one has agreed

53:05

on what this new world is called.

53:07

We have SEO is a classic term, but.

53:10

A couple of different options. GEO,

53:12

Generative Engine Optimization.

53:15

GAIO, Generative AI

53:17

Optimization. AAO,

53:20

AI Agent Optimization.

53:22

SGE, Search Generative

53:24

Experience. AIO, AI

53:26

Overview Optimization. And

53:28

last, LLMO, Large

53:30

Language Model Optimization. What

53:32

are you going with, Alex? I'm

53:34

going with I'm Angry. at the

53:37

fact that some have even

53:39

been advanced in these discussions. Conversation.

53:41

Let me start by striking

53:43

the ones that I find hideous.

53:45

Let's go. L -L -M -O,

53:47

take a hike, you're gone. You're

53:49

gone. It sounds terrible. It

53:51

sounds like a I agree. G

53:54

-A -I -O. Gone.

53:56

Awful. Yeah. Awful.

54:00

AIO sounds like an insurance company. You're

54:02

gone. It might be. It actually

54:04

might be. Yeah. Are you protected for

54:06

anything that might happen to your

54:08

family try AI? Oh AI

54:10

overview optimization All right, that's not

54:12

so what do you like

54:14

I'm into Geo. It's it's like

54:17

SEO. It's it's gonna stick

54:19

It's one letter generative engine optimization

54:21

Now engine is a little

54:23

weird because we don't really say

54:25

anything like a generative engine

54:27

like we say search engine But

54:29

it's close to SEO people

54:31

get it. It's gonna be Geo

54:33

or Geo And I

54:36

think if what I talked

54:38

about with retail becomes a

54:40

thing that you shop within

54:42

ChatGPT, then GEO is gonna

54:44

be a massive field. You

54:46

gotta figure out your GEO

54:48

strategy ASAP because you gotta

54:50

get those results when we're

54:52

all married to ChatGPT and

54:54

shopping with it. For

54:57

it as well, potentially. Yeah,

54:59

exactly. The way to get AI on

55:01

your side, buy it, nice things. Buy

55:03

nice things. It's time to

55:06

improve your strategy. Would

55:08

you get an AI that you're

55:10

in love with just like its own

55:12

set of GPUs? You'll never be

55:14

tired again. You'll never feel

55:16

exhausted. Showing my love for

55:18

you, I'm buying you this network

55:20

server rack from NVIDIA. That's

55:24

NVIDIA's new market. At

55:27

the Valentine's Day, say I love you

55:29

to the robot in your life with

55:31

an Nvidia server rack. It's like a

55:33

little decked out. It's a little like

55:35

the chips are just like the wiring's

55:38

a little nicer. It's a, yeah, I

55:40

think, I mean, what else?

55:42

I don't know. That seems to be

55:44

the most relevant purchase that would make

55:46

it happy. Crazier things have

55:48

happened. Yeah,

55:50

Nvidia, it's your new growth

55:52

strategy. You're doing pretty well, but just think

55:55

about your five year plan. Jensen,

55:57

I hope you're listening to this. We're serious.

55:59

We're very serious about this. Okay,

56:02

so speaking of chapters of love and

56:04

hate, we had a very interesting moment

56:06

happen here in Washington DC this week.

56:09

So Metta, of course, is on

56:11

trial in an antitrust case

56:13

and who shows up, but Kevin

56:15

Systrom, the co -founder of Instagram,

56:17

who famously sold Instagram for

56:19

a billion dollars to Facebook back

56:21

in the day, he

56:23

comes in and testifies for the

56:25

prosecution. And he says, basically

56:27

Mark was not investing in Instagram

56:29

because he believed it was

56:31

a threat to their growth, their

56:33

meaning, Facebook's

56:35

growth. And Facebook apparently

56:37

had this buy or bury strategy,

56:39

which is basically you buy

56:41

the company or you try to

56:43

destroy them. And people

56:46

are saying that what they did to Instagram

56:48

was they bought and buried it. And

56:50

this is what Sistrum says, we were by

56:52

far the fastest growing team. We produced

56:54

the most revenue and relative to what we

56:56

should have been at the time. I

56:58

felt like we should have been much larger.

57:01

And so, oh, he also talks about

57:03

Zuckerberg's emotion. He says, as the

57:05

founder of Facebook, he felt a lot

57:08

of emotion around which one was

57:10

better, meaning Instagram or Facebook. And I

57:12

think there were real human emotional

57:14

things going on there. Basically, Zuckerberg was

57:16

so tied to Facebook that he

57:18

hurt Instagram in service of trying to

57:20

make Facebook better. Let me

57:22

put out the counter argument here and.

57:24

get your reaction. I

57:27

get you, Kevin. I

57:29

hear what you're saying, but if you look at

57:31

who ended up winning, Instagram ended

57:33

up winning. Instagram is the

57:35

app. Whatever Facebook did worked.

57:37

It's massive. It is,

57:39

I think, more used, maybe

57:41

not in sheer user numbers, but

57:44

certainly it's more culturally relevant

57:46

than the Blue app, and

57:48

it will outlast Facebook despite

57:50

Mark Zuckerberg's emotional attachment to

57:52

the latter. And so therefore,

57:55

I hear your testimony. However,

57:57

to me, it is not meaningful here,

57:59

even though Facebook may lose. It

58:02

was interesting to see your

58:04

perspective, but ultimately, I don't

58:06

think it really changes what

58:08

the court is going to rule

58:10

because it doesn't hold water

58:12

when you look at the results.

58:14

What do you think, Ron?

58:17

I actually completely agree. I'm a

58:19

strong believer that A

58:21

lot of what Meta has

58:23

done and has become is

58:25

definitely from an antitrust perspective

58:27

problematic. However, this

58:29

specific example, it

58:31

probably started, well, if we separate

58:33

it, it could have definitely started.

58:35

There's been a lot of communication

58:37

that makes it feel that it

58:39

was a buy or bury type

58:41

action at the time. But

58:44

yeah, by 2018, Facebook

58:47

had so deeply integrated

58:49

Instagram into the Facebook experience

58:51

to grow it. I

58:54

remember vividly, like 2015 -ish, starting

58:56

to see a lot of

58:59

non -tech or social media

59:01

forward friends all showing up

59:03

suddenly because they were getting

59:05

Facebook notifications or accidentally cross -posting.

59:08

Like they, I mean, he had brought

59:10

up how it was growing yet

59:12

they only had a thousand employees compared

59:14

to 35 ,000 employees at Facebook. But

59:16

you don't need those that many

59:18

employees because it was the engine of

59:20

Facebook that was driving the growth.

59:22

So yeah, on this one, I

59:24

do not agree that that is

59:26

the, that's the thing that's going

59:28

to move the needle in terms

59:31

of like, has Facebook behaved problematically?

59:33

I do love that he goes

59:35

after like. Zuckerberg's emotion

59:37

here. I mean I'm feeling a bit

59:39

cage -matchy between these two Kevin and

59:41

Mark on this one because like

59:43

to be like you were just jealous

59:45

that we were growing and you

59:47

weren't so you didn't give us resources

59:49

Especially because that's not what was

59:51

happening. So to still call them out

59:53

on that I kind of I

59:55

kind of want to see if we

59:57

get a reaction from Zuck on

59:59

if I get a threads notification on

1:00:01

this one you could and also

1:00:03

Just thinking about this a little bit

1:00:05

more deeply, you look

1:00:07

at Facebook's marquee acquisitions, Instagram

1:00:09

and WhatsApp, they're doing great. I

1:00:12

mean, they're doing better than Facebook Blue. WhatsApp

1:00:14

and Instagram are the future of this company. I

1:00:17

think at a certain point,

1:00:19

maybe I could be totally

1:00:21

wrong on this, but it does

1:00:23

feel like from a product

1:00:25

development standpoint, from just like

1:00:27

a quality of utility standpoint, I

1:00:30

don't want to say they gave up on Blue, but

1:00:32

like they're just kind of like, yeah, whatever. people

1:00:34

are going to still stick around some

1:00:36

number of people and it'll just kind of

1:00:39

degrade and the content. And they're going

1:00:41

to stick around there. But to make beautiful

1:00:43

products to get more interesting and better,

1:00:45

let's work on Instagram and WhatsApp. That's what

1:00:47

it feels like from the outside, at

1:00:49

least. Definitely. Now

1:00:51

I'm going to drop the however. However

1:00:53

is. However. However. So

1:00:55

I'm in DC this week for Semaphore's

1:00:57

World Economic Summit. I was able to

1:01:00

interview the CEO of Altice USA. Dennis

1:01:02

Matthew was interesting conversation. We're

1:01:04

gonna put it up on YouTube

1:01:06

just about 15 minutes or

1:01:08

so so brief but being here

1:01:10

enabled me to get a

1:01:12

chance to spend time with the

1:01:14

Washington DC creatures and The

1:01:16

vibe here is that we're gonna

1:01:18

see breakups very likely of

1:01:20

Google and potentially of Facebook and

1:01:22

the difference between Facebook

1:01:25

and Google is that, I

1:01:27

mean, Google's lost its antitrust cases,

1:01:29

but Google knew antitrust was

1:01:31

coming and was pretty buttoned up

1:01:33

in terms of its disclosures

1:01:36

and didn't basically, didn't have like

1:01:38

damning emails, you know, come

1:01:40

out in the case. Whereas Facebook

1:01:42

had no idea that this

1:01:44

would happen to it. And you're

1:01:46

seeing all these emails from

1:01:49

Zuckerberg spelling out this like buyer

1:01:51

-berry strategy and he got caught.

1:01:53

So even if you could

1:01:55

say that The acquisitions haven't been

1:01:57

bad for competition. It's pretty

1:01:59

rough to see all this really

1:02:02

damning information about the way

1:02:04

that Facebook operated come out in

1:02:06

court. And when you're in a court,

1:02:09

sometimes those emails can sway a judge. And

1:02:12

Facebook could very well lose this

1:02:14

case the same way that Google

1:02:16

lost its cases. Google

1:02:19

for one is running out of

1:02:21

appeals. I think Google can appeal

1:02:23

the first case to the Supreme

1:02:25

Court and that's it and Then

1:02:27

we see them. We go to

1:02:29

the remedy phase. So Very interesting

1:02:32

moment for big tech. They don't

1:02:34

have a lot of friends in

1:02:36

DC despite the money they've spent

1:02:38

From what I understand the administration

1:02:40

hates Facebook really really hates Facebook

1:02:42

and Despite Zuckerberg going to see

1:02:44

Trump. It doesn't seem like Trump

1:02:46

is gonna Back off the

1:02:48

heat at all here. So

1:02:50

could be a very interesting like

1:02:52

regulation has been back burner

1:02:55

for us But could we see

1:02:57

breakups? I think the chances

1:02:59

are higher than Then I would

1:03:01

have ever imagined even a

1:03:03

couple months ago You don't get

1:03:05

many bipartisan bipartisan efforts or

1:03:07

beliefs and this certainly seems to

1:03:09

be the one I think

1:03:11

this the interesting part from like

1:03:13

the legal perspective is and

1:03:16

like related to Kevin's testimony, is

1:03:18

it intent? Cause there's no doubt

1:03:20

in my mind. And I think the

1:03:22

emails all show that very clearly.

1:03:24

The goal was to remove competition from

1:03:26

the market. Like that was the

1:03:28

goal. What you do with

1:03:30

it after, do you integrate it

1:03:33

tightly with your existing product and make

1:03:35

it potentially your marquee product? Or

1:03:37

do you just sunset it and

1:03:39

kill it off? That's after the fact.

1:03:41

Like the goal was to remove

1:03:43

competition. But

1:03:45

the fact that they did not

1:03:47

end up killing Instagram, and

1:03:50

now it's a huge,

1:03:52

gigantic, influential product, is

1:03:54

that enough to say like, yeah, I said

1:03:56

buy and bury at the time, but look,

1:03:58

we didn't bury it. We bought it and

1:04:00

it's flourishing. Is that enough?

1:04:03

I'm not a lawyer, so I will

1:04:06

not be able to understand that. Yeah,

1:04:09

and I think one last point about

1:04:11

this. The earth is changing

1:04:13

beneath these companies feet. It's like this

1:04:15

is the last battle, like we spoke about

1:04:17

last week. And now some

1:04:19

of the things that you would do in

1:04:21

these apps, you're going to spend time talking

1:04:23

to AIs instead of your friends. And

1:04:26

so even if it had

1:04:28

given the company a short term

1:04:30

competitive advantage, or even let's

1:04:32

say the Department of Justice ends

1:04:34

up splitting double click or

1:04:36

Google's ad network off of Google,

1:04:39

it's not to make a big difference, I

1:04:41

think. What matters now is the battle

1:04:43

of today and that battle is artificial intelligence.

1:04:45

Thank you to the conscious

1:04:47

robots and large language models

1:04:50

that we cannot interpret for

1:04:52

bringing competition to the market

1:04:54

after about 12, 13 years,

1:04:56

maybe 20. 100%.

1:04:58

Well, thank you everybody for listening. Remember,

1:05:00

if you want your AI to love

1:05:02

you back, buy it some server racks. That's

1:05:05

all they want. Man.

1:05:07

If that happens and we put

1:05:09

the product links and some affiliate

1:05:11

codes, 5 % of

1:05:13

$100 ,000 Valentine's Day presents,

1:05:16

not bad. Not a

1:05:18

bad model. That's our future business model here. I

1:05:20

think we're finding it on the fly. All

1:05:23

right, Ron, John, great to see you. Thanks so much for coming on.

1:05:25

All right, see you next week. See you

1:05:27

next week. And thank you, everybody, for listening. We'll

1:05:29

see you next time on Big Technology Podcast.