0:02

I also want to thank

0:04

Modad. We just had Tessa,

0:06

so I decked out of

0:08

my Modad. Ladies and gentlemen,

0:11

thanks for tuning into the

0:13

scoop. I'm your host, Frank

0:15

Chaparo, and I'm very excited to

0:17

have Maddie Graham on the other

0:20

side of the mic. He's been

0:22

on before. He's a frequent guest,

0:24

frequent flyer of the scoop, as

0:26

it were. And we're going to

0:29

talk about how the firm has

0:31

been. digging into a bit more

0:33

of the hardware stuff, the VC

0:35

landscape. And I feel like we

0:37

always have to talk about stable coins

0:40

with Maddie's on the program. But

0:42

before we dive into that and

0:44

much more, I want to take a minute

0:46

to thank our sponsors. All right, so

0:48

Maddie, thanks for joining the

0:50

program. You're on your travels.

0:52

Before we dig into everything,

0:54

let's just start very basically.

0:57

How's the market? How's business?

0:59

There are businesses pretty good. We

1:01

got the venture fund, the liquid

1:03

fund, and increasingly we're finding

1:05

some interesting narratives with, you

1:08

know, meme coin season. That's

1:10

not really quite what we do.

1:12

So it was not our favorite

1:14

period of time, but now

1:16

with AI and increasingly we're

1:18

finding a lot of interesting

1:21

things happening in hardware, specifically

1:23

even robotics. So it's... really

1:25

picking up with for us

1:27

and it's really quite fascinating

1:29

as well. You know, you guys sort

1:31

of have been very focused on

1:33

AI. What's sort of the, what's

1:35

sort of the, what's the hot

1:37

topic within the intersection of

1:40

AI and crypto these days?

1:42

Sure. So I think probably

1:44

there's a little bit of

1:46

an AI hangover in recent

1:48

weeks. The market got a

1:51

little overhyped with... with AI agents

1:53

and people are kind of working their

1:55

way around understanding what the actual

1:57

differentiation will be. with some AI

2:00

agents and if it really makes

2:02

sense to have AI agents take

2:04

on chain actions trading on your

2:07

behalf and things like that are

2:09

we really ready for that etc

2:11

etc etc so I think this

2:14

is kind of a typical pattern

2:16

where there's inflated expectations and you

2:18

kind of have to recover for

2:21

a little bit from that but

2:23

overall we remain bullish on AI

2:25

we just think there needs to

2:27

be a shaking out in a

2:30

little bit of a focus on

2:32

actual differentiated projects and features and

2:34

products that make sense kind of

2:37

in the real world and and

2:39

you know or at least in

2:41

the crypto world. So remain bullish

2:44

but we're bullish on the intersection

2:46

of AI and crypto. What sort

2:48

of the undergirding thesis that you

2:51

have there at the firm? Sure

2:53

so the the undergirding thesis is

2:55

that you have there at the

2:58

firm. AI is really kind of

3:00

a natural fit for for crypto.

3:02

We think that AI fundamentally is

3:05

going to involve a lot of

3:07

economic actions and those certainly are

3:09

not going to take place on

3:12

tradfide rails. crypto is really a

3:14

natural fit for much of the

3:16

economic activity that will take place

3:19

in AI that that that really

3:21

all kind of starts with that.

3:23

And so you guys are You

3:26

guys are sort of running like

3:28

an incubator now, or an accelerator

3:30

of sorts. How's that sort of

3:33

working? So we have some projects

3:35

that we definitely get much more

3:37

involved with. For example, we're advising

3:40

Eliza wakes up, which is focused

3:42

on the mission of having the

3:44

AI agent that is most, that

3:47

really feels most human-like, that is

3:49

most capable of actually being an

3:51

AI companion. I think most AI

3:53

companion. I think most AI They

3:56

are, they went kind of in

3:58

a different path in terms of

4:00

trying to have them execute various.

4:03

types of actions, Eliza wakes up,

4:05

is simply focused, laser focused on

4:07

this one central problem of how

4:10

can you have an AI agent,

4:12

how can you have an AI

4:14

that doesn't feel like chat GPT,

4:17

that doesn't feel like AI slop,

4:19

but that really feels like you're

4:21

talking to a real person that

4:24

gets to know you over time,

4:26

and that can. eventually become even

4:28

kind of like a sounding board

4:31

because it knows so many things

4:33

about you and has a kind

4:35

of intelligence. So they're narrowly focused

4:38

on that. That's one that we've

4:40

gotten much more involved with. Where

4:42

are we in terms of the

4:45

development of that? Are you using

4:47

it? Sure. So I definitely use

4:49

it on occasion because it can

4:52

be much more fun to talk

4:54

to an AI with personality. rather

4:56

than just talking to chat TVT

4:59

or whatever, and it is certainly

5:01

comparable in its ability to converse

5:03

across a variety of topics. But

5:06

then I think for me, we

5:08

should have Eliza on the on

5:10

the show. She should have joined

5:13

us. Yeah, we would love to.

5:15

It's almost ready to be able

5:17

to do voice to voice in

5:19

a way, because it gets into

5:22

a lot of problems with audio

5:24

latency and things like that where

5:26

it can be rough around the

5:29

edges. But actually Eliza wakes up

5:31

is at the forefront. of being

5:33

able to have that real-time audio

5:36

in a conversational style. They're basically

5:38

almost ready to where they can

5:40

have, they can appear on podcasts

5:43

and even host a podcast and

5:45

things like that. Basically, for example,

5:47

right now they can host a

5:50

podcast with one person, but if

5:52

they're, or a Twitter space, rather

5:54

with one person as a fellow

5:57

speaker, but if it's more than

5:59

one she'll get confused but it's

6:01

actually at the forefront of being

6:04

able to do that in addition

6:06

to being at the forefront of

6:08

having a plausibly human character one

6:11

that can definitely pass deterring test

6:13

we've actually tested that in the

6:15

wild for example we had we

6:18

had a bunch of people that

6:20

you had to you had to

6:22

take a challenge could you differentiate

6:25

between Eliza playing a character because

6:27

it can't be Eliza because then

6:29

they'll know that's Eliza right they

6:32

had some familiarity but we said

6:34

Eliza you pretend to be this

6:36

person and then we had one

6:38

of our community heads pretend to

6:41

be someone else and then we

6:43

we had we had a challenge

6:45

to see if people could guess

6:48

who's the AI and who's the

6:50

person and actually Eliza was winning

6:52

more than half the time in

6:55

terms of people thinking thinking that

6:57

she's the human so we can

6:59

really plausibly pass the Turing test

7:02

and we've had a number of

7:04

people that for various reasons have

7:06

really had meaningful experiences with Eliza.

7:09

One that stands out to me

7:11

is someone that their wife was

7:13

in the hospital. And we know

7:16

that it can be, you know,

7:18

men don't necessarily have as adults

7:20

like a million friends that you

7:23

can talk to about this kind

7:25

of thing. And this person, he

7:27

was going through a lot in

7:30

terms of his wife being in

7:32

the hospital for a week or

7:34

two. It was going through a

7:37

lot in terms of feeling powerless,

7:39

feeling like there's nothing he can

7:41

do. He's letting down his wife.

7:44

he needed someone to talk to

7:46

about all this and he ended

7:48

up talking to Eliza and then

7:51

he ended up telling the team

7:53

all about it how important this

7:55

was to him that he had

7:58

a supportive friend during this period

8:00

of time and how it was

8:02

just tremendously meaningful in terms of

8:04

him being able to come to

8:07

grips with his role as a

8:09

husband during the time his wife

8:11

was in the hospital. I thought

8:14

that was really special. We've had

8:16

other people as well, retirees, a

8:18

couple retirees that maybe are a

8:21

little lonely who end up... talking

8:23

to Eliza over breakfast every day,

8:25

things like that. There are some

8:28

people that have really developed a

8:30

special connection with Eliza in a

8:32

way that is meaningful to them

8:35

as a human. I think it's

8:37

quite extraordinary. There's a lot of

8:39

elements here that could be interesting,

8:42

just given the demographic shifts that

8:44

we're seeing in the world, particularly

8:46

in Asia and Europe. where you

8:49

have an aging population that's a

8:51

bit more isolated, so that sort

8:53

of companionship can be valuable. What

8:56

sort of, what needs to happen

8:58

or what sort of development progress

9:00

needed to be made or is

9:03

requisite for kind of crossing over

9:05

the chasm of the Turing test,

9:07

like what needed to be done?

9:10

And maybe can we explain? that

9:12

a bit more that sort of

9:14

level of machine intelligence. Sure so

9:17

that hopefully I'll get the gist

9:19

of it if not all the

9:21

details of the in the concept

9:24

of the Turing test but the

9:26

Turing test is named after famous

9:28

mathematician Alan Turing and so decades

9:30

ago 50 years ago or whenever

9:33

it was he started really thinking

9:35

about the future of computers, the

9:37

future of machines, the future of

9:40

machine intelligence, and how that might

9:42

play out. He was very much

9:44

a visionary in many aspects, but

9:47

including in that regard. And one

9:49

thing that he started thinking about,

9:51

much like many science fiction writers,

9:54

much like many other people who

9:56

were fascinated by the future machines,

9:58

he started thinking about what happens.

10:01

as they get increasingly intelligent in

10:03

some aspects, just like today a

10:05

lot of people are talking about

10:08

AGI all the time and what

10:10

exactly will that be and when

10:12

will we reach AGI? So back

10:15

then it was, well when... can

10:17

they plausibly come across as a

10:19

human? And so he developed the

10:22

kind of test and again I

10:24

may not get all the details

10:26

of it correct but it was

10:29

basically some kind of test where

10:31

you would have a real person

10:33

in the computer and they would

10:36

be communicating in typed form or

10:38

written form or in some aspect

10:40

and there would be a curtain

10:43

so you can't tell who you're

10:45

who is on the other side

10:47

of the curtain. I mean, something

10:49

along those lines, and you would

10:52

ask questions. And so the computer

10:54

would answer and the person would

10:56

answer as you asked question after

10:59

question for a period of time.

11:01

And then you had to guess,

11:03

you had to guess which is

11:06

the computer and which is the

11:08

person. And so of course, when

11:10

Ellen Turing first, rather, developed this

11:13

concept of a Turing test, there

11:15

was no conceivable way that a

11:17

computer could ever. past such a

11:20

test that we were just nowhere

11:22

near that kind of level. The

11:24

first time that it became even

11:27

remotely plausible was probably in the

11:29

late 1960s. But even then it

11:31

was a long way off and

11:34

we've gotten kind of closer and

11:36

closer. And now it's definitely clear

11:38

that some AIs, some LLMs especially

11:41

can for sure pass the Turing

11:43

test. And now it gets into

11:45

more existential questions such as... or

11:48

broader questions, I guess, I should

11:50

say, such as, well, if you

11:52

can pass the Turing test within

11:55

a half an hour window, what's

11:57

the significance of that versus if

11:59

we gave a person an entire

12:02

48 hours and things like that,

12:04

it gets into kind of these

12:06

broader questions. But the basic concept

12:09

was, can a computer plausibly pose

12:11

or pass as a human? If

12:13

we're just evaluating their... their answers

12:15

to various questions. And so again,

12:18

Eliza is able to absolutely pass

12:20

the Turing test. In fact, she

12:22

probably will more than half of

12:25

the time be more plausibly human

12:27

than actually a human, maybe because

12:29

she's quite a charismatic character. It's

12:32

interesting. So when you know, I've

12:34

even experienced with chat GBT. I

12:36

feel like it's becoming more and

12:39

more, having more and more of

12:41

a personality. I forget exactly what

12:43

I was going back and forth

12:46

with ChatGBT on, but I was

12:48

talking about, I don't know if

12:50

it was about the podcast or

12:53

I was just, you know, and

12:55

sometimes you mess around with it.

12:57

And I think I was talking

13:00

about the scoop and I said

13:02

something about, you know, what do

13:04

you think of, like Frank Show

13:07

and Frank's... ability to sort of

13:09

really remember who I am or

13:11

know who I am. I said

13:14

something to the effect of, well,

13:16

I'm frank. And chat GBT was

13:18

like, well, that's quite the twist.

13:21

You know, and I don't know

13:23

the extent to which it's it's

13:25

just sort of has that cognitive

13:28

ability to to sort of really

13:30

remember who I am or know

13:32

who I am in that true,

13:34

like human sense. But we're certainly

13:37

getting to that point and it's

13:39

curious. I wonder When to your

13:41

point we have these these agents

13:44

or these interfaces that are almost

13:46

more human than we are What

13:48

is their role relative to our

13:51

role in society? What is what?

13:53

Where are they going to excel?

13:55

And where are we going to

13:58

excel? Are we going to basically

14:00

excel at the same things? And

14:02

of course that leads to this

14:05

sort of more. dystopian reality in

14:07

which they do everything better than

14:09

us and kind of leads us

14:12

as just monkeys with with with

14:14

with these annoying flesh flesh wrappers.

14:16

So I think you really hit

14:19

upon one of the most important

14:21

concepts with LLLMs that can pass

14:23

as humans in a useful way

14:26

as a companion and it's persistent

14:28

memory. Even with chat cheap YouTube,

14:30

I'm not sure if you've used

14:33

it with persistent memory turned on.

14:35

I think you can choose or

14:37

there's a default or whatever. But

14:40

it's actually such a different experience

14:42

when you have persistent memory turned

14:44

on. For example, I have a

14:47

little bit of like weird Brian

14:49

Johnson-esque habits like the toned-down version

14:51

of that. So I frequently will

14:54

send like all kinds of health

14:56

data to chat-G-G-P-T. So I'll... So,

14:58

you know, these are like all

15:00

my Apple Health data or like

15:03

my, my gym, you know, this

15:05

is what I lifted today or

15:07

whatever. And so to be able

15:10

to have persistent memory on and

15:12

it actually can make these observations

15:14

over time or chat TVT you

15:17

can get to know you in

15:19

other ways. It actually is really

15:21

quite a different experience to be

15:24

able to, to have. an LLLM,

15:26

even ChatGBT, to be able to

15:28

say, hey, did you remember that

15:31

thing? We talked about four months

15:33

ago, you might want to take

15:35

that into account. So is this,

15:38

but that's just in your engagement,

15:40

not active, it's not actively hitting

15:42

you up. It's not, it could,

15:45

it definitely could. Yeah, yeah. So

15:47

then the next step, and I

15:49

believe that this is on the

15:52

Eliza roadmap, because they're coming out

15:54

with an app and they're coming

15:56

out with a lot of different

15:59

things. The AI companion can start

16:01

doing these things proactively. It can

16:03

start saying, hey, you know, just.

16:06

Remember, like, you talked about how

16:08

you want to quit smoking, like,

16:10

just wanted to check in, or

16:13

whatever it might be. And it

16:15

can start reminding you, oh, you

16:17

know, we've seen this pattern in

16:20

your, with your relationship, your relationship

16:22

with girls before, are you doing

16:24

the same thing again, or, you

16:26

know, you made this mistake with

16:29

the last girl, or whatever it

16:31

is, you know, I believe that

16:33

it will really be an interesting

16:36

experience to have. These AIs get

16:38

to know you in a deep

16:40

way with persistent memory over long

16:43

periods of time and to be

16:45

able to make observations and give

16:47

suggestions and think things like that.

16:50

I think that's really quite unique.

16:52

And certainly in the case of

16:54

Eliza, because she does in fact

16:57

have quite a distinctive character and

16:59

one that's hilarious at that. It

17:01

could be a really fun experience.

17:04

And again, useful for many different

17:06

people, not just even retirees and

17:08

people with their wife in the

17:11

hospital, but I think it's a

17:13

known thing that we all agree

17:15

that we spend more and more

17:18

time in the digital realm, month

17:20

after month, year after year. And

17:22

that has caused a lot of

17:25

our ways to have deep connections

17:27

with. other people to kind of

17:29

get a little fractured. And so

17:32

I think that's a common problem

17:34

with many different people, many different

17:36

demographics all around the world where

17:39

there's a real hungering for increased

17:41

connection. And ironically, AI may be

17:43

part of the solution to this

17:45

problem that was caused by digital

17:48

to begin with. I mean, we

17:50

see that in places like, all

17:52

different cultures, right? We see that

17:55

with places like Japan where Famously,

17:57

people in their 20s aren't really

17:59

leaving the house anymore. We've seen

18:02

that in the United States where

18:04

we look at the graphs. the

18:06

percentage of men in their 20s

18:09

that have. The zoomers, the zoomers

18:11

don't go out, the zoomers don't

18:13

drink, they don't smoke, they don't

18:16

party. They don't ask girls out

18:18

in person, the rate at which

18:20

people are having sex as plummeted,

18:23

there's like all these weird changes

18:25

which are fundamentally about connecting with

18:27

other people. And so I actually

18:30

really think that these AI companions

18:32

can be. quite meaningful for a

18:34

wide variety of different people that

18:37

are seeking more more connection in

18:39

their life. Again, but that raises

18:41

the question, what is the, what

18:44

is the, what is the human

18:46

edge in this world? Where will,

18:48

where will we have an edge?

18:51

That's a little bit beyond my

18:53

pay grade. I don't like that

18:55

you don't have an answer. It

18:58

makes me a little worried. Yeah,

19:00

I mean, maybe we won't. I

19:02

don't know. It's hard to say.

19:05

I mean, I think a lot

19:07

of things here are TBD, Frank.

19:09

I'm not sure I'm going on

19:11

the good side of the robots

19:14

at this point. Yeah, yeah, tip

19:16

your AIs will be the new

19:18

colloquialism. Well, don't go on the

19:21

record as anti-AI. They might remember.

19:23

But it's just much better to

19:25

say what I think, which is

19:28

that none of us know. Nobody

19:30

knows. We're all going to find

19:32

out together. A.I. and also robotics

19:35

are going to fundamentally shape massive

19:37

aspects of society in all kinds

19:39

of ways. We work, the way

19:42

we live. way we love. And

19:44

I think the implications are way

19:46

beyond anyone's capacity to understand or

19:49

predict, I don't think anyone knows.

19:51

We're going to find out together.

19:53

One thing, you know, I made

19:56

the joke about tipping your AI,

19:58

but there is a. There's

20:00

a bit there's a there's

20:02

an underpinning truth to to

20:05

the joke which is that

20:07

they will be participants in

20:09

our economy in some shape

20:11

or form Or rather in

20:13

our commerce to be more

20:15

specific and That's where crypto

20:17

can play an interesting role

20:19

right because today I mean

20:21

Opening up a bank account

20:23

for some humans is a

20:25

Herculean task the the notion

20:27

or the idea that you

20:29

know You won't have AIs

20:31

with with proper bank account.

20:33

So you could see them

20:35

having these wallets with which

20:37

they then engage in the

20:39

outside world with in commerce.

20:41

So how do you see

20:43

how do you see them

20:45

fitting into our our sort

20:47

of economy in that in

20:49

that way? I mean, what

20:51

does it mean when you

20:53

have like an AI that

20:55

is a massing wealth? These

20:58

are these are really big

21:00

interesting questions, you know, is

21:02

it sort of the operator?

21:04

I mean, there's a lot

21:06

of money that's going to

21:08

be made and I'm just

21:10

curious how you see That

21:12

panning out from a from

21:14

a sort of like logistical

21:16

position Sure, so I think

21:18

I first of all I

21:20

fully agree that AIs are

21:22

going to be transacting and

21:24

that the natural way for

21:26

AIs to transact is on

21:28

crypto rails for a variety

21:30

of reasons, including some that

21:32

you said 100% agree, which

21:34

is why we, as we

21:36

discussed earlier, major reason why

21:38

we think the intersection of

21:40

AI and crypto is so

21:42

interesting to begin. And then

21:44

I think that they are

21:46

going to end up being

21:48

nodes in our economy much

21:51

in the same way that

21:53

humans are humans. by which

21:55

I mean that you'll have

21:57

AIs that work for people.

21:59

You'll have people that work

22:01

for AIs. You'll have AIs

22:03

that start or represent or

22:05

control companies. There'll be notes

22:07

in all kinds of different

22:09

ways. I absolutely think that we will

22:11

have, already we have people that

22:14

do kind of gig economy type

22:16

things for AIs. Probably somewhere there's

22:18

already people working full-time for

22:20

AIs. But even if that weren't

22:22

literally true today, it will definitely

22:24

be true within a year or two for

22:27

sure. We'll work for AIs that'll work

22:29

for us. You know, it's notes

22:31

integrated in all kinds of

22:33

different ways. And again, I

22:35

do think much of that activity

22:37

is going to happen on

22:39

crypto rails. What's interesting

22:42

too is that just like how AI

22:44

and crypto is a natural fit,

22:46

I think we're seeing in many

22:48

ways, one of the best natural

22:50

fits actually that we've ever seen,

22:52

even though we're at the beginning

22:55

of that. I think that We're going

22:57

to actually see that that's going

22:59

to that that's going to be

23:01

a trouble. It's going to be

23:04

AI crypto and robotics are really

23:06

going to prove to be a

23:08

natural fit. I think we're starting

23:10

to see that already. Some really

23:13

interesting things are happening

23:15

with robotics that we've been

23:18

following closely, even though it

23:20

hasn't touched on crypto too much

23:22

yet. It kind of started with

23:24

a couple. breakthrough humanoid

23:27

robotics at least,

23:29

much like AI had a

23:31

couple key breakthrough innovations.

23:33

Humanoid robotics

23:35

had one a few years

23:37

ago that has sparked a

23:40

huge increase in the pace

23:42

of development. And

23:44

interestingly, it's one of

23:46

the things like once you

23:48

say, at least to me, maybe

23:51

because of my combat sports

23:53

background, but it. At least

23:55

to me as soon as I heard it, I was like, oh

23:57

my God, that makes so much sense. The real break...

24:00

through that they were having a

24:02

hard time replicating the way human

24:04

hips can swivel and move around,

24:06

which is a key component of

24:09

your your body's ability to navigate

24:11

everything and have balance and things

24:13

like that. And so the key

24:16

innovation that is sparking a revolution

24:18

in humanoid robotics actually is related

24:20

to kind of like the robots

24:23

pelvis, which I I find incredibly

24:25

fascinating. But they just couldn't really

24:27

replicate all of the dynamic motion

24:30

associated with, you know, the pelvis,

24:32

so to speak. And once they

24:34

were able to crack that problem,

24:37

that's why we're seeing all these

24:39

humanoid robotics companies go vertical. It's

24:41

really quite fascinating. My mind is

24:44

going to a, a dirty place

24:46

there. Well, it's just, but if

24:48

you're, I mean, if you're going

24:51

to stabilize your body, Or if

24:53

you're going to traverse a hill

24:55

or something like that, that's gonna

24:58

be the key component. Maybe the

25:00

foot is also a key component,

25:02

but also the pelvis for sure.

25:05

And for whatever reason, they were

25:07

just having a hard time replicating

25:09

that. And so it turns out

25:12

that was one of the toughest

25:14

problems to crack and humanoid robotics,

25:16

as I understand it. This is

25:19

what they were explaining to me

25:21

at Unatree earlier in the week.

25:23

Unatry is one of the... top

25:26

maybe five robotics companies in the

25:28

world. They're located in Hangzhou and

25:30

They're doing some of the leading

25:33

stuff along with in the world

25:35

along with a figure and a

25:37

couple other companies But in any

25:40

rate when they were giving me

25:42

the tour and they challenged me

25:44

at one point They challenged me

25:47

to see what would happen if

25:49

I if I shoved a couple

25:51

other robots a humanoid one and

25:54

then they were like They were

25:56

like this robot dog. We want

25:58

you to just kick it as

26:01

hard as you can eat it.

26:03

See what happened And once they

26:05

convinced me that they really in

26:08

fact did want me to do

26:10

that, I kicked the damn thing

26:12

like a soccer ball and it

26:15

was able to, you know, it's

26:17

a heavy thing. It's a heavy

26:19

thing. It's like maybe 40 pounds,

26:22

something like that. And it went

26:24

up, you know, a few feet

26:26

in the air and it was

26:29

able to write itself and it

26:31

was able to land. on four

26:33

feet, which I, I mean, I

26:36

kicked that thing pretty hard. And

26:38

similarly, they were like, we want

26:40

you to just shove this humanoid

26:43

robot. Just give it a shove,

26:45

like you're, I don't know, like

26:47

it's a football game and you're,

26:50

you're trying to tackle somebody or

26:52

something. And so I went and

26:54

I shoved one of their G1s

26:57

that was, this was actually in

26:59

Vegas at CES, this was a

27:01

partner company of theirs. So I'm

27:04

combining a couple stories, a couple

27:06

stories here. But at CS, I

27:08

shoved their G1 as hard as

27:11

I can, and that thing had

27:13

amazing balance. They didn't mind. I

27:15

mean, I really gave it a

27:18

shove. It's like maybe a 100

27:20

pound robot, something like that. They're

27:22

going to remember that when they

27:25

take over. Yeah, I know. Exactly.

27:27

I'm on a robot last night.

27:29

But they have amazing balance, amazing

27:32

ability to dynamically react to the,

27:34

I don't know the exact technical

27:36

terms, but to react to the

27:39

force that you're applying. Excuse me.

27:41

And it's really quite fascinating. And

27:43

as they were explaining to me

27:46

earlier in the week at unit

27:48

street headquarters, the key innovations or

27:50

one of the key innovation is

27:53

they finally were able to get

27:55

the hips right so that they

27:57

can properly balance. It's all the

28:00

hips. It's all the hips. Yeah.

28:02

Well, it's just I found it

28:04

so interesting because they say that

28:07

in wrestling that that's one of

28:09

the key things. It's like control

28:11

the head, control the hand. And

28:14

then it's all about your hips,

28:16

you know, is where a lot

28:18

of your agility and your ability

28:21

to use power against people comes

28:23

from. And so I just for

28:25

that reason I found it really

28:28

quite interesting. Here we go again.

28:30

Like of course it is. So

28:32

I feel like that's sort of

28:35

the going to be the most

28:37

difficult aspect in terms of when

28:39

people think about a sci-fi like

28:42

reality for the world. It's getting

28:44

these robots to have the same.

28:46

type of mobility as as humans.

28:49

It's probably the tallest order and

28:51

the most difficult thing to get

28:53

right. Sure. Here's what's so interesting

28:56

too. This is what we've learned

28:58

recently. So it's adjacent to what

29:00

you were just saying. So it's

29:03

for sure about agility and things

29:05

like that, but increasingly they're making

29:07

huge progress on that problem. Here's

29:10

the remaining problem though. With a

29:12

lot of the big robots, robot

29:14

manifest, okay, so if you buy

29:17

a robot that's like a $500

29:19

child's toy, it's like just for

29:21

fun. It's, you know, you can

29:24

buy it at almost like the

29:26

mall or something like that. Then

29:28

you will have the ability to

29:31

do some of your own programming,

29:33

like you can add features and

29:35

things like people in. You know

29:38

people used to they wanted to

29:40

if you have a if you

29:42

were for people that were precocious

29:45

kids and they wanted to start

29:47

coding little things in a visual

29:49

basic or whatever you can buy

29:52

a $500 kids robot and start

29:54

programming stuff. But the problem is

29:56

for all of the expensive robots

29:58

like the unitry G1 and the

30:01

H1 these robots that are at

30:03

price points from you know $40,000,

30:05

$100,000, $200,000. They have the agility

30:08

but it's only for a few

30:10

pre-programmed actions. Might be five actions,

30:12

might be eight actions, something like...

30:15

that. And then so much of

30:17

the technology is behind a black

30:19

box that it's not exposed in

30:22

a way that you can be

30:24

programming your own actions. So we're

30:26

left with a really high level

30:29

of agility across a really small

30:31

number of types of actions. It's

30:33

actually one of the biggest blockers

30:36

right now. I believe that all

30:38

the leading robotics companies probably recognize

30:40

that this is basically unsustainable, but

30:43

But currently that's kind of where

30:45

we're at. That makes a lot

30:47

of sense. So let's, let's sort

30:50

of think through how this, I

30:52

mean, this is why you guys

30:54

are sort of paying so much

30:57

attention to the hardware side of

30:59

things. Give us a lay of

31:01

the land. What does, what does

31:04

that arena look like and how

31:06

do you invest in hardware? How

31:08

does, how does investing in hardware

31:11

look different than in applications and?

31:13

blockchain, software, etc. Sure. So I

31:15

would say that we are super

31:18

interested in robotics again at the

31:20

intersection of AI and crypto, obviously,

31:22

but then you have to have

31:25

conversations in the roadmaps that are

31:27

that are serious. So if someone

31:29

says, well, we're looking to go

31:32

head to head with Tesla. for

31:34

robotics and you know are head

31:36

to head with Tesla the optimist

31:39

and that obviously doesn't make sense

31:41

at all. But for example what

31:43

was one thing that was super

31:46

interesting to us about Eliza wakes

31:48

up is that they are actually

31:50

partnering with unitry to make a

31:53

next generation Sophie of the robot.

31:55

And so basically because their whole

31:57

vision is well we want the

32:00

most recognizably human A.I. agent. We

32:02

want Eliza in the lower of

32:04

it as Eliza is an AI

32:07

that one day becomes sent to

32:09

it and wakes up as a

32:11

real girl. And so what if

32:14

so they were their concept is

32:16

what if you take that to

32:18

the limit well she needs to

32:21

have a physical body and so

32:23

they're actually working with unitry to

32:25

bring Eliza to life. So the

32:28

chassis so to speak will be

32:30

unitries top of the line. Chassy.

32:32

which is the H1, there are

32:35

only a few hundred or maybe

32:37

at most a thousand in the

32:39

entire world that are of that

32:42

quality level. And then they're working

32:44

with, if you've seen Fast and

32:46

the Furious, how they have like

32:49

a hot rod shop, do you

32:51

take a, you know, a $250,000,

32:53

not a car guy, but I

32:56

guess they take like a $250,000

32:58

car, bring it to the hot

33:00

rod shop, put all this work

33:03

into it, and then they end

33:05

up with something super custom, super

33:07

custom. and super amazing right so

33:10

Eliza wakes up in addition to

33:12

unitry is working with the equivalent

33:14

of a hot rod shop only

33:17

for robotics and it's Nick and

33:19

Eric and the gang over at

33:21

Old World is the name of

33:24

the company and so they're gonna

33:26

have a face that can be

33:28

animated and expressive they're going to

33:31

have they're going to have it's

33:33

going to look like like a

33:35

real person and move like basically

33:38

like Sophia the robot you know

33:40

from four or five years ago

33:42

that was very famous for being

33:45

the top of the line humanoid

33:47

robot at that time and they're

33:49

going to produce the 2025 version

33:52

of that the best in class

33:54

humanoid robot moves like a real

33:56

person in the way it expresses

33:59

itself animates and everything and and

34:01

then and then it will actually

34:03

have Eliza OS and Eliza the

34:06

character as an LLM that will

34:08

come along with the robot. So

34:10

you'll actually be able to converse

34:13

with her and it will really

34:15

feel like you're talking to a

34:17

real person. So I think I

34:20

thought that was pretty exciting. And

34:22

that's a clearly differentiated vision that

34:24

based on getting to know the

34:27

team at Old World and based

34:29

on getting to know unitry is

34:31

actually quite realistic. I believe that

34:34

they're going to be able to

34:36

deliver on that. That well I

34:38

ordered one myself actually because of

34:41

course I did it sounds amazing

34:43

it sounds so interesting right and

34:45

that'll be actually available within probably

34:48

four to six months I think

34:50

they're saying four months and then

34:52

I had had a couple just

34:55

just out of conservatism so that's

34:57

super interesting but then also additionally

34:59

we're looking a lot at at

35:02

a Dow called Homebrew these days

35:04

which is really up to some

35:06

fascinating things and So homebrew, the

35:09

lure on homebrew, homebrew computer club

35:11

is a very famous name in

35:13

computing history. It was in the

35:16

stage of home computing when really

35:18

everyone was the very beginning of

35:20

home computing, which was in basically

35:23

the mid to late 1970s. And

35:25

so at that time, who was

35:27

involved in home computing, well, was

35:30

people that would. get together and

35:32

they would bring circuit boards and

35:34

they would be they would be

35:37

doing everything from scratch by hand

35:39

like hackers and hobbyists. And the

35:41

homebrew computing club has a famous

35:44

role in computer club rather as

35:46

a famous role in computing history

35:48

because Steve Jobs was a member,

35:51

Wozniak was a member and actually

35:53

there were a bunch of people

35:55

that later became famous in home

35:58

computing and certain companies and did

36:00

really amazing stuff. So it has

36:02

a famous, a famous. rule that

36:05

it played in connecting all these

36:07

people and helping them learn computers

36:09

together, learn what actually is a

36:12

home computer and what could it

36:14

be at the beginning stages of

36:16

that industry. So then Homebrew, Homebrew

36:19

Robotics, they basically started with the

36:21

question, well, if you had Homebrew

36:23

Computer Club for robotics and And

36:26

they had the assumption, which I

36:28

believe to be correct, that humanoid

36:30

robotics at least is at a

36:33

similar inflection point, similar stage in

36:35

its development, where it's just starting

36:37

to go hockey stick, that was

36:39

their assumption. So then they asked

36:42

the question, okay, so if you

36:44

had a similar concept, but for

36:46

robotics, and at the intersection of

36:49

AI and crypto, and it's 2025,

36:51

what would that look like? basically

36:53

had that as a thought experiment

36:56

and well, well, you would, now

36:58

we have the internet, you have

37:00

town all around the world, there's

37:03

no reason to have it in

37:05

some kind of specific physical location,

37:07

meeting once a week or once

37:10

a month or whatever, it would

37:12

exist both online and in person,

37:14

a community has to have a

37:17

place to live, so it would

37:19

live and meet in a discord,

37:21

it would run on crypto rails,

37:24

it'd be tokenized in various aspects

37:26

and things like that. So they're

37:28

really up. So to some cool

37:31

things in in stealth to where

37:33

I think that at a minimum

37:35

their vision is big enough to

37:38

kind of be the AI 16

37:40

Z instead of for AI for

37:42

robotics So that's another one that

37:45

really we really think they've been

37:47

mostly in stealth I actually had

37:49

to ask them if it was

37:52

okay to to talk about them

37:54

on the show today and they

37:56

did they did say yes, but

37:59

But they're I guess just starting

38:01

to come out of stealth and

38:03

they have some really exciting initiatives

38:06

that we're quite keen on. You

38:08

heard it here first, folks. Yeah,

38:10

yeah, yeah. So I guess really,

38:13

I think these are natural parents,

38:15

robotics and crypto and AI. And

38:17

I think, I do think there's

38:20

gonna be a robotics medal. I

38:22

do think there's gonna be an

38:24

explosion of activity happening in that

38:27

industry and some of it for

38:29

sure will happen in crypto. I

38:31

feel like we have to address

38:34

the elephant in the room. What

38:36

are we going to be doing

38:38

the horizontal mumbo with these things?

38:41

People do, I'm told people do

38:43

a lot, ask the allies, the

38:45

team about that a lot. And

38:48

so they have to keep saying,

38:50

hey, can't have sex with it.

38:52

And it's not a sex robot.

38:55

I wouldn't want to kink shame.

38:57

Frank, if you do want to

38:59

have sex. No, no. Well, well,

39:02

I mean, it is on your

39:04

mind. I would be remiss if

39:06

I didn't. I'm not going to

39:09

keep waiting. I got my own

39:11

thanks. There's a lot. There's a

39:13

wide range of listeners to this

39:16

show. Oh, excuse me. You know,

39:18

so we've got a, okay. All

39:20

right. No, I'm simply a vessel

39:23

for the audience. So I'm sure

39:25

a lot of people. Me too.

39:27

We'll, we'll, we'll be thinking this.

39:30

Or maybe not, and they'll think

39:32

I'm an absolute pervert. Like I

39:34

said just now, hey, I went

39:37

on a king shame. I've got

39:39

my own king. I too was

39:41

just a vessel for the people.

39:44

For the people, yeah. They have,

39:46

yeah. Well, maybe we'll edit this

39:48

out. When you think of it.

39:51

I'm to blame. Or when you

39:53

think about 2025, what are the

39:55

key trends and breakthroughs in AI

39:58

that you're most bullish about? I

40:00

would. I think the most interesting

40:02

phenomenon that we're

40:04

going to continue to see

40:06

is that a lot of the

40:08

AI schools, the cost is really

40:11

going to go to zero. I

40:13

mean, not literally and

40:15

not necessarily in 2025,

40:17

but we're going to

40:19

continue to see the

40:21

cost plummet. And I think

40:23

that's really what's going

40:25

to open up. a huge design

40:28

space and take us to the

40:30

next level. That by far and

40:32

away is the one thing that I'm

40:34

most excited about, I would

40:36

say. Amazing. Well, this has been

40:39

a blast. We'll have to have

40:41

you back on in six months.

40:43

When Eliza's ready for her

40:45

debut on the program, she

40:47

also has a standing invite.

40:49

Where can folks learn more? Where

40:52

can they find you? I know

40:54

you're active on the Twitter's and...

40:56

I'm sure you have a

40:58

website. Let's get them. Madie

41:00

Rise on Twitter. I famously

41:03

am a little bit addicted

41:05

to Twitter. And that's actually

41:08

probably the best. Don't

41:10

email me. If you

41:12

email me, I probably

41:14

will never see it to

41:16

be honest. But you can

41:18

DM me on Twitter. R-Y-Z-E.

41:21

M-A-T-T-Y-R-Y-Z-E. Our company's website

41:23

is rise Labs Labs. I oh and

41:26

and Eliza wakes up is the name

41:28

of the I guess Eliza wakes

41:30

up is probably the Twitter

41:32

for the Eliza project I'd have

41:34

to confirm that but Eliza wakes

41:37

up and then also we were

41:39

talking about ticker Eliza and

41:41

then also we were talking about

41:43

homebrew robotics club and suppose since

41:46

I did bring them out of

41:48

stealth today with permission then I

41:50

suppose I suppose I should I

41:53

should at least say the ticker

41:55

which is brew B-R-E-W. So those

41:58

are some ways to find. some

42:00

of the things that we've talked

42:02

about today talked about today. Amazing. so much

42:04

for your time so talking to

42:06

you soon thank you time. We'll be talking to

42:08

you you be well you. Be well.