0:06

Welcome to the Making Sense podcast.

0:08

This is Sam Harris. Just a note

0:10

to say that if you're hearing this,

0:12

you are not currently on our subscriber

0:15

feed, and we'll only be hearing the

0:17

first part of this conversation. In order

0:19

to access full episodes of the Making

0:21

Sense podcast, you'll need to subscribe at

0:23

samharis.org. There you'll find our private RSS

0:26

feed to add to your favorite podcatcher,

0:28

along with other subscriber-only content. We don't

0:30

run ads on the podcast, and therefore

0:32

it's made possible entirely through the support

0:34

of our subscribers. So if you enjoy

0:37

what we're doing here, please consider

0:39

becoming one. Welcome to

0:41

the Making Sense Podcast.

0:43

This is Sam Harris.

0:45

Today I'm speaking with Ben

0:47

Lamb. Ben is a technology

0:49

and software entrepreneur

0:52

who has been featured in many

0:54

publications, the Wall Street

0:56

Journal, New York Times,

0:59

Forbes, discussing topics related

1:01

to innovation and innovation

1:04

and innovation. He is also

1:06

the co-founder and CEO of

1:08

Colossel Biosciences, a company he

1:10

started with biologist George Church

1:13

for the purpose of resurrecting extinct

1:15

species like the woolly mammoth and

1:17

the Tasmanian tiger and the dodo.

1:20

And they aim to reintroduce

1:22

them into the wild. Ben is also

1:24

a fellow of the Explorers Club and

1:26

serves on the Scientific Advisory

1:29

Board of the Planetary Society.

1:31

But we focus on his work at

1:33

Colossel. We discussed the difference

1:35

between their approach and Jurassic

1:38

Park, the details of resurrecting the

1:40

mammoth and other species, the relevance

1:42

of this work to human health, the

1:44

role of artificial intelligence here,

1:47

what it would take to reintroduce

1:49

mammoths and Tasmanian tigers and

1:51

dodos back into the wild,

1:53

the environmental and business case

1:55

for doing this, and other topics. Anyway,

1:57

the future appears to be almost...

2:00

here. And now I bring you,

2:02

Ben Lamb. I am here with

2:04

Ben Lamb. Ben, thanks for joining

2:06

me. Thanks so much for having

2:08

me. So we're going to talk

2:10

about some amazing stuff that you're

2:12

doing over there at Colossel, your

2:14

biotech company. But before we get

2:16

there, how do you summarize your

2:18

career and interest at this point?

2:20

How did you give me the...

2:22

potted bio that gets us to

2:24

the topic at hand. Well, I'm

2:26

definitely insatially curious. And so I'm

2:28

always, you know, I'm not really

2:30

a technologist, I'm not really an

2:32

engineer. I try to look at

2:34

things from a systems design perspective,

2:36

and I'm always fascinated with how

2:38

things work and how things can

2:40

be improved. And I always like

2:42

to find new interesting projects. And

2:44

so I've been in everything from

2:46

mobile gaming before that was quite

2:48

big. I built some precursors to

2:50

large language models. that we were

2:52

actually calling conversational operating systems at

2:54

the time. My last company was

2:56

actually satellite software in defense. So

2:59

we actually built a common operating

3:01

picture to understand and track everything

3:03

in the sky all the way,

3:05

actually lower the orbit all the

3:07

way down to the surface of

3:09

the sea and work closely with

3:11

the US Air Force and Space

3:13

Force and some of our global

3:15

partners on that. And then I

3:17

met George Church and, you know,

3:19

I actually kind of fell into

3:21

the extinction. I reached out to

3:23

him because I'm curious and I

3:25

thought that the intersection of synthetic

3:27

biology and AI and computational biology

3:29

and you know quantum which I

3:31

hear is only two years away

3:33

every two years will eventually you

3:35

know kind of give us dominion

3:37

to engineer life and do directed

3:39

evolution on a scale that you

3:41

know is unprecedented for you know

3:43

human advancement and so I got

3:45

massively excited about the opportunities there

3:47

and and then I asked George

3:49

the question and I said if

3:51

you had one. project with unlimited

3:53

capital that you could focus on

3:55

for the rest of your life.

3:57

You know, what would it be,

3:59

George? And, you know, did know

4:02

what I would get out of

4:04

George. Is it going to, you

4:06

know, another star system or what?

4:08

And his feedback was I would

4:10

bring back Willie Mammus and help

4:12

reintroduce them back into the ecosystem

4:14

to help biodiversity and the ecosystem

4:16

as well as develop technologies for

4:18

both human health care and species

4:20

preservation. And at that moment I

4:22

was pretty hooked. Yeah, George is

4:24

a very impressive scientist. I've met

4:26

him, I think, it might have

4:28

only been once, maybe twice, at

4:30

a conference, but... Is he still

4:32

at Harvard? He's still at Harvard.

4:34

So I do get to monopolize

4:36

a decent amount of his time,

4:38

but we do share him with

4:40

Harvard and a handful of other

4:42

initiatives he's co-founded. So the company

4:44

is colossal biosciences, is that the

4:46

the full name? Correct. And so

4:48

what are you doing over there

4:50

at colossal? Yeah, so we decided

4:52

that we wanted to build the

4:54

world's first deextinction and species preservation

4:56

company because... If you look at

4:58

some of these stats and kind

5:00

of the trend line that we're

5:02

seeing for biodiversity loss and what

5:05

the impacts to ecosystems can and

5:07

will be especially from a keystone

5:09

perspective, it's pretty terrifying. And when

5:11

we started the company, our original

5:13

pitch deck, all the data we

5:15

could find showed that if without

5:17

massive human intervention or massive new

5:19

technologies, that we could lose up

5:21

to 15, 1, 5% of biodiversity

5:23

between now and 2050. What's terrifying

5:25

is in 2024 that number has

5:27

been up to 50% 5-0. So

5:29

that's not a very good trend

5:31

line. And so George had this

5:33

vision, and I just feel like

5:35

I'm kind of the steward and

5:37

helper with it, we could go

5:39

build a company that could, you

5:41

know, one, build tools and technologies

5:43

that could be capable of bringing

5:45

back lost species, as well as

5:47

applying those technologies and innovation to

5:49

conservation, giving that to the world

5:51

for free. And all these species

5:53

have direct applications, those technologies like

5:55

genetic engineering and others, to human

5:57

health care. So we really had

5:59

this interesting opportunity to build. company

6:01

that hopefully could inspire people, create

6:03

true impact, but also create massive

6:05

value creation around the way. And

6:08

which species are you focused on

6:10

first? So we've announced three species

6:12

today, the woolly mammoth, which

6:14

George was actually working on

6:16

for about eight years before

6:18

I showed up, collecting samples

6:20

in Siberia, working on computational

6:22

analysis and elephants. The Tasmanian

6:25

tiger, also known as the

6:27

thylacine, which went extinct in

6:29

1936 in Australia due to

6:31

human hunting, the Australian government

6:33

actually put a bounty on

6:35

eradicating the species. And then, you know,

6:37

we wanted a bird species, we wanted

6:39

to recruit Besh Shapiro, who's our chief

6:41

science officer. So we did the dodo,

6:43

because there's probably not a more

6:45

iconic species than the dodo that

6:47

symbolizes deextinction. So how is this

6:50

different from Jurassic Park? I mean,

6:52

you know, I don't think anyone

6:54

would really associate it with Jurassic

6:56

Park until you bring in the

6:58

mammoth and then all of a

7:01

sudden the... We're talking about charismatic

7:03

megafauna and we're hoping for a

7:05

T-Rex. To what degree does that

7:07

vision account for some of your

7:09

enthusiasm around this? And I mean

7:12

obviously there's a difference between reintroducing

7:14

animals to the wild and setting

7:16

up a theme park. Was Jurassic

7:18

Park a formative idea for you or

7:20

is that or you've arrived where you

7:23

are by a different path? So we

7:25

get the Jurassic Park question quite

7:27

a bit as you, as that

7:29

may not surprise you. Like occasionally

7:31

when I go on stage to

7:34

speak, they'll play the music. You know,

7:36

we've seen every meme with like George's

7:38

face on it or my face on

7:40

it. So we've heard this a time

7:43

or two. George will tell you that

7:45

in a weird way he thinks

7:47

that Michael Crichton was actually

7:49

inspired by him because if you

7:51

go look in the original... Jurassic

7:53

Park novel. There's actually a DNA

7:56

sequence early in the in the

7:58

in the work in the. And

8:00

it actually is George's work with only

8:02

one letter changed. And George will argue

8:04

that statistically he... It's still plagiarism. It's

8:07

still... And George loves, you know, many

8:09

of Crichton's novels, right, and it's a

8:11

very inspiring author that he was. And

8:14

but George will tell you that, you

8:16

know, he laughs and says, maybe I

8:18

inspired Jurassic Park, because a lot of

8:21

his original work in yeast is actually

8:23

shows up in the book. I will

8:25

tell you from my perspective, you know,

8:28

growing up, you know, born in the

8:30

80s, childily 80s and 90s, you know,

8:32

I think one, you know, I love

8:34

science fiction, I love Jurassic Park. That's

8:37

not necessarily why I got into this,

8:39

but it sure makes it a lot

8:41

easier to connect with people because even

8:44

though we have the memes and all

8:46

the jokes that come around colossal versus

8:48

Jurassic Park, which was this dystopian movie,

8:51

at least it taught people about there's

8:53

this thing called DNA. and there's this

8:55

thing called genetic engineering. And so like

8:58

moms in Iowa know that there's this

9:00

ability to manipulate the genome because of

9:02

Mr. DNA, right? And so we also,

9:05

we a lot of times use Jurassic

9:07

Park as an example of how we're

9:09

doing it exactly inverse, meaning that we're

9:12

not trying to fill the gaps in

9:14

a ancient DNA that with the holes

9:16

that you get from, you know, frogs

9:18

or whatnot. we're trying to truly understand

9:21

the genomes so that we could selectively

9:23

choose the genes that we then want

9:25

to engineer into that of a living

9:28

species. So it's almost like reverse drastic

9:30

park. And when we say that to

9:32

the kind of average public in that,

9:35

in some journalists and whatnot, when we're

9:37

explaining the process and the science, they

9:39

really resonate with it. Because I think

9:42

that movie does have such a, was

9:44

the right movie with the right technology

9:46

and the right story at the right

9:49

time, that really connects with people. So

9:51

let's go over those details again. So

9:53

what was being proposed as the scientific,

9:55

you know, bioengineering basis for Jurassic Park

9:58

and and what exactly are you doing

10:00

with paleogenomics and going out into the

10:02

wild and getting DNA samples, however imperfectly

10:05

preserved, and integrating them with living species?

10:07

What is your approach and how is

10:09

it different from what was being? It's

10:12

been a long time since I saw

10:14

the film. I actually never read the

10:16

novels. I don't know if the films

10:19

depart from the novel in their logic.

10:21

And I know nothing about... any of

10:23

the errors that Crichton might have made

10:26

with respect to his molecular biology if

10:28

he made any. So what was proposed

10:30

there and what are you guys actually

10:32

doing? So in Drusk Park, they propose

10:35

that you could go find pieces of

10:37

like amber, which by the way is

10:39

a very porous material. It is not

10:42

a good DNA store, not that we've

10:44

tried, but then magically in amber you'd

10:46

get insects and specifically mosquitoes. that had

10:49

been trapped for over 65 million years.

10:51

And while that's true, there isn't DNA

10:53

from that. Amber has images of very

10:56

porous material. It is not a great

10:58

DNA store. Typically the best DNA stores

11:00

for us for ancient DNA are cold

11:03

dry places. So animals that passed away

11:05

in a cave and a very dry

11:07

cave that stayed consistent without other animals

11:09

in it. That's kind of optimal for

11:12

us. And so then they would take

11:14

this DNA. that they extracted from a

11:16

mosquito that lived, you know, a hundred

11:19

million years ago and been a dinosaur,

11:21

and they would extract in the movie

11:23

actual blood, which also is impossible. And

11:26

then they would take that blood, use

11:28

computers, which is very similar to what

11:30

we do, which I'll get into, and

11:33

then fill in the holes of the

11:35

ancient DNA, because ancient DNA is very,

11:37

very fragmented, with that of in the

11:40

movie Frog DNA, amongst some other, many

11:42

other things. But the problem with that...

11:44

Number one is there is an ancient

11:47

Dino DNA, you know, the oldest DNA

11:49

that we're able to collect is, you

11:51

know, a little bit over a million

11:53

years. some fragments and stuff that are

11:56

older, but you know, for the most

11:58

part, we're working in thousands and tens

12:00

of thousands of years, not, you know,

12:03

millions of years, because DNA degrades very,

12:05

very quickly. It starts to break down

12:07

the minute it leaves your body, and

12:10

so when you layer in like radiation,

12:12

heat, acidification, other animals defecation, other animals

12:14

dying on it, it starts to break

12:17

down, and it also gets massively contaminated.

12:19

It's on truly endogenous at that point,

12:21

right. And so what we do is.

12:24

Instead of going and taking a bunch

12:26

of different pieces of a mammoth, assembling

12:28

it and saying what's missing, and how

12:30

do we plug that with a frog

12:33

or elephant DNA, we do it almost

12:35

exactly in reverse. So the first thing

12:37

that we did is we went out

12:40

and we looked at phylogenetically, so on

12:42

that tree of life that we've all

12:44

seen, some version of it, you know,

12:47

in science textbooks and today on the

12:49

internet, we say what is the closest

12:51

living relative to the mammoth in this

12:54

case? and that's actually the Asian elephant.

12:56

It's 99.6% the same genetically. It's actually

12:58

closer genetically to an Asian elephant than

13:01

an Asian elephant is to an African

13:03

elephant and that's kind of a fun

13:05

party trivia for you. And then we

13:07

spend a lot of time trying to

13:10

do comparative genomics, truly use a bunch

13:12

of software, use AI, some of our

13:14

custom models to understand what is the

13:17

difference even from an African elephant to

13:19

an Asian elephant. What is the difference

13:21

from a population level? So we actually

13:24

sequenced a lot of different Asian elephants.

13:26

So what is truly Asian elephant versus

13:28

population diversity in those genomes? Because not

13:31

all genomes are obviously exact copies of

13:33

each other. And then how do we

13:35

compare that to the mammoth? And then

13:38

we can identify, okay, where are these

13:40

regions of the genome that are vastly

13:42

different? And what do we know about

13:44

that from scientific research, from other peer-reviewed

13:47

papers? from actually doing molecular and functional

13:49

assays, actually growing stem cells and testing

13:51

our hypothesis. So you have to do

13:54

a lot of work to then kind

13:56

of verify what we think. the core

13:58

genes that made a mammoth a mammoth

14:01

were so that then we can engineer

14:03

them into that of an Asian elephant

14:05

cell and that's not just taking pieces

14:08

and pushing it in there that's actually

14:10

just changing existing code so we fundamentally

14:12

don't need long-term pieces of these DNA

14:15

we don't need all these dead samples

14:17

we just need the code in the

14:19

computer so do we have the complete

14:22

genome of the willy mammoth I mean

14:24

is that something that's disputed or did

14:26

we get enough samples of sufficient integrity

14:28

such that we just know we've got

14:31

the full mammoth genome? We have enough.

14:33

So we have we have about 65

14:35

mammoth genomes. Most of those aren't published.

14:38

Most of those aren't published. Most of

14:40

those are Siberian and Russian mammoth samples.

14:42

We're now doing a lot of work

14:45

with Alaskan mammoths as well. And we

14:47

work with about 17 universities across the

14:49

world, one of which is the University

14:52

of Stockholm and Louva Dolan research in

14:54

the world. And so... We've taken all

14:56

of his different samples and it's about

14:59

a 700,000 year difference between all the

15:01

different samples to kind of fill that

15:03

in. But we have enough of the

15:05

protein coding regions of it as well

15:08

as Colombian mammists, step mammists, and others.

15:10

And we have a pretty cool paper

15:12

that I hope will come out mid-next

15:15

year about this that shows the comparative

15:17

genomics that we know enough of the

15:19

mammoth genome that we can identify the

15:22

core areas around cold tolerance, fat, hair,

15:24

curved tus. So we actually have enough

15:26

to do our work. It is not

15:29

as complete as our thylizing genome, which

15:31

we recently announced is 99.5% complete, or

15:33

sorry, 99.9% complete, which is truly incredible

15:36

for any genome, let alone ancient DNA.

15:38

That's the Tasmanian tiger? Correct. So are

15:40

you using CRISPR technology to insert mammoth

15:42

code into an Asian elephant zygote? Or

15:45

what is the step there that would

15:47

produce a living mammoth? Yeah, so we

15:49

start with an Asian elephant cell, right?

15:52

And we actually had this spend a

15:54

lot of time getting the culture conditions

15:56

right, actually immortalizing this. cells, one of

15:59

the things that, you know, before we

16:01

get into the genetic engineering side, one

16:03

of the things that's interesting about elephants

16:06

and blue whales in a handful of

16:08

other species is they actually get cancer

16:10

a fraction of what we do based

16:13

on age and body weight of which

16:15

they grow to. And the leading theory

16:17

of that, and we're seeing this also

16:19

being verified in our lab, is they

16:22

have an over-expression of a protein called

16:24

P53, about seven times more than we

16:26

have in mice, which I'm sure you're

16:29

familiar with. And what's interesting is we've

16:31

actually had to learn how to regulate

16:33

that, because any time we want to

16:36

go make those changes, which we'll get

16:38

into, the cell would just seness. So

16:40

not only do we have to build

16:43

immortalization constructs to keep the cells growing

16:45

and living and healthy, we also had

16:47

to figure out how we can quote

16:50

unquote turn down P53 so that we

16:52

could edit the cells and then be

16:54

able to turn it back up because

16:57

you don't want to produce cancer in

16:59

elephants, right? And so we had to,

17:01

there's a lot of prep work before

17:03

we even get to the point that

17:06

we can do the engineering itself. And

17:08

as you can probably guess, you know,

17:10

because your deep background in science, you

17:13

know, CRISPR has become a catchall for

17:15

all genetic engineering. They're like, oh, it's

17:17

just CRISPR, right? We just, we just

17:20

CRISPR it. But what's interesting is we

17:22

use a combination of tools, some of

17:24

which are proprietary, So in some cases,

17:27

we're changing the individual nucleotides, the individual

17:29

letters on that double helix, in other

17:31

cases we're knocking out certain genes, and

17:34

in other cases we're actually synthesizing big

17:36

blocks of DNA, where if there's like

17:38

a bunch of changes along one kind

17:40

of strand, it's actually more efficient for

17:43

us to synthesize that block, knock that

17:45

block out, and then insert this new

17:47

block so that you have less likelihoods

17:50

of off-target effects or unintended consequences. from

17:52

your editing. And I'd say the last

17:54

thing that we're doing that on the

17:57

editing front that is our kind of

17:59

I. I think the thing that sets

18:01

us apart from a genomics perspective

18:04

is we're trying to become the

18:06

biggest pioneer of multiplex editing, meaning

18:08

editing all over the genome at

18:10

the same time. So instead of making

18:12

one edit, maybe you can make 20

18:15

edits, 50 edits, 1,000 edits, all with

18:17

a very high degree of efficiency,

18:19

versus having to synthesize entire

18:21

giant blocks. I do believe

18:23

that technology will get here,

18:25

being able to synthesize even

18:27

full chromosomes at some point.

18:29

But we're not, we as

18:31

humanity aren't quite there yet.

18:33

So editing is the most

18:35

efficient kind of current modality

18:37

that we've been pursuing. So

18:39

at what point did this

18:42

actually become technically feasible? And

18:44

what year would you say

18:46

this became something that you

18:48

could actually start on and

18:50

it seems to be just a

18:52

piece of science fiction? Yeah, so

18:54

I think, you know, people have

18:56

been talking about, you know, version

18:58

of genetic engineering from the 80s,

19:00

right? But it was like, it was,

19:02

I don't remember the exact year, but

19:04

it was like what, 2012, 2014, somewhere

19:07

around there, where we had the

19:09

true kind of discovery around, you

19:11

know, CRISPR, and the idea that you

19:13

could, you know, target a part of

19:15

the genome, successfully knock it out and

19:17

have it repair itself. And I think

19:20

from there, you've seen work like David

19:22

Liu's work in prime and base

19:24

editing, where you can change individual

19:26

letters. you've seen kind of this

19:28

like pre-Cambrian explosion, you know, to

19:30

use our Jurassic, you know, use

19:32

our, some of our Jurassic fun

19:34

terms of genetic engineering tools and

19:36

technologies, because we've all been promised

19:38

from the 80s and 90s gene

19:40

therapies and genetic engineering capabilities that

19:42

allow us to do all kinds

19:45

of stuff, right, that have never

19:47

really manifested. But I think that that

19:49

really, in the last, you know, 10

19:51

years has been where those technologies have

19:54

been viable. I don't believe before that

19:56

kind of 2012 2015 time frame of

19:58

like that that crisper. race with, you

20:00

know, fang and, and, and, and, and,

20:02

and, and, and, and, all of them,

20:04

right, that are just, they're, and George,

20:06

included, which were all incredible scientists, I,

20:08

I don't believe that this would have

20:10

been a viable undertaking. And, and, and

20:12

now, after that, it became viable, but

20:14

it, you know, you saw have compute,

20:16

you saw have AI, there's a lot

20:19

of other components to it. And it's

20:21

just becomes very, very, very, very, very,

20:23

very, very costly. I think is, I

20:25

think we're still a little bit early,

20:27

but we're in kind of the right

20:29

kind of five years to truly be

20:31

able to deliver. So is AI a

20:33

necessary component of the process? It is.

20:35

And you know, we're learning every day

20:37

new ways that we can apply. You

20:39

know, my background has been mostly in

20:41

software, right? And so, you know, we're

20:43

finding every day new ways to apply

20:45

these technologies around it. Like we actually

20:47

have a tool that we built internally

20:49

that we've been giving it this feedback

20:51

So we built a cool little model

20:53

that probably doesn't apply to most people,

20:55

but for us we find it fascinating,

20:57

that will actually give us the right

20:59

recommendation that's over 90% accurate of what

21:01

tool we should use for the specific

21:03

edit that we're going after. And that's

21:05

awesome when you think about biology, because

21:07

if you're going to make an edit,

21:09

you then have to go see if

21:11

that edit worked, you then have to

21:13

grow those cells, those cells have to

21:15

live. Then you have to sequence those

21:17

cells, you've got to wait a couple

21:19

weeks, in some cases, if you don't

21:21

have sequencing cores internally, to get that

21:23

data back. And so the feedback loop,

21:25

if you've made some, if you've made

21:27

the wrong edit using the wrong tour,

21:29

at least the most efficient tool, you

21:31

know, can be months of lost scientific

21:33

experiment time, both costly in terms of

21:35

go to market and in terms of

21:37

your research and in all the reagents

21:39

and stuff that you had to go

21:41

use in that, right. And so we're

21:43

now using AI not just for comparative

21:45

genomics for comparative genomics. But even in

21:47

what selection of what editing tool we

21:49

should use for the editing job that

21:51

we're trying to go pursue. So now

21:53

how far have you? gotten and now

21:55

I'm not I'm not

21:57

asking just about the

21:59

but you but you

22:01

can talk about the

22:03

dodo the Tasmanian tiger

22:05

anything else you've experimented

22:07

with. What have

22:09

you produced in the

22:12

lab? and is it all still in still

22:14

in do you have a do you

22:16

have a elephant elephant that There

22:18

is no a name? pregnant Asian secret

22:20

mammoth unfortunately unfortunately. be the I would

22:22

be the first be more be

22:25

more excited to share share it

22:27

with you was. was. So, so, de-extinction is a

22:29

is a right? There's There's computational

22:31

analysis, or there's If you'd

22:33

like to you'd like to continue listening to

22:35

this conversation, you'll need to subscribe

22:37

at at .org. Once Once you do, you'll

22:40

get access to all full -length episodes

22:42

of the Making of the Making Sense along

22:44

with other with other content, including

22:46

bonus episodes and AMAs, and the

22:48

conversations I've been having on

22:50

the having on the Waking The Making Sense

22:52

podcast is ad is ad-free, relies

22:54

entirely on on support. And you can

22:56

subscribe now now at Sam .org.