Latest / Core Memory / Fixing American Science Funding

Transcript

Fixing American Science Funding · Core Memory

• 0:00
  SPEAKER_02: Do you mind,
• 0:01
  SPEAKER_02: just because I haven't fully tested that mic before,
• 0:04
  SPEAKER_02: do you mind just telling me what you did yesterday in a semi?
• 0:08
  SPEAKER_01: Yesterday I was in Healdsburg,
• 0:10
  SPEAKER_01: California,
• 0:10
  and I was having some meetings and drinking some coffee and
• 0:15
  walking a dog.
• 0:15
  SPEAKER_00: You did not drink coffee.
• 0:18
  SPEAKER_01: Seeing beautiful California landscape.
• 0:20
  SPEAKER_02: So you guys were doing the Edge Esmeralda.
• 0:23
  SPEAKER_02: That's right.
• 0:26
  SPEAKER_02: Bellamy?
• 0:28
  SPEAKER_02: Can you scoot in a little bit?
• 0:30
  SPEAKER_02: Is that okay?
• 0:30
  SPEAKER_02: Yep.
• 0:33
  SPEAKER_02: And then, yeah, sorry, and tell me what you had for breakfast.
• 0:35
  SPEAKER_02: I had delicious hash browns at the place next door.
• 0:38
  I wanted to have a gluten-free crepe,
• 0:42
  but Anastasia had some orange juice.
• 0:50
  SPEAKER_02: What was your first car?
• 0:51
  SPEAKER_01: Gray Volvo.
• 0:56
  SPEAKER_01: How long did you drive it?
• 0:58
  SPEAKER_01: Only a couple of years because it broke down on the highway.
• 1:01
  SPEAKER_00: We could switch too if that would make that easier.
• 1:03
  SPEAKER_02: If that's okay.
• 1:04
  SPEAKER_02: Sorry, guys.
• 1:05
  SPEAKER_02: Nope.
• 1:05
  SPEAKER_00: We're all good.
• 1:06
  SPEAKER_01: Let's see.
• 1:07
  SPEAKER_01: Okay.
• 1:09
  SPEAKER_02: All right.
• 1:10
  SPEAKER_02: We've done it.
• 1:10
  SPEAKER_01: Do you like my gray Volvo too?
• 1:12
  SPEAKER_01: I've never seen your gray Volvo.
• 1:13
  SPEAKER_01: Okay.
• 1:16
  SPEAKER_02: Cool.
• 1:16
  SPEAKER_02: Okay.
• 1:17
  SPEAKER_02: You guys feel good?
• 1:18
  SPEAKER_02: That sounds great.
• 1:19
  SPEAKER_02: All right.
• 1:19
  SPEAKER_02: I'm going to do, I'll do like a little bit of an intro and, and,
• 1:26
  make it awkward so you have to sit there listening to me for a
• 1:29
  second okay here we go
• 1:47
  SPEAKER_02: Adam Marblestone, Anastasia Gamak.
• 1:50
  SPEAKER_02: Thank you for joining me.
• 1:52
  SPEAKER_02: How are you guys?
• 1:53
  SPEAKER_02: I'm great.
• 1:54
  SPEAKER_02: Doing really well.
• 1:55
  SPEAKER_02: Doing well.
• 1:55
  SPEAKER_02: You've been at Edge Esmeralda.
• 1:58
  For people who have no idea what that is,
• 2:00
  could you give us like the one minute
• 2:02
  SPEAKER_00: It's a pop-up village in which they did not actually build a village.
• 2:08
  It's just a normal city where a bunch of highly agentic tech
• 2:14
  weirdos move to be
• 2:16
  SPEAKER_00: each other's neighbors and hang out for a month.
• 2:19
  And like do projects and...
• 2:21
  Everything from...
• 2:22
  We just did normal meetings there.
• 2:25
  SPEAKER_00: People built solar...
• 2:28
  SPEAKER_00: water filtration systems.
• 2:30
  SPEAKER_00: Someone built a solar A-frame talks.
• 2:33
  SPEAKER_00: Audrey Tang was there, which is really exciting.
• 2:36
  And mostly I think the point is that you're walking down the
• 2:38
  street and you run
• 2:39
  SPEAKER_00: into these other people who are building and doing very cool things.
• 2:42
  Yeah,
• 2:43
  I've been hearing about this for a couple of years and now I
• 2:45
  need to go,
• 2:46
  I think.
• 2:47
  SPEAKER_02: I will try to make it this year if I can.
• 2:50
  Okay,
• 2:50
  to give people some context,
• 2:51
  you both co-founded and run Convergent Research
• 2:56
  will get into what it does but but people are going to be
• 2:59
  hearing a lot about the
• 3:01
  concept of a fro a focused research organization and what that
• 3:05
  means to science i
• 3:06
  SPEAKER_02: am going to take you out of fro land at different times as well um adam
• 3:13
  SPEAKER_02: I learned things about you last night that I didn't know.
• 3:15
  SPEAKER_02: You were a gymnast, I think, as a young person.
• 3:18
  SPEAKER_03: You found that out.
• 3:19
  SPEAKER_02: Yes,
• 3:19
  SPEAKER_02: found that out,
• 3:20
  SPEAKER_02: became a scientist,
• 3:23
  SPEAKER_02: worked at Ed Boyden's lab for a while,
• 3:25
  SPEAKER_02: worked at Google DeepMind,
• 3:27
  SPEAKER_02: worked at Kernel with Brian Johnson,
• 3:29
  SPEAKER_02: a little sort of Neuralink,
• 3:30
  SPEAKER_02: sort of consulting stint or something in there,
• 3:33
  SPEAKER_02: I think.
• 3:33
  SPEAKER_01: Very early, yeah.
• 3:35
  SPEAKER_02: And then now...
• 3:36
  have gone into um well you're just always a super fascinating
• 3:41
  person to talk to
• 3:42
  about all things science looking forward to that are pursuing
• 3:45
  it now anastasia um
• 3:47
  studied at ucla and was not like directly neuroscience or
• 3:51
  something at the the
• 3:52
  beginning i don't think no i studied astrophysics right and
• 3:55
  then i learned i don't
• 3:56
  SPEAKER_00: like math okay
• 4:00
  And you were very early at Neuralink and helped them kind of
• 4:05
  set up,
• 4:06
  get going.
• 4:07
  SPEAKER_02: And then you worked at a company that I covered.
• 4:09
  SPEAKER_02: I'm pretty sure I wrote the first story on ever,
• 4:12
  SPEAKER_02: which was Momentum Machines,
• 4:13
  SPEAKER_02: which became...
• 4:14
  SPEAKER_02: Creator.
• 4:15
  SPEAKER_02: Creator.
• 4:15
  SPEAKER_02: And they were making a robot for like burger making.
• 4:19
  SPEAKER_00: And we made robots that make cheeseburgers.
• 4:21
  SPEAKER_00: Yeah.
• 4:21
  SPEAKER_00: Highly impactful work.
• 4:23
  SPEAKER_02: It was actually awesome.
• 4:24
  SPEAKER_02: I loved cheeseburgers.
• 4:26
  SPEAKER_02: Who doesn't love cheeseburgers?
• 4:27
  I mean,
• 4:27
  it was cool and it worked,
• 4:28
  but it sort of flamed out,
• 4:29
  I think,
• 4:30
  in the end.
• 4:30
  SPEAKER_02: And then,
• 4:32
  SPEAKER_02: very famously,
• 4:33
  SPEAKER_02: I think,
• 4:33
  SPEAKER_02: for a lot of people during the COVID,
• 4:36
  SPEAKER_02: when COVID was just first kicking off,
• 4:38
  SPEAKER_02: we're at Curative,
• 4:40
  SPEAKER_02: which became really well-known for actually figuring out
• 4:43
  SPEAKER_02: testing in a hurry and doing it at a massive, massive scale.
• 4:47
  SPEAKER_02: And I think you were,
• 4:48
  SPEAKER_02: my sense when I was doing my research is I've known you,
• 4:51
  SPEAKER_02: but like just digging in,
• 4:52
  SPEAKER_02: I mean,
• 4:52
  you seem like an operations whiz because you seem to get things
• 4:55
  off the ground.
• 4:56
  SPEAKER_00: Yeah.
• 4:57
  I was in the DRC and someone called me as I was flying home and
• 5:01
  I built their
• 5:01
  SPEAKER_00: supply chain from like...
• 5:04
  500 test kits a week to a million test kits a week in about
• 5:07
  three months three
• 5:08
  months which like no government organization seems remotely
• 5:12
  capable of doing so
• 5:14
  well done and then and then you both obviously ended up um
• 5:17
  working on on convergent
• 5:19
  together um okay so that's the intro i will briefly thank ewan
• 5:23
  ventures for making
• 5:24
  SPEAKER_02: this podcast possible um they've been a big
• 5:27
  supporter of the show thank you and then off we go um there's a
• 5:30
  ton of things i
• 5:31
  want to ask you guys about i think we should i feel bad because
• 5:34
  you have to explain
• 5:35
  froze all the time and you've had to do this for years but um i
• 5:38
  do think we should
• 5:38
  give the audience like a just some context on what this idea is
• 5:43
  i know you've done
• 5:43
  it a hundred times but i there are there are people really who
• 5:47
  still don't really
• 5:48
  know this concept um yeah who wants to take a crack well i
• 5:51
  gotta go first sure okay
• 5:53
  SPEAKER_00: um
• 5:55
  So there's this notion that maybe we prematurely standardized
• 6:00
  on how we do R&D in the US.
• 6:02
  SPEAKER_00: There are academic labs.
• 6:04
  SPEAKER_00: There is industry and startups.
• 6:07
  SPEAKER_00: Maybe there's DARPA.
• 6:08
  SPEAKER_00: That's basically how we do science and technology.
• 6:12
  There are a handful of problems that are not easily solved
• 6:15
  within those institutions.
• 6:17
  Things that need a multidisciplinary team working together over
• 6:22
  a certain length of
• 6:23
  SPEAKER_00: time towards a very focused goal where...
• 6:27
  SPEAKER_00: There isn't necessarily a for-profit mechanism on the other side,
• 6:30
  SPEAKER_00: and it doesn't necessarily fit into publishing or academic incentives.
• 6:35
  And so Adam and Sam Rodriguez came up with this concept called
• 6:39
  a focused research
• 6:40
  SPEAKER_00: organization,
• 6:41
  SPEAKER_00: which is special built to solve just that problem.
• 6:44
  If you identify a fundamental capability that we should build
• 6:50
  to spin up a new
• 6:51
  SPEAKER_00: organization only to tackle that goal.
• 6:53
  And these are like big things that could crack fields wide open
• 6:58
  and really advance
• 6:59
  SPEAKER_02: them in new ways where,
• 7:00
  SPEAKER_02: and I mean,
• 7:01
  SPEAKER_02: you just said it,
• 7:02
  but sometimes a company might not be willing to spend the money
• 7:06
  because you can't
• 7:07
  SPEAKER_02: tell when you're going to make money from this thing.
• 7:08
  SPEAKER_02: A university,
• 7:09
  it might be too much money for them to like really go all in on
• 7:13
  this wild idea and
• 7:14
  SPEAKER_02: see where it goes.
• 7:15
  SPEAKER_02: And so, and then you put them on, you do try to contain it.
• 7:18
  SPEAKER_02: I mean, they get a fair bit of money, but you put kind of this five-year
• 7:22
  SPEAKER_02: time limit around it,
• 7:23
  SPEAKER_02: I'm sure,
• 7:23
  SPEAKER_02: so that these academics are just sitting there forever,
• 7:27
  SPEAKER_02: or engineers trying to make everything absolutely perfect,
• 7:30
  SPEAKER_02: but to get these ideas off the ground and make more ideas.
• 7:36
  SPEAKER_00: Yeah.
• 7:37
  SPEAKER_00: Something like, we know the thing we want to build.
• 7:41
  Going in,
• 7:41
  we're pretty convinced we can do it in five to six years with
• 7:44
  $30 to $50 billion.
• 7:45
  SPEAKER_00: And, you know,
• 7:49
  SPEAKER_00: Being at early stage startups,
• 7:50
  SPEAKER_00: there is like an existential forcing function to get there,
• 7:54
  SPEAKER_00: that there isn't necessarily at other larger institutions.
• 7:58
  SPEAKER_00: So mimicking that kind of existential,
• 8:00
  SPEAKER_00: I have to build this thing,
• 8:02
  SPEAKER_00: I have to ship a product,
• 8:04
  SPEAKER_00: I think is what we're trying to do with this time-limited nature.
• 8:09
  SPEAKER_02: I don't know how sensitive this is.
• 8:10
  I hope it's not too sensitive,
• 8:12
  but I know Eric Schmidt has publicly been a big backer.
• 8:15
  SPEAKER_02: James Fickle, I see on a lot of the companies that you guys are backing.
• 8:19
  SPEAKER_02: Thank you to James, helper to many people.
• 8:23
  SPEAKER_02: So people who are interested in science,
• 8:25
  SPEAKER_02: seeing science flourish,
• 8:26
  SPEAKER_02: seeing big ideas flourish,
• 8:28
  SPEAKER_02: are willing to put their money behind this.
• 8:31
  I've written about a few of these companies and the ones that
• 8:35
  I've gravitated
• 8:36
  SPEAKER_02: towards are in the brain neuroscience area.
• 8:38
  So E11 is trying to build a more detailed,
• 8:43
  richer map of the brain by using viruses to
• 8:48
  SPEAKER_02: to put colors on neurons so we can trace their paths.
• 8:53
  SPEAKER_02: Forrest Neuro has a brain implant using,
• 8:57
  SPEAKER_02: I think,
• 8:57
  ultrasound to see a bit deeper into the brain than some of the
• 9:01
  current implants
• 9:01
  SPEAKER_02: from Neuralink and others.
• 9:03
  And then I know I visited you guys up near Boston and there's
• 9:07
  other stuff around
• 9:08
  SPEAKER_02: what,
• 9:08
  SPEAKER_02: like enzymes and I don't know.
• 9:10
  SPEAKER_02: I get focused on the brain stuff.
• 9:13
  SPEAKER_02: What is the other scope of what you guys are interested in?
• 9:17
  SPEAKER_01: Yeah,
• 9:17
  SPEAKER_01: often,
• 9:18
  SPEAKER_01: I mean,
• 9:19
  SPEAKER_01: there's a broader frame which we can come back to,
• 9:20
  which is sort of what are the most important missing
• 9:23
  fundamental capabilities.
• 9:25
  SPEAKER_01: We think a lot about the tech tree.
• 9:26
  SPEAKER_01: But basically, often we're making a tool system or data set which makes
• 9:32
  catalytically makes downstream work in some area dramatically
• 9:35
  easier or faster.
• 9:37
  SPEAKER_01: So one of the projects that you saw in Boston was called PTI.
• 9:41
  SPEAKER_01: It's on proteomics.
• 9:43
  So proteins are kind of like the little machines and cells that
• 9:46
  do most of the work.
• 9:47
  SPEAKER_01: Most of our technology in biology focuses on the genome.
• 9:52
  SPEAKER_01: nucleic acids, so DNA or RNA.
• 9:55
  And so this is a project that's trying to make proteomics about
• 9:58
  100 to 1,000 times
• 9:59
  SPEAKER_01: cheaper per sample.
• 10:01
  SPEAKER_01: So that way you could do many,
• 10:03
  SPEAKER_01: many samples,
• 10:04
  SPEAKER_01: which you might want to do,
• 10:05
  SPEAKER_01: for example,
• 10:06
  SPEAKER_01: if each sample is a single cell,
• 10:08
  you might want to look at many different single cells to
• 10:09
  understand how they're
• 10:10
  SPEAKER_01: different.
• 10:11
  SPEAKER_01: And so,
• 10:12
  SPEAKER_01: so far,
• 10:12
  SPEAKER_01: PTI has increased the cost performance of proteomics by a factor of 27,
• 10:18
  SPEAKER_01: and they're hoping over the five years to do 100 to 1,000.
• 10:22
  SPEAKER_02: And it's super cool.
• 10:23
  And I'm forgetting the exact name of it,
• 10:24
  but you just put this map of unsolved problems.
• 10:28
  SPEAKER_02: Is that the... Of gaps.
• 10:29
  SPEAKER_02: Of gaps.
• 10:29
  SPEAKER_02: We have a gap map.
• 10:30
  SPEAKER_02: The gap map, which is incredible.
• 10:32
  SPEAKER_02: It's on your website, Convergent Research.
• 10:33
  SPEAKER_02: It's really cool.
• 10:34
  I mean,
• 10:34
  you very specifically outline in all these different fields
• 10:38
  what needs to be done.
• 10:40
  SPEAKER_02: I would fully recommend that people go check it out.
• 10:42
  SPEAKER_02: It just got my mind.
• 10:43
  SPEAKER_02: It gave me just free story ideas, really, of people to go explore.
• 10:48
  SPEAKER_02: And so...
• 10:50
  SPEAKER_02: When you list that out,
• 10:51
  SPEAKER_02: I mean,
• 10:51
  SPEAKER_02: you are doing research either adjacent or in some of those areas,
• 10:56
  SPEAKER_02: but this was just stuff that you guys wished people would also tackle.
• 11:00
  SPEAKER_01: Yeah, it's interesting.
• 11:00
  SPEAKER_01: It exists across all fields.
• 11:02
  SPEAKER_01: I think the gap map is sort of a digest of what we've
• 11:07
  downloaded by talking to hundreds and hundreds of scientists
• 11:10
  across different
• 11:11
  SPEAKER_01: fields,
• 11:11
  SPEAKER_01: scientists and engineers.
• 11:12
  SPEAKER_01: I mean,
• 11:13
  SPEAKER_01: at the highest level,
• 11:15
  I think what makes our approach unique is actually something
• 11:21
  that it seems fairly
• 11:22
  SPEAKER_01: obvious,
• 11:23
  SPEAKER_01: but it's just the recognition that
• 11:27
  We have capabilities,
• 11:28
  like let's say making computers as humans,
• 11:31
  making computers.
• 11:32
  SPEAKER_01: Those capabilities depend on more fundamental capabilities.
• 11:35
  SPEAKER_01: This is the idea of a technology tree,
• 11:37
  SPEAKER_01: but I really think of it as a technology and science tree.
• 11:42
  The ability to make computer chips depends on the ability to
• 11:44
  process silicon and
• 11:45
  SPEAKER_01: understand solid state physics.
• 11:47
  Doing those things depend on being able to see the atomic
• 11:49
  structure of materials
• 11:50
  SPEAKER_01: and things like that.
• 11:51
  And so as you kind of go further and further back towards if
• 11:55
  you want the root or
• 11:56
  the trunk of this technology tree or technology and science
• 11:59
  tree of more and more
• 12:00
  SPEAKER_01: fundamental capabilities,
• 12:03
  SPEAKER_01: our everyday products and services are sort of the leaves of this tree,
• 12:05
  SPEAKER_01: right?
• 12:09
  SPEAKER_01: As you go further back in that tree,
• 12:12
  our main hypothesis is that you can actually just identify what
• 12:16
  the missing
• 12:17
  fundamental capabilities are that would be kind of thick
• 12:20
  branches on this tree,
• 12:23
  SPEAKER_01: whereas most research
• 12:24
  SPEAKER_01: It's more just thinking forward, right?
• 12:27
  SPEAKER_01: It's like, what would be the next interesting thing to do?
• 12:29
  SPEAKER_01: What would be the next curiosity-driven thing to do?
• 12:32
  SPEAKER_01: Or it's thinking very economically driven,
• 12:33
  what would be the next product or service that I can make that
• 12:35
  I can get money for?
• 12:36
  SPEAKER_01: We just say,
• 12:37
  SPEAKER_01: okay,
• 12:37
  SPEAKER_01: well,
• 12:38
  let's look at these different missing spots on the tech tree
• 12:41
  where we could sprout
• 12:42
  SPEAKER_01: a new branch.
• 12:44
  those more fundamental capabilities,
• 12:46
  we think these are identifiable by road mapping.
• 12:48
  SPEAKER_01: But road mapping also just means going and talking to people and saying,
• 12:51
  what would be the capabilities that would change your fields if
• 12:54
  you somehow could
• 12:55
  get the money and the team and the organization and the
• 12:57
  operations to build that
• 12:58
  SPEAKER_01: capability?
• 12:59
  SPEAKER_01: And so the gap map is kind of just,
• 13:01
  SPEAKER_01: we've gone and asked lots of people that question,
• 13:03
  SPEAKER_01: and it's a way of organizing it by thinking about research gaps.
• 13:06
  SPEAKER_01: But really, the capabilities is kind of the essence of it.
• 13:09
  Sometimes people ask us what tools or processes we use to do
• 13:15
  roadmapping.
• 13:17
  And I will say that is kindness and networking and asking
• 13:22
  different questions.
• 13:23
  SPEAKER_00: Did you say kindness and networking?
• 13:25
  SPEAKER_02: Yeah.
• 13:26
  SPEAKER_02: Yes.
• 13:26
  SPEAKER_02: Okay.
• 13:27
  I really think that there's an extent to which Adam has
• 13:30
  traversed a very large
• 13:32
  network of scientists very generously and asked them a
• 13:35
  different set of questions
• 13:36
  SPEAKER_00: than they normally get asked.
• 13:38
  SPEAKER_00: Instead of saying, like, oh, what are you doing next?
• 13:41
  SPEAKER_00: What is cool for you?
• 13:42
  SPEAKER_00: What are you interested in?
• 13:43
  SPEAKER_00: How does this work?
• 13:44
  SPEAKER_00: He's like, well, what is the, like...
• 13:45
  SPEAKER_00: fundamental thing you wish could be built.
• 13:47
  SPEAKER_00: What would happen in a different system?
• 13:49
  SPEAKER_00: What is missing from this field?
• 13:51
  SPEAKER_00: And oftentimes people aren't asking themselves this question,
• 13:54
  SPEAKER_00: and so they don't really have an answer.
• 13:56
  SPEAKER_00: But then they go away and they think about it,
• 13:58
  SPEAKER_00: and they're like,
• 13:58
  SPEAKER_00: oh,
• 13:58
  SPEAKER_00: actually,
• 13:59
  SPEAKER_00: this is like,
• 14:00
  SPEAKER_00: if my lab works differently or if the system worked differently,
• 14:03
  SPEAKER_00: this is what I wish existed.
• 14:05
  SPEAKER_00: This would allow me to do all of these downline things.
• 14:08
  SPEAKER_00: And by going and asking that over and over and over again,
• 14:12
  SPEAKER_00: we've collected information that then we put out via the gap map.
• 14:16
  SPEAKER_00: I don't think we knew that we were building that in the first place.
• 14:19
  SPEAKER_00: Adam was talking to cool people about cool things.
• 14:22
  And it came out as this map of fundamental capabilities that
• 14:27
  then would solve these
• 14:29
  SPEAKER_00: gaps in science and technology.
• 14:30
  Yeah,
• 14:31
  it's just cool to have somebody focused on doing something like
• 14:33
  that.
• 14:33
  SPEAKER_02: It's funny,
• 14:35
  SPEAKER_02: just as you were talking about the gap map,
• 14:37
  SPEAKER_02: and I hadn't thought of this when I first saw it.
• 14:39
  SPEAKER_02: I remember when I was working on the
• 14:43
  SPEAKER_02: This must have been like 2014 or 20, somewhere vaguely around there.
• 14:47
  SPEAKER_02: There was this guy, Jonathan Huebner.
• 14:50
  SPEAKER_02: He published this paper that was kind of controversial,
• 14:54
  SPEAKER_02: I guess,
• 14:54
  SPEAKER_02: at the time,
• 14:55
  SPEAKER_02: but also I felt like people were mostly agreeing with it.
• 14:57
  SPEAKER_02: He was arguing that most of the big ideas had already been found.
• 15:01
  And Peter Thiel really jumped on this at the time and was like,
• 15:04
  yeah,
• 15:04
  he's right.
• 15:05
  SPEAKER_02: And people were depressed.
• 15:06
  SPEAKER_02: And then now I'm just bubbling through my head.
• 15:10
  SPEAKER_02: I mean, it feels like we're in, maybe it's real, maybe it's not.
• 15:12
  SPEAKER_02: This is what I want to ask you.
• 15:13
  I feel like we're in such a different place because even if you
• 15:16
  just look at AI,
• 15:17
  SPEAKER_02: everyone's like,
• 15:17
  SPEAKER_02: oh,
• 15:17
  SPEAKER_02: this is going to unlock all these things.
• 15:19
  SPEAKER_02: With the gap map, there's very clear things, specific stuff that...
• 15:23
  SPEAKER_02: you're pointing to that could open up new fields,
• 15:25
  SPEAKER_02: the fro idea in of itself is sort of- I've never really grokked fully.
• 15:32
  SPEAKER_01: I think you can talk about productivity of different systems of science.
• 15:35
  I do think that this institutional question is key and the
• 15:39
  question of what sorts
• 15:39
  SPEAKER_01: of activities can you actually do in order to unlock new ideas.
• 15:43
  SPEAKER_01: And sometimes it's not really about an actually new idea,
• 15:47
  SPEAKER_01: if that makes sense,
• 15:48
  SPEAKER_01: or that's not the limiting factor,
• 15:50
  SPEAKER_01: right?
• 15:50
  SPEAKER_01: So there's much research that is limited by new ideas.
• 15:54
  I think a lot of academic research is limited by new ideas and
• 15:57
  creativity,
• 15:58
  SPEAKER_01: maybe fundamental math research.
• 15:59
  SPEAKER_01: We can come back to that.
• 16:00
  SPEAKER_01: We have a FRO related to math,
• 16:01
  SPEAKER_01: but a lot of fundamental math research,
• 16:03
  SPEAKER_01: yeah,
• 16:04
  SPEAKER_01: maybe related to ideas.
• 16:05
  But if you think about the Hubble Space Telescope,
• 16:07
  for example,
• 16:09
  it's kind of obvious that you
• 16:12
  SPEAKER_01: if you want to boost the idea-generating capacity of astronomy,
• 16:15
  SPEAKER_01: you need a bigger mirror in space that collects more light.
• 16:19
  SPEAKER_01: And that,
• 16:19
  SPEAKER_01: you know,
• 16:20
  SPEAKER_01: by that time you could have a CCD camera and so on,
• 16:22
  SPEAKER_01: and you could beam it back to Earth,
• 16:23
  SPEAKER_01: right?
• 16:24
  SPEAKER_01: And so a big mirror in space is sort of...
• 16:28
  almost obvious that that would be a catalytic boost to the
• 16:32
  field of astronomy,
• 16:33
  SPEAKER_01: but how are you going to go build that?
• 16:35
  SPEAKER_01: Who's going to provide the money?
• 16:36
  SPEAKER_01: Who's going to run the project?
• 16:37
  SPEAKER_01: How are you going to actually get it into space?
• 16:40
  I think our observation in part is that there's a lot of things
• 16:43
  like that,
• 16:43
  SPEAKER_01: not necessarily as big as the Hubble Space Telescope.
• 16:47
  The way astronomers do it is they actually have this decadal
• 16:50
  survey process where
• 16:51
  SPEAKER_01: the astronomers all get together and they say,
• 16:53
  SPEAKER_01: what are the few things that we want to ask the NSF and NASA and so on?
• 16:56
  SPEAKER_01: The National Academies of Sciences coordinates some of this process.
• 16:59
  SPEAKER_01: But in other fields, there isn't a decadal survey.
• 17:03
  And also the projects you might need,
• 17:04
  they're not necessarily billion-dollar megaprojects.
• 17:07
  SPEAKER_01: Even getting 20 or 30 people to coordinate and say,
• 17:10
  SPEAKER_01: if we had this missing capability,
• 17:12
  SPEAKER_01: we could boost that.
• 17:13
  SPEAKER_01: So,
• 17:13
  SPEAKER_01: yeah,
• 17:15
  I just kind of naturally just see the world in terms of these
• 17:17
  missing capabilities,
• 17:18
  SPEAKER_01: some of which are actually fairly straightforward,
• 17:21
  SPEAKER_01: but somehow nobody's doing them.
• 17:24
  SPEAKER_01: And so why is that?
• 17:25
  SPEAKER_01: Is that institutional or financial?
• 17:27
  SPEAKER_01: Yes, it's not really intellectual.
• 17:30
  SPEAKER_01: There is a chicken and egg problem, though, which is interesting.
• 17:32
  I mean,
• 17:32
  so Tom Khalil,
• 17:33
  who's our board chair,
• 17:35
  often talks about,
• 17:36
  you know,
• 17:37
  SPEAKER_01: you don't spend a lot of time,
• 17:39
  SPEAKER_01: um,
• 17:39
  SPEAKER_01: planning for what would happen if you won the lottery,
• 17:41
  SPEAKER_01: you know,
• 17:41
  SPEAKER_01: exactly how would you spend the money?
• 17:43
  SPEAKER_01: And similarly,
• 17:44
  SPEAKER_01: if you're a scientist and you're thinking,
• 17:45
  SPEAKER_01: well,
• 17:46
  SPEAKER_01: to,
• 17:47
  SPEAKER_01: to get my next grant,
• 17:48
  I need to have a project that's roughly the size and shape of a
• 17:51
  couple of graduate
• 17:52
  SPEAKER_01: students and postdocs in my lab.
• 17:54
  SPEAKER_01: Um,
• 17:56
  that's the type of thing you're mostly thinking about the
• 17:59
  missing capabilities are,
• 18:00
  SPEAKER_01: are,
• 18:01
  SPEAKER_01: are that size and shape.
• 18:02
  SPEAKER_01: Right.
• 18:03
  SPEAKER_01: Um, but, uh,
• 18:06
  If someone came to you and said,
• 18:07
  OK,
• 18:07
  well,
• 18:08
  here,
• 18:09
  think for a while about this.
• 18:11
  SPEAKER_01: You can be the CEO of a 30 or 50-person engineering-focused,
• 18:16
  almost like a nonprofit deep tech startup to develop any
• 18:19
  capability that your field
• 18:20
  SPEAKER_01: really needs.
• 18:24
  or at least some people would actually come up with what would
• 18:26
  be those ideas,
• 18:27
  SPEAKER_01: but they're not being asked that.
• 18:28
  SPEAKER_01: But similarly,
• 18:29
  if you're a funder and people aren't sending you those ideas
• 18:32
  because no one ever
• 18:32
  SPEAKER_01: asked them and what would be the point,
• 18:34
  SPEAKER_01: you wouldn't plan how you won the lottery if no one was asking you,
• 18:38
  SPEAKER_01: how would you spend the money?
• 18:40
  SPEAKER_00: I think one version of this is even if you think about PTI,
• 18:43
  SPEAKER_00: if you were a grad student or a postdoc in Nikolai's lab,
• 18:47
  SPEAKER_00: you'd be like,
• 18:48
  SPEAKER_00: great,
• 18:48
  SPEAKER_00: what is the next piece?
• 18:49
  SPEAKER_00: Like,
• 18:50
  SPEAKER_00: okay,
• 18:50
  SPEAKER_00: now I do more mass tags,
• 18:52
  SPEAKER_00: or now I do a different type of protein or a different type of imaging,
• 18:55
  SPEAKER_00: and then...
• 18:56
  SPEAKER_00: That's how you think about the problem.
• 18:58
  SPEAKER_00: Instead of, I need to make single-cell proteomics 100 times cheaper.
• 19:02
  SPEAKER_00: How do I do that?
• 19:03
  SPEAKER_00: The answer involves $44 million.
• 19:05
  SPEAKER_00: That's not actually how they think about solving the problem.
• 19:09
  And so by giving them a different problem to solve,
• 19:12
  you get a different answer.
• 19:13
  SPEAKER_02: So you guys see this as just because of that,
• 19:17
  SPEAKER_02: this angle of attack is very different to,
• 19:20
  if I think historically of Xerox PARC or Bell Labs or the
• 19:24
  university system or
• 19:28
  SPEAKER_02: company R&D Labs or even things like Janelia Farms,
• 19:33
  SPEAKER_02: Howard Hughes.
• 19:33
  SPEAKER_02: Yeah.
• 19:35
  So there are places that have historically allowed for this
• 19:37
  open-ended,
• 19:38
  well-funded research.
• 19:40
  The United States obviously has funded science quite
• 19:43
  successfully in many ways for a long time.
• 19:45
  But it's this intention,
• 19:47
  this angle of attack that you're doing that you feel like is
• 19:49
  different.
• 19:50
  I think there are some similarities to what some of those
• 19:52
  organizations have done
• 19:53
  SPEAKER_01: in at least parts of their research.
• 19:55
  SPEAKER_01: So Bell Labs had a sort of fundamental development track.
• 19:58
  SPEAKER_01: We also like this term fundamental development.
• 20:00
  SPEAKER_01: It's not just research on new ideas, but you need to really...
• 20:04
  SPEAKER_01: do the engineering to make it possible to use that as a capability,
• 20:08
  SPEAKER_01: whether in further research or in something more commercial.
• 20:12
  SPEAKER_01: Janelia,
• 20:12
  SPEAKER_01: HHMI,
• 20:14
  SPEAKER_01: this research campus in Virginia,
• 20:15
  some of the work that they do looks more like open-ended
• 20:17
  academic research,
• 20:19
  SPEAKER_01: but some of it
• 20:20
  SPEAKER_01: They did really great work on the fly connectome, map of the fly brain.
• 20:25
  SPEAKER_01: That looks more like this directed missing capabilities development.
• 20:29
  SPEAKER_01: But there's only a few such institutions.
• 20:30
  There's not really a general mechanism either for finding what
• 20:33
  would be the most
• 20:34
  SPEAKER_01: important ones or for saying,
• 20:36
  SPEAKER_01: okay,
• 20:37
  SPEAKER_01: what if you don't work at Janelia?
• 20:38
  SPEAKER_01: What if you don't work at the Allen Institute?
• 20:40
  How do you actually create a mechanism to identify and then
• 20:44
  build against one of
• 20:45
  SPEAKER_01: those fundamental capabilities?
• 20:47
  SPEAKER_01: Yeah.
• 20:48
  SPEAKER_02: And I don't mean this to sound as loaded as it probably will come off,
• 20:52
  SPEAKER_02: but if I,
• 20:54
  SPEAKER_02: as an outside observer of China,
• 20:56
  SPEAKER_02: you know,
• 20:56
  SPEAKER_02: it,
• 20:56
  SPEAKER_02: it seems like they,
• 20:58
  SPEAKER_02: they think a lot in some of these terms.
• 21:01
  SPEAKER_02: And it's like, these are areas we care about.
• 21:03
  SPEAKER_02: We care about quantum computing.
• 21:04
  SPEAKER_02: We're not just going to fund one quantum computing effort.
• 21:06
  We'll,
• 21:06
  we'll fund 10,
• 21:07
  you know,
• 21:08
  on every approach and then whatever shakes out,
• 21:11
  shakes out.
• 21:12
  SPEAKER_02: Um, am I reading?
• 21:16
  SPEAKER_02: Yeah.
• 21:17
  SPEAKER_01: The cultural, what Anastasia said early on of sort of the U.S.
• 21:20
  SPEAKER_01: has prematurely standardized,
• 21:21
  and I think that Western Europe and so on maybe somewhat is in
• 21:24
  a similar mold,
• 21:25
  SPEAKER_01: but the U.S.
• 21:26
  SPEAKER_01: has maybe prematurely standardized on sort of scaling up this,
• 21:29
  SPEAKER_01: what was originally Vannevar Bush,
• 21:31
  SPEAKER_01: kind of heavily university academic PI based,
• 21:33
  plus national labs and a few other commercial companies and so
• 21:37
  on mechanisms.
• 21:38
  And China has a different,
• 21:40
  I don't know in that much detail,
• 21:42
  but they do have a different
• 21:43
  SPEAKER_01: set of maybe strengths and weaknesses, right?
• 21:45
  SPEAKER_01: So if they want to do focus capability development,
• 21:47
  SPEAKER_01: let's say for the next generation of computer hardware,
• 21:49
  SPEAKER_01: new kinds of semiconductors,
• 21:50
  SPEAKER_01: quantum computers,
• 21:52
  it may be easier from an incentive structure and funding
• 21:55
  perspective in their
• 21:56
  SPEAKER_01: system to do that compared to ours.
• 21:58
  SPEAKER_01: Yeah.
• 21:58
  SPEAKER_00: To do something like roadmap,
• 22:00
  SPEAKER_00: what are the top five potential fundamental capabilities here,
• 22:03
  and then just decide to allocate resources towards those top
• 22:06
  five and see what
• 22:07
  SPEAKER_00: happens,
• 22:08
  SPEAKER_00: I think is actually quite tough in the US right now.
• 22:11
  yeah yes okay which is leading me to another question i want to
• 22:14
  ask if you know no
• 22:15
  idea actually if this gets you guys in hot water or not um
• 22:19
  because i i don't fully
• 22:20
  know where all the complexity lies but but as we look at what's
• 22:24
  kind of been
• 22:24
  happening to science funding over the last few months um i
• 22:30
  talked to people
• 22:32
  SPEAKER_02: who I think are fairly rational and actually all pro-science.
• 22:36
  SPEAKER_02: And they actually have mixed opinions.
• 22:39
  SPEAKER_02: I mean,
• 22:40
  there's a whole camp of people I talk to who are big science
• 22:43
  funders who are like,
• 22:44
  SPEAKER_02: you know what?
• 22:45
  SPEAKER_02: The NIH is getting what it,
• 22:47
  SPEAKER_02: deserves, the universities are getting what they deserve.
• 22:49
  SPEAKER_02: They were bloated.
• 22:50
  I know this sucks,
• 22:51
  but something's gotta be done and this is the change required.
• 22:54
  SPEAKER_02: And then obviously there's other people who are just like,
• 22:56
  SPEAKER_02: this is,
• 22:56
  SPEAKER_02: this is horrible or this is gonna have,
• 23:00
  SPEAKER_02: you know,
• 23:00
  SPEAKER_02: that may even be the case,
• 23:01
  SPEAKER_02: but we're gonna have lasting effects from what we're,
• 23:03
  SPEAKER_02: we're doing.
• 23:04
  And we're also scaring away,
• 23:06
  um,
• 23:06
  the best of Brian's people in the process.
• 23:08
  SPEAKER_02: So, um, yeah, you guys seem like very well positioned.
• 23:11
  SPEAKER_01: I'll start, I'll start with a couple of things.
• 23:12
  SPEAKER_01: So, so one, one thing to note is that we, for
• 23:16
  SPEAKER_01: saw this gap,
• 23:19
  this gap in this sort of mid-scale fundamental capabilities
• 23:21
  development and
• 23:23
  SPEAKER_01: identification,
• 23:23
  SPEAKER_01: if you want FRO-shaped problems,
• 23:26
  SPEAKER_01: right?
• 23:27
  SPEAKER_01: Even we started doing this back in like 2020, right?
• 23:31
  SPEAKER_01: Previous administration.
• 23:34
  SPEAKER_01: This was a pre-existing problem.
• 23:36
  SPEAKER_01: So thinking of this as a gap that needs to be filled or as an additional
• 23:40
  SPEAKER_01: diversification of the research enterprise or ecosystem,
• 23:45
  SPEAKER_01: that pre-exists any of this stuff.
• 23:47
  SPEAKER_01: So our original intention certainly has been to add something new.
• 23:54
  SPEAKER_01: The more mechanisms,
• 23:54
  SPEAKER_01: the better,
• 23:55
  SPEAKER_01: because I actually do think that university research is actually really,
• 23:59
  SPEAKER_01: really pivotal.
• 24:01
  To my mind,
• 24:02
  there's no other way to do certain types of things that
• 24:05
  universities do.
• 24:06
  So imagine that your goal,
• 24:07
  imagine I told you there was a scientific field and
• 24:11
  SPEAKER_01: In order to make progress in it,
• 24:12
  you need to spend a long time kind of almost cloning the
• 24:15
  thought patterns of an
• 24:15
  SPEAKER_01: expert in that field,
• 24:17
  while at the same time being really strongly incentivized to
• 24:21
  create your own
• 24:22
  SPEAKER_01: diverse angle on it,
• 24:23
  SPEAKER_01: right?
• 24:23
  SPEAKER_01: Yeah.
• 24:25
  Think about Jeff Hinton's students that went off to start these
• 24:29
  AI companies.
• 24:31
  SPEAKER_01: They kind of cloned Jeff Hinton's brain and made their own version.
• 24:34
  SPEAKER_01: That's academic research.
• 24:35
  SPEAKER_01: That's what it's like to get a PhD in a research field.
• 24:38
  SPEAKER_01: And I think that's a really important part of the system.
• 24:41
  SPEAKER_01: So the FROs don't solve...
• 24:43
  SPEAKER_01: that.
• 24:44
  They saw this directed development of these missing,
• 24:47
  almost more engineerable capabilities.
• 24:50
  SPEAKER_01: But we still think that the university system is really important.
• 24:53
  SPEAKER_01: On the other hand,
• 24:55
  SPEAKER_01: at large,
• 24:55
  SPEAKER_01: I mean,
• 24:55
  except for this philanthropic intervention and kind of manually
• 24:59
  going to funders
• 25:00
  SPEAKER_01: and saying,
• 25:00
  SPEAKER_01: hey,
• 25:01
  SPEAKER_01: will you help us out here?
• 25:03
  SPEAKER_01: The system really hasn't
• 25:06
  made a fro type mechanism so the change that's going on may be
• 25:12
  an instability that
• 25:13
  allows more change at the federal level yeah okay so okay that
• 25:18
  is a um possibly
• 25:20
  optimistic take on things but but but writ large i mean you get
• 25:24
  to talk to so many
• 25:25
  scientists and and you probably i know we know a lot of the
• 25:28
  same people so um
• 25:30
  Yeah,
• 25:30
  I mean,
• 25:30
  is there any sense to what has been going on,
• 25:34
  or where do you fall on that?
• 25:37
  SPEAKER_00: I've spent a lot of time with metascience.
• 25:41
  SPEAKER_00: I've learned that there's two versions of metascience.
• 25:42
  There's metascience that happens by analyzing citations and
• 25:47
  describing the way that
• 25:47
  SPEAKER_00: science happens.
• 25:48
  And then there's like this,
• 25:49
  Ben Reinhart called it once,
• 25:51
  metascience pirates.
• 25:52
  SPEAKER_00: And Michael Nielsen calls it metascience entrepreneurs.
• 25:55
  SPEAKER_00: But people running experiments.
• 25:57
  I think that it was really hard to run experiments like golden
• 26:01
  tickets at the NSF.
• 26:02
  SPEAKER_00: or lottery-based grant allocation.
• 26:08
  And I think that those things should be easier to run
• 26:12
  experiments about how do we
• 26:15
  SPEAKER_00: make decisions about allocating funding,
• 26:16
  SPEAKER_00: about doubling down on funding,
• 26:18
  SPEAKER_00: and the types of projects that people can work on.
• 26:21
  SPEAKER_00: There are certainly a lot of people who say things like,
• 26:23
  SPEAKER_00: I couldn't get my PhD work funded today that I did in the 80s or the 70s.
• 26:29
  SPEAKER_00: And so I think that there is some amount of needing to encourage
• 26:33
  SPEAKER_00: innovation in how science is chosen to be done.
• 26:37
  SPEAKER_00: It feels amazing to spend time in the U.S.
• 26:42
  with the best scientists in the world who all want to work
• 26:45
  together and want to
• 26:46
  SPEAKER_00: come together and feel comfortable traveling in and out of the U.S.
• 26:49
  SPEAKER_00: and bringing as much talent into one dense ecosystem.
• 26:55
  And I think that it would be a real shame if we ended up losing
• 26:58
  that really dense
• 26:59
  SPEAKER_00: talent ecosystem in science and technology.
• 27:02
  SPEAKER_02: Right.
• 27:02
  SPEAKER_02: And then do you think there's any,
• 27:06
  SPEAKER_02: just because you two do study this and these types of organizations,
• 27:10
  SPEAKER_02: is there validity to the critiques on the overhead at these universities?
• 27:18
  SPEAKER_02: that I feel like we're Elon Musk-ing science right now.
• 27:21
  And it's kind of like,
• 27:22
  well,
• 27:22
  you guys need to clean up your act,
• 27:24
  deal with this overhead thing.
• 27:25
  SPEAKER_01: There are certainly problems.
• 27:27
  There are certainly problems in the existing system,
• 27:30
  major inefficiencies,
• 27:31
  but I think
• 27:35
  You have to be really thoughtful in terms of surgical
• 27:38
  interventions because you
• 27:40
  don't want to throw the baby out with the bathwater in the
• 27:42
  process of doing that.
• 27:42
  SPEAKER_01: A lot of scientists are really struggling right now.
• 27:45
  That doesn't mean the system is perfect and it doesn't mean it
• 27:47
  doesn't need major
• 27:48
  SPEAKER_01: catalysts for structural change.
• 27:50
  But I don't think just decreasing the amount is the way to get
• 27:54
  that structural change.
• 27:56
  SPEAKER_00: I have one question.
• 27:57
  One thing that annoys me about the word overhead is that
• 28:01
  actually often captured in
• 28:02
  an overhead number is something like the machine that everybody
• 28:05
  uses or like the
• 28:07
  SPEAKER_00: air handlers that allow you to do BSL-2 work in your lab.
• 28:13
  SPEAKER_00: And that's captured in overhead.
• 28:15
  SPEAKER_00: but it is essential for the project.
• 28:17
  SPEAKER_00: And so I think that actually the way that we describe things,
• 28:20
  SPEAKER_00: the way that billing happens,
• 28:22
  SPEAKER_00: the way that even like grassmaking happens,
• 28:23
  because you can't actually get a grant to support this like
• 28:26
  shared piece of
• 28:26
  SPEAKER_00: equipment.
• 28:27
  SPEAKER_00: A grant has to go towards equipment that is dedicated to the project.
• 28:31
  SPEAKER_00: And so therefore shared equipment has to go into overhead.
• 28:34
  Like the system makes it very opaque to answer this question to
• 28:38
  the point where I
• 28:39
  SPEAKER_00: don't know the answer to your question.
• 28:41
  SPEAKER_00: They don't understand the system.
• 28:42
  SPEAKER_02: Yeah,
• 28:43
  SPEAKER_02: I just,
• 28:43
  SPEAKER_02: you know,
• 28:44
  last week I went to talk to a group of Stanford aerospace
• 28:47
  students and they were
• 28:49
  SPEAKER_02: really sweet.
• 28:50
  SPEAKER_02: They just took me on a tour of the lab afterwards.
• 28:53
  SPEAKER_02: It was like, oh, this professor, all his programs have been shut down.
• 28:57
  SPEAKER_02: This professor, all his programs have been shut down.
• 28:59
  SPEAKER_02: And then my friend,
• 29:00
  SPEAKER_02: and then there was another kid there who was,
• 29:02
  SPEAKER_02: well,
• 29:02
  SPEAKER_02: I saw the next day who had studied at Caltech in California.
• 29:06
  SPEAKER_02: You would know better than I,
• 29:07
  SPEAKER_02: and I'll try to get this right,
• 29:08
  but I guess Caltech is mostly NSF-funded is what I was told and
• 29:12
  is maybe like a
• 29:13
  SPEAKER_02: little more immune to these things.
• 29:16
  SPEAKER_02: But even you're saying,
• 29:17
  SPEAKER_02: oh,
• 29:17
  SPEAKER_02: Francis Arnold,
• 29:18
  SPEAKER_02: oh,
• 29:19
  SPEAKER_02: God,
• 29:19
  SPEAKER_02: I don't know if I'm going to get someone in trouble.
• 29:20
  SPEAKER_02: I think she's a Nobel Prize winner.
• 29:22
  SPEAKER_02: Her lab was under duress, and so I came away.
• 29:25
  SPEAKER_01: Multiple Nobel Prize winners we've talked to struggle for funding.
• 29:28
  SPEAKER_01: So I came away super depressed.
• 29:30
  Both in a preexisting system and especially as a result of
• 29:32
  what's happening.
• 29:33
  SPEAKER_00: Yeah.
• 29:34
  SPEAKER_00: People that I think,
• 29:36
  should easily be getting funding who are doing work that will
• 29:39
  change society are
• 29:40
  SPEAKER_00: currently struggling to maintain their grad students and postdocs.
• 29:44
  So I tweeted about this because I had this visceral reaction
• 29:46
  from these two
• 29:47
  SPEAKER_02: conversations right next to each other.
• 29:49
  SPEAKER_02: And then again, a bunch of friends of mine who are
• 29:54
  SPEAKER_02: extremely pro science were just in the comments saying,
• 29:57
  SPEAKER_02: this is what they get,
• 29:58
  SPEAKER_02: you know,
• 29:58
  SPEAKER_02: this is,
• 29:58
  SPEAKER_02: this is what needs to happen.
• 30:00
  SPEAKER_02: And, and yeah, I didn't know.
• 30:01
  SPEAKER_02: I mean,
• 30:02
  SPEAKER_02: that seemed to actually,
• 30:03
  SPEAKER_02: I didn't tally everything up,
• 30:04
  SPEAKER_02: but that seemed to be the majority opinion.
• 30:07
  SPEAKER_02: And then, yeah, I guess, um,
• 30:09
  So there's a proposal from Caleb Watney at the Institute for
• 30:13
  Progress that's called
• 30:14
  SPEAKER_01: X-Labs,
• 30:16
  SPEAKER_01: which has to do with diversifying,
• 30:18
  SPEAKER_01: setting up non-traditional institutional structures.
• 30:21
  So one type of X-Lab would look sort of like things like the
• 30:23
  ARC Institute or other
• 30:25
  SPEAKER_01: large institutes.
• 30:26
  SPEAKER_01: One would look maybe like an individual focused research organization,
• 30:29
  SPEAKER_01: let's say,
• 30:29
  SPEAKER_01: or something roughly like that.
• 30:30
  And one category would look like things like convergent
• 30:33
  research,
• 30:33
  which is a sort of
• 30:35
  SPEAKER_01: engine for creating and operating these focused research organizations.
• 30:40
  SPEAKER_01: But it might be for something other than an FRO.
• 30:41
  SPEAKER_01: It might be sort of they're kind of meta organizations.
• 30:45
  SPEAKER_01: I think that type of proposal is a really great idea.
• 30:47
  SPEAKER_01: Let's try some more experiments.
• 30:50
  SPEAKER_01: Let's try funding in different structures.
• 30:53
  But there's a lot of work going on in the existing structures
• 30:55
  where you have to be
• 30:56
  SPEAKER_01: very careful not to throw the baby out with the bathwater.
• 30:58
  SPEAKER_01: I think the people are struggling right now.
• 31:01
  SPEAKER_01: It's a problem.
• 31:02
  There's a line from Hamilton,
• 31:05
  which is,
• 31:05
  revolution is easy,
• 31:07
  but leading is harder.
• 31:07
  SPEAKER_00: You've got to sing it.
• 31:08
  SPEAKER_00: No.
• 31:09
  SPEAKER_00: It's because I would have to wrap it, actually, I think.
• 31:11
  SPEAKER_00: That would be bad for everybody.
• 31:14
  But the notion that it is one thing to be like,
• 31:17
  there is real bloat in science.
• 31:19
  SPEAKER_00: There are real institutional or bureaucratic problems.
• 31:23
  SPEAKER_00: The university systems have been around for hundreds of years.
• 31:26
  SPEAKER_00: It's...
• 31:27
  Michael Nielsen and Kendrin have an amazing article on
• 31:31
  metascience and sort of how
• 31:33
  SPEAKER_00: institutions change over time.
• 31:35
  SPEAKER_00: And there are issues that come from that.
• 31:39
  SPEAKER_00: It just means that now you have to build something better.
• 31:42
  SPEAKER_00: And I think that building is a big,
• 31:44
  SPEAKER_00: consciously making decisions about what to fund,
• 31:47
  SPEAKER_00: how to fund them,
• 31:48
  SPEAKER_00: how to keep the momentum going is like the moment that we're in.
• 31:51
  SPEAKER_00: And making sure that someone is doing that is important.
• 31:55
  do you guys think when i look at how you're funded somewhere
• 31:58
  like howard hughes you
• 31:59
  know um often it takes this incredibly wealthy person who
• 32:03
  happens to care about
• 32:04
  this particular thing and and then amazing things happen they
• 32:07
  you know how are used
• 32:08
  to going on for a long time doing pretty awesome stuff um i
• 32:11
  think you were a member
• 32:13
  you did the hertz fellowship right i mean hurt when i look at
• 32:16
  the list of people
• 32:17
  SPEAKER_02: who have done that
• 32:18
  SPEAKER_02: That's what?
• 32:19
  SPEAKER_02: That's like a rental car magnate from ages ago.
• 32:21
  SPEAKER_01: That's right.
• 32:22
  SPEAKER_01: It's relatively inexpensive.
• 32:24
  SPEAKER_01: Right.
• 32:24
  SPEAKER_01: Dario, who's the CEO of Anthropic, you know.
• 32:27
  SPEAKER_02: Andy Bechtelsheim, I think, was a Hertz fellow.
• 32:29
  SPEAKER_02: I mean, you know, he invented like half the stuff the internet runs on.
• 32:32
  It's a very interesting intervention,
• 32:35
  actually,
• 32:35
  the Hertz Fellowship,
• 32:36
  because it...
• 32:38
  it funds the student directly it doesn't fund the university pi
• 32:41
  that the student
• 32:42
  works for it just funds the student so that means that they can
• 32:44
  then move between
• 32:45
  labs and it also really encourages them to sort of up level
• 32:49
  their ambition through
• 32:50
  through the community of hertz fellows and so the hertz
• 32:52
  foundation is an example of
• 32:53
  just a purely additive intervention you're not taking away
• 32:56
  anyone's anything you're
• 32:57
  just giving these high potential people this extra kind of
• 33:00
  wiggle room within the
• 33:01
  system that's really powerful right and it's hard to know
• 33:04
  what's a selection effect
• 33:05
  SPEAKER_01: versus a treatment effect
• 33:06
  SPEAKER_01: It's really cool.
• 33:07
  SPEAKER_01: The results seem to be very interesting.
• 33:09
  SPEAKER_02: Nathan Mirvold,
• 33:10
  I think maybe it was what I urge people to go on the Wikipedia
• 33:12
  list and look at the
• 33:14
  SPEAKER_02: list of the Hertz fellows.
• 33:15
  SPEAKER_02: But what I was kind of thinking is one of my life theses,
• 33:21
  SPEAKER_02: I guess,
• 33:21
  SPEAKER_02: is that,
• 33:22
  SPEAKER_02: you know,
• 33:22
  SPEAKER_02: we live in the age of the engineer.
• 33:25
  SPEAKER_02: The world's wealthiest people are tech people.
• 33:27
  SPEAKER_02: Yeah.
• 33:30
  SPEAKER_02: This has been the story of the last, maybe call it 30, 30 years.
• 33:34
  SPEAKER_02: And, and clearly these tend to be people that care about science.
• 33:38
  They've,
• 33:38
  they've been in the tech industry and clearly they are funding
• 33:41
  some of these things.
• 33:41
  SPEAKER_02: However,
• 33:42
  SPEAKER_02: you know,
• 33:42
  SPEAKER_02: it's,
• 33:43
  SPEAKER_02: it's not,
• 33:43
  SPEAKER_02: I,
• 33:44
  SPEAKER_02: what I was curious about is like,
• 33:45
  SPEAKER_02: yeah,
• 33:45
  SPEAKER_02: do you guys feel like enough?
• 33:47
  SPEAKER_02: I see, I see everybody on the giving pledge.
• 33:49
  SPEAKER_02: They're going to give their money away when they die.
• 33:52
  SPEAKER_02: I see people griping about... This one's easy.
• 33:55
  SPEAKER_01: No, it's still not enough.
• 33:57
  SPEAKER_01: We're very grateful.
• 33:58
  SPEAKER_01: We have 30-something donors now.
• 34:02
  SPEAKER_01: And that has been the decisive factor in being able to do this.
• 34:08
  SPEAKER_01: It's interesting.
• 34:08
  SPEAKER_01: It's actually...
• 34:11
  SPEAKER_01: It's expensive, but not that expensive.
• 34:13
  SPEAKER_01: You mentioned the Hubble Space Telescope.
• 34:14
  SPEAKER_01: I don't know the exact cost,
• 34:15
  but my guess is on the order of 10 billion plus dollars to do
• 34:20
  something like the
• 34:20
  SPEAKER_01: Hubble or CERN.
• 34:21
  SPEAKER_01: You could probably do it cheaper.
• 34:23
  SPEAKER_01: Right, exactly.
• 34:24
  SPEAKER_01: You could probably do it cheaper.
• 34:26
  SPEAKER_01: If you imagine each FRO is 20 or 30 people times five years,
• 34:30
  SPEAKER_01: something like that,
• 34:30
  that's on the order of tens of millions of dollars if you add
• 34:32
  lab space and
• 34:34
  SPEAKER_01: equipment and so on.
• 34:38
  SPEAKER_01: So imagine there are, let's say, 20 or 30 fields of science.
• 34:42
  SPEAKER_01: Treat neuroscience as different from molecular biology,
• 34:47
  but don't treat Alzheimer's as different from schizophrenia
• 34:50
  within neuroscience.
• 34:50
  SPEAKER_01: There's like 20 buckets.
• 34:51
  SPEAKER_01: I don't know how many departments Caltech has, something on that order.
• 34:55
  SPEAKER_01: And imagine for each one of those,
• 34:56
  SPEAKER_01: there's like five to 10 pivotal fundamental missing capabilities.
• 35:01
  SPEAKER_01: As important as just mapping the wiring of the brain is for neuroscience,
• 35:04
  SPEAKER_01: like with connectomics,
• 35:06
  SPEAKER_01: right?
• 35:06
  SPEAKER_01: Or as big telescopes are for astronomy.
• 35:08
  SPEAKER_01: And so imagine there was sort of five froze times 20 fields.
• 35:12
  SPEAKER_01: That's 100 froze.
• 35:13
  SPEAKER_01: Each one of those, a few tens of millions of dollars.
• 35:16
  SPEAKER_01: Well,
• 35:16
  that's like single digit billion dollars to like in some sense
• 35:20
  do all of the focus
• 35:21
  SPEAKER_01: research organizations in a period of time.
• 35:25
  SPEAKER_01: So,
• 35:25
  SPEAKER_01: you know,
• 35:25
  SPEAKER_01: you need,
• 35:26
  SPEAKER_01: you know,
• 35:26
  SPEAKER_01: five to $10 billion,
• 35:27
  SPEAKER_01: something like that to kind of solve this categorically.
• 35:32
  SPEAKER_01: That's something that's within the reach of philanthropy,
• 35:34
  SPEAKER_01: but we're nowhere near that in terms of people actually committing.
• 35:37
  SPEAKER_01: Yeah.
• 35:38
  And I,
• 35:39
  I mean,
• 35:39
  if you look at the way the world is going to me,
• 35:41
  at least,
• 35:42
  you know,
• 35:45
  SPEAKER_02: People have gotten extraordinarily wealthy.
• 35:47
  SPEAKER_02: You see this swing.
• 35:49
  SPEAKER_02: Again, this is in my head.
• 35:50
  SPEAKER_02: I mean, please correct it if you think differently.
• 35:53
  But when I look at like today's data centers being built by the
• 35:58
  AI companies,
• 35:58
  SPEAKER_02: when I first started as a tech reporter,
• 36:00
  almost all the largest supercomputers were built by the
• 36:03
  Department of Energy or a
• 36:04
  SPEAKER_02: national lab or a university.
• 36:07
  SPEAKER_02: And now I don't think...
• 36:11
  whether it's the NSA or the DOE or whoever could even remotely
• 36:14
  kind of compete with
• 36:15
  SPEAKER_02: a Google or a Facebook or an open AI data center.
• 36:19
  I think about stuff like Tesla building like a nationwide
• 36:22
  charging network,
• 36:23
  SPEAKER_02: SpaceX,
• 36:24
  SPEAKER_02: outflanking anything the US military or government could do.
• 36:28
  SPEAKER_02: And like in my head then,
• 36:31
  SPEAKER_02: you know,
• 36:31
  so there's a bit of responsibility for all this wealth that
• 36:35
  you've gained because,
• 36:36
  SPEAKER_02: and then if you're going to sit there and complain,
• 36:39
  SPEAKER_02: well,
• 36:39
  SPEAKER_02: like China's,
• 36:40
  SPEAKER_02: racing ahead of us.
• 36:41
  SPEAKER_02: They're doing all these things.
• 36:43
  I feel like some of this burden has shifted from the government
• 36:45
  and the university
• 36:46
  SPEAKER_02: because of partly just how powerful tech has gone,
• 36:49
  SPEAKER_02: where the wealth has shifted.
• 36:51
  SPEAKER_02: And yeah, I don't know.
• 36:52
  SPEAKER_02: Somebody has to backfill this,
• 36:55
  SPEAKER_02: in my mind,
• 36:55
  SPEAKER_02: if it's not going to be centered at the universities anymore.
• 36:59
  Some of the big AI labs,
• 37:00
  of course,
• 37:01
  are ultimately nonprofits or public benefit corporations.
• 37:04
  SPEAKER_01: They could do a huge impact on science.
• 37:06
  SPEAKER_01: Even the nonprofits.
• 37:07
  SPEAKER_01: As could their
• 37:11
  SPEAKER_02: But that's interesting.
• 37:13
  SPEAKER_02: I mean, you guys must think about this all the time, though.
• 37:15
  SPEAKER_02: I mean,
• 37:15
  SPEAKER_02: if AI really does advance in the way these people are saying,
• 37:18
  SPEAKER_02: I mean,
• 37:19
  then there would be five companies that theoretically should be
• 37:22
  the epicenter of
• 37:23
  SPEAKER_02: scientific breakthrough and how that gets doled out,
• 37:26
  SPEAKER_02: I guess.
• 37:27
  SPEAKER_02: I mean, I guess people can tap into their systems, but...
• 37:29
  SPEAKER_00: I think that stopping to think through how does AI change our thesis?
• 37:36
  SPEAKER_00: So we started in 2020.
• 37:38
  SPEAKER_00: We're like, we're going to do two of these.
• 37:40
  SPEAKER_00: Let's go.
• 37:40
  SPEAKER_00: It's like, how do I hire anyone?
• 37:43
  SPEAKER_00: What does a contract look like?
• 37:45
  SPEAKER_00: It was very much so like...
• 37:47
  SPEAKER_00: Can we make this happen?
• 37:48
  SPEAKER_00: And then it was like, well, can we repeat this?
• 37:51
  SPEAKER_00: We have E11 and we have cultivarium.
• 37:54
  SPEAKER_00: Can we grow more?
• 37:55
  Now we're at 10 projects across literally math,
• 38:01
  biology,
• 38:03
  marine-based carbon removal.
• 38:06
  SPEAKER_00: We have this big...
• 38:08
  SPEAKER_00: big, big portfolio.
• 38:09
  SPEAKER_00: And we're like, great.
• 38:10
  What is the right thing to do given where we think we are on
• 38:13
  the timeline?
• 38:15
  And I think that it's really had us take a step back and think
• 38:17
  about the future.
• 38:18
  SPEAKER_00: What projects are going to be easily obsoleted by AI?
• 38:23
  If I do something now in like five years,
• 38:26
  would it have been done way faster and way cheaper?
• 38:29
  SPEAKER_00: What will work?
• 38:30
  SPEAKER_00: What will be like accelerated or will accelerate AI?
• 38:34
  SPEAKER_00: I think that there's this big
• 38:36
  SPEAKER_00: like ground truth problem.
• 38:37
  SPEAKER_00: I used to call it a data problem.
• 38:39
  SPEAKER_00: Maybe it's like a ground truth problem.
• 38:41
  SPEAKER_00: If you can get AI models to predict things,
• 38:43
  SPEAKER_00: how do you know they're...
• 38:44
  SPEAKER_00: Like,
• 38:45
  SPEAKER_00: how do you verify that they're predicting the right things?
• 38:49
  SPEAKER_00: How do you fine-tune them to the real world?
• 38:51
  SPEAKER_00: You actually need to go and collect data from those systems.
• 38:55
  SPEAKER_00: How do we...
• 38:56
  SPEAKER_00: How do we mitigate some of the downside risks of AI?
• 39:00
  SPEAKER_00: How do we build better biosecurity?
• 39:02
  SPEAKER_00: What are the technical problems on that side?
• 39:05
  How do we solve the fact that we're going to have potentially
• 39:08
  amazing hackers in
• 39:10
  SPEAKER_00: everyone's pockets?
• 39:11
  SPEAKER_00: What are the cyber security systems?
• 39:12
  SPEAKER_00: What are the technical problems there?
• 39:14
  And how do we build a better society given that we all have to
• 39:19
  now interact with a
• 39:21
  SPEAKER_00: crazy AI mediated world.
• 39:23
  How do we do epistemics or decision making or maybe even like
• 39:28
  some cultural things?
• 39:28
  SPEAKER_00: What will it do to the economy?
• 39:30
  SPEAKER_00: And so I think that it has like really updated our thinking of like,
• 39:33
  SPEAKER_00: well,
• 39:33
  SPEAKER_00: now that we have this framework of fundamental capabilities,
• 39:36
  now that we know that we can launch projects and we can do this
• 39:38
  road mapping,
• 39:39
  SPEAKER_00: how do we apply that to 2025?
• 39:42
  SPEAKER_02: And so you guys have placed these 10 bets.
• 39:44
  SPEAKER_02: Are you waiting to see how those play out?
• 39:48
  SPEAKER_02: Or are you making continued bets?
• 39:51
  SPEAKER_02: Especially with that context in mind.
• 39:54
  SPEAKER_01: Yeah, we're trying to make continued bets.
• 39:57
  SPEAKER_01: But with a certain...
• 40:00
  SPEAKER_01: increased clarity of focus given just,
• 40:02
  SPEAKER_01: in some sense,
• 40:03
  SPEAKER_01: the urgency of dealing with the changes due to AI.
• 40:07
  SPEAKER_01: But it even affects some of our existing projects.
• 40:09
  SPEAKER_01: Can I give you one example?
• 40:10
  SPEAKER_01: Yeah, I would love that.
• 40:11
  SPEAKER_01: I'm just going to ask you for it.
• 40:13
  SPEAKER_01: So we have one of our projects is called the Lean Theorem Prover,
• 40:16
  SPEAKER_01: so the Lean Focus Research Organization.
• 40:19
  SPEAKER_01: And this is weirdly, it is...
• 40:22
  It's kind of like the Hubble Space Telescope analogy for
• 40:24
  astronomy,
• 40:24
  but this is for math.
• 40:25
  SPEAKER_01: We didn't even know there was something like a scientific infrastructure,
• 40:29
  SPEAKER_01: engineering infrastructure for math,
• 40:30
  SPEAKER_01: right?
• 40:30
  SPEAKER_01: Because what do mathematicians do?
• 40:32
  They go on the whiteboard and they have good ideas and they
• 40:33
  talk to each other and
• 40:34
  SPEAKER_01: it's not very expensive or anything.
• 40:36
  SPEAKER_01: They just kind of do interesting work.
• 40:43
  SPEAKER_01: When mathematicians mostly write proofs,
• 40:45
  so a proof basically trying to deduce some mathematical
• 40:49
  conclusion from some
• 40:50
  SPEAKER_01: underlying assumptions and you write down all the steps,
• 40:54
  SPEAKER_01: they're typically doing this basically still on pencil and paper.
• 40:56
  And then that gets published in an article where you can see
• 41:00
  the full typeset math symbols.
• 41:02
  And then some other mathematician goes and they read the paper
• 41:04
  and they say,
• 41:05
  SPEAKER_01: okay,
• 41:05
  SPEAKER_01: do I think this is right or wrong?
• 41:07
  SPEAKER_01: Have you missed a step here?
• 41:09
  SPEAKER_01: Well,
• 41:09
  SPEAKER_01: it turns out that with proofs,
• 41:11
  SPEAKER_01: with mathematical proofs,
• 41:12
  SPEAKER_01: if you write it in a certain programming language,
• 41:14
  SPEAKER_01: this is the lean programming language that this is based on,
• 41:18
  you can actually just click a button on the computer and the
• 41:20
  computer says,
• 41:20
  SPEAKER_01: is the proof correct?
• 41:21
  SPEAKER_01: You can just check,
• 41:23
  do all the steps add up from the assumptions to the conclusions
• 41:26
  with no missing
• 41:26
  SPEAKER_01: pieces?
• 41:27
  SPEAKER_01: Is it a complete proof?
• 41:28
  SPEAKER_02: This is what the FRO you're funding, they've built that application.
• 41:32
  SPEAKER_01: So we're building the software infrastructure of Lean.
• 41:34
  SPEAKER_01: They had already started doing this,
• 41:36
  but they only had a couple people working on it as an open
• 41:38
  source project,
• 41:39
  and they now have 20 so people in an organized engineering
• 41:42
  project building out
• 41:43
  SPEAKER_01: this Lean theorem prover.
• 41:44
  SPEAKER_02: And did Lean exist before?
• 41:46
  SPEAKER_01: It did exist before, and we made a FRO as a way to accelerate.
• 41:51
  SPEAKER_01: their progress.
• 41:53
  And so now they're organized as a fro versus a kind of ad hoc
• 41:56
  open source set of people.
• 41:59
  And anyway,
• 42:00
  so you can now verify these math proofs if you write them in
• 42:03
  Lean.
• 42:04
  SPEAKER_01: This has a couple of different types of implications.
• 42:06
  So imagine I'm one mathematician and I want to know if I trust
• 42:08
  the other mathematician's proof.
• 42:09
  SPEAKER_01: Just say, did the Lean compile, right?
• 42:12
  SPEAKER_01: Check.
• 42:12
  SPEAKER_01: Okay, then I know the proof is correct.
• 42:15
  SPEAKER_01: So that's interesting for mathematical collaboration.
• 42:17
  SPEAKER_01: Is it like 100%?
• 42:20
  SPEAKER_01: If the proof is correct, it will tell you that it is correct.
• 42:23
  And if the proof is missing something or is wrong,
• 42:25
  it will tell you it is wrong.
• 42:26
  SPEAKER_01: It doesn't necessarily tell you what you have to do to make it correct.
• 42:30
  SPEAKER_01: That's where some AI would come in, is can AI suggest the next step?
• 42:33
  SPEAKER_01: But okay, now you have the verifier, right?
• 42:35
  SPEAKER_01: So this is a little bit like in playing the game of Go, right?
• 42:39
  SPEAKER_01: Where DeepMind previously had big breakthroughs.
• 42:44
  SPEAKER_01: Because you can tell in the computer quickly who won the game of Go,
• 42:48
  SPEAKER_01: you can now crank out reinforcement learning on this,
• 42:52
  SPEAKER_01: train AI based on this verified ground truth reward of who won.
• 42:57
  Now here you can train AI on the verified ground truth reward
• 43:00
  of did you correctly
• 43:01
  SPEAKER_01: prove the math theorem?
• 43:04
  SPEAKER_01: So that's interesting in a few ways.
• 43:05
  SPEAKER_01: So I mean, one is, can just the AI start to do math?
• 43:09
  SPEAKER_01: So far,
• 43:10
  SPEAKER_01: everything we've been able to do,
• 43:12
  SPEAKER_01: reinforcement learning from a ground truth verified reward,
• 43:15
  SPEAKER_01: eventually,
• 43:16
  SPEAKER_01: you get basically superhuman.
• 43:18
  If you have a good in silico,
• 43:19
  in the computer,
• 43:20
  you can compute the verified reward quickly.
• 43:22
  SPEAKER_01: So now does that mean math?
• 43:24
  SPEAKER_01: proving math theorems,
• 43:25
  SPEAKER_01: including ones no human has ever been able to prove,
• 43:29
  SPEAKER_01: conjectures no one has ever been able to prove,
• 43:30
  SPEAKER_01: would fall to this.
• 43:32
  SPEAKER_01: So that's all interesting for math.
• 43:33
  But then now with AI,
• 43:35
  we're thinking about this in a different way,
• 43:37
  a couple of different ways.
• 43:38
  So one is that's a verifiable source of true reasoning,
• 43:42
  complex reasoning.
• 43:45
  SPEAKER_01: Lean will tell you if the thing hallucinated a math proof, right?
• 43:48
  SPEAKER_01: So it can reward it for not hallucinating.
• 43:49
  SPEAKER_01: That's part of it.
• 43:51
  SPEAKER_01: But also,
• 43:51
  SPEAKER_01: if you think about software,
• 43:53
  SPEAKER_01: software is just basically some complicated mathematical object.
• 43:56
  SPEAKER_01: A software program is just basically math.
• 43:59
  SPEAKER_01: And so...
• 44:01
  SPEAKER_01: What if you wanted to prove that the software couldn't be hacked?
• 44:04
  Even if you had a super intelligent AI that was trying to hack
• 44:07
  your software,
• 44:08
  I have a mathematical proof now that the software cannot be
• 44:12
  hacked no matter how
• 44:13
  SPEAKER_01: smart you are.
• 44:14
  SPEAKER_01: It just can't be hacked in certain ways.
• 44:16
  SPEAKER_01: It's mathematically impossible.
• 44:18
  Superintelligence can't get around a mathematical impossibility
• 44:21
  no matter how
• 44:22
  SPEAKER_01: intelligent it is.
• 44:22
  SPEAKER_01: So that's this idea of formally verifying software.
• 44:25
  SPEAKER_01: So now we're thinking,
• 44:25
  SPEAKER_01: okay,
• 44:26
  SPEAKER_01: Lean,
• 44:26
  which we thought was just a pure math project in some sense
• 44:29
  actually might be
• 44:29
  SPEAKER_01: really important.
• 44:30
  SPEAKER_01: What if we could formally verify and have defense-dominant cybersecurity?
• 44:34
  SPEAKER_01: Things could not be hacked.
• 44:37
  SPEAKER_01: in the limit,
• 44:37
  SPEAKER_01: because if you can have AI to prove that the software can be hacked,
• 44:40
  SPEAKER_01: or synthesize the software and prove that it can't be hacked,
• 44:44
  SPEAKER_01: software becomes defense dominant now.
• 44:47
  SPEAKER_01: So now we're thinking a lot about,
• 44:49
  SPEAKER_01: should we go into formal methods,
• 44:50
  SPEAKER_01: formal verification of software,
• 44:52
  SPEAKER_01: rather than just thinking about this as math?
• 44:54
  SPEAKER_01: And that's because we're worried,
• 44:56
  SPEAKER_01: obviously,
• 44:57
  SPEAKER_01: that when AI gets advanced,
• 44:59
  SPEAKER_01: people are going to be hacking right and left,
• 45:01
  SPEAKER_01: and our societal infrastructure is going to be very vulnerable,
• 45:03
  SPEAKER_01: too.
• 45:04
  SPEAKER_02: Yeah, sure.
• 45:05
  SPEAKER_02: And were they one of your first projects you funded?
• 45:08
  SPEAKER_01: They're kind of in the middle of our,
• 45:09
  SPEAKER_01: somewhere in the middle of the nine or ten that we have now.
• 45:12
  SPEAKER_01: Okay.
• 45:13
  SPEAKER_00: It's been amazing to watch their user base grow.
• 45:17
  SPEAKER_00: DeepMind has used them in their research.
• 45:19
  SPEAKER_00: Terry Tao is a Fields Medalist,
• 45:21
  SPEAKER_00: probably one of the leading mathematicians right now,
• 45:24
  SPEAKER_00: and has disproved one of his own proofs using Lean,
• 45:27
  SPEAKER_00: and is like actively using Lean.
• 45:29
  SPEAKER_00: Harsh.
• 45:30
  SPEAKER_00: And he's like live streaming using Lean.
• 45:32
  SPEAKER_00: I didn't know it was the thing you could do.
• 45:36
  SPEAKER_01: Getting data and what Terry Tao actually does with it is actually a very
• 45:39
  SPEAKER_01: interesting data source.
• 45:40
  SPEAKER_02: Like a proof influencer or something?
• 45:45
  SPEAKER_02: We're going to move off froze in a second.
• 45:48
  And don't do E11 or Forrest because I'm going to take you guys
• 45:52
  into neuroscience land.
• 45:53
  SPEAKER_02: But let's just make this concrete for people one more time.
• 45:56
  SPEAKER_02: Give me one other question.
• 45:58
  and why that the way you're solving this problem fits with the
• 46:05
  hardcore philosophy
• 46:06
  SPEAKER_02: of the organization.
• 46:07
  SPEAKER_00: Yeah.
• 46:09
  SPEAKER_00: Let's talk about the climate.
• 46:12
  SPEAKER_00: So assuming
• 46:14
  SPEAKER_00: climate change is real,
• 46:16
  SPEAKER_00: lots of efforts to reduce emissions,
• 46:19
  SPEAKER_00: we probably also need to draw carbon out of the atmosphere.
• 46:23
  SPEAKER_00: If you do the math,
• 46:24
  a lot of the ways that we can draw carbon out of the atmosphere
• 46:27
  are actually really
• 46:28
  SPEAKER_00: energy intensive.
• 46:29
  SPEAKER_00: One of the least intensive ways is to use an open system,
• 46:32
  SPEAKER_00: so something where carbon already fluxes in and out of the atmosphere.
• 46:36
  The biggest carbon sink is the ocean,
• 46:40
  and tons of carbon regularly move back and forth.
• 46:44
  If you can increase the alkalinity of the ocean,
• 46:46
  it then actually can draw down more carbon.
• 46:49
  SPEAKER_00: It's a big buffer, basically.
• 46:51
  We're already pulling a ton of carbon into the ocean as more
• 46:55
  CO2 goes into the
• 46:56
  SPEAKER_00: atmosphere,
• 46:56
  SPEAKER_00: which is like why ocean acidification is coming.
• 46:58
  SPEAKER_00: So basically, you can take a big...
• 47:01
  SPEAKER_00: a boat full of alkaline material,
• 47:03
  SPEAKER_00: a boat full of tums,
• 47:05
  SPEAKER_00: toss it into the ocean,
• 47:07
  SPEAKER_00: and we should have carbon removed.
• 47:10
  SPEAKER_00: And it should be removed in this way that is like abiologic.
• 47:12
  SPEAKER_00: It shouldn't really affect biological systems.
• 47:18
  But I have no idea how much carbon is removed because it will
• 47:22
  happen over the next decade.
• 47:25
  SPEAKER_00: That ton of Tums may float along and then sink to the bottom.
• 47:31
  SPEAKER_00: It has to go out through the estuary.
• 47:32
  There's all sorts of physical processes going on and then
• 47:35
  chemical processes going on.
• 47:37
  SPEAKER_00: No idea how much carbon is actually removed.
• 47:42
  SPEAKER_00: You just can't measure it.
• 47:43
  SPEAKER_00: And so it has held back.
• 47:45
  SPEAKER_00: the industry,
• 47:45
  SPEAKER_00: because if you're a buyer,
• 47:47
  SPEAKER_00: you want to know how much carbon removal you have purchased.
• 47:52
  SPEAKER_00: And these companies who are spinning up in the space can't tell you.
• 47:57
  SPEAKER_00: And so we have a FRO called Seaworthy,
• 48:00
  which is building a set of physics-based modeling tools to tell
• 48:04
  you that if you did
• 48:05
  a perturbation intervention and you put in an alkaline material
• 48:10
  in this estuary
• 48:12
  SPEAKER_00: during this season,
• 48:14
  SPEAKER_00: where that carbon flux will happen over the next decade.
• 48:19
  SPEAKER_00: And it's an incredibly complex physics problem.
• 48:23
  SPEAKER_00: There is the very surface level chemistry problem,
• 48:26
  SPEAKER_00: there is the local estuary flows,
• 48:29
  SPEAKER_00: the regional flows,
• 48:29
  SPEAKER_00: and then actually into global flows,
• 48:32
  SPEAKER_00: and building models at all of these levels.
• 48:34
  SPEAKER_00: They exist on each level,
• 48:35
  but acting between the different levels of granularity is
• 48:39
  incredibly mathematically
• 48:40
  SPEAKER_00: complicated.
• 48:42
  SPEAKER_00: There was a bunch of orgs,
• 48:44
  SPEAKER_00: NCAR,
• 48:44
  SPEAKER_00: NOAA,
• 48:45
  SPEAKER_00: who can do pieces of this,
• 48:48
  SPEAKER_00: but the problem of solving,
• 48:50
  SPEAKER_00: how do I solve?
• 48:52
  measurement and verification for marine-based carbon removal is
• 48:56
  just,
• 48:56
  like,
• 48:56
  not their agenda.
• 48:57
  SPEAKER_00: So no one is going to build this giant modeling software.
• 49:01
  SPEAKER_00: And so there is a team in Boulder,
• 49:03
  SPEAKER_00: Colorado,
• 49:04
  SPEAKER_00: led by Matt and Alicia,
• 49:05
  SPEAKER_00: who are just killing it,
• 49:07
  SPEAKER_00: building only this infrastructure.
• 49:09
  SPEAKER_00: So over the last couple of years,
• 49:11
  SPEAKER_00: they have released a map of all of the different kind of carbon fluxes.
• 49:16
  You can check out their website,
• 49:17
  and they're working with partners to develop this tooling.
• 49:21
  SPEAKER_02: Okay.
• 49:21
  SPEAKER_02: So that's great because I think some people might,
• 49:24
  SPEAKER_02: there can be startups that come out of this research,
• 49:27
  SPEAKER_02: but there don't always have to be.
• 49:28
  SPEAKER_02: And I mean,
• 49:28
  SPEAKER_02: this is an example where like there may be a business case,
• 49:30
  SPEAKER_02: there may not,
• 49:31
  but like we want this to happen and you guys are willing to
• 49:34
  help make it happen.
• 49:36
  SPEAKER_02: And sorry,
• 49:37
  SPEAKER_02: I will push on passes,
• 49:38
  SPEAKER_02: but just it occurs to me,
• 49:40
  SPEAKER_02: we've talked about this and people who might not.
• 49:43
  SPEAKER_02: they might want to know what the business case is around some of this.
• 49:45
  SPEAKER_02: I mean,
• 49:46
  SPEAKER_02: uh,
• 49:47
  SPEAKER_02: the people who work on these technologies,
• 49:49
  SPEAKER_02: if it's,
• 49:50
  SPEAKER_02: um,
• 49:51
  SPEAKER_02: well,
• 49:52
  SPEAKER_02: any of these throws,
• 49:52
  SPEAKER_02: I mean,
• 49:53
  SPEAKER_02: they are able to,
• 49:55
  SPEAKER_02: after the fact,
• 49:56
  SPEAKER_02: like if they see a company in,
• 49:58
  SPEAKER_02: uh,
• 49:59
  And they would have,
• 50:00
  I guess,
• 50:00
  like a head start over anybody else in this field.
• 50:05
  SPEAKER_01: If it makes sense for the impact,
• 50:08
  SPEAKER_01: for them to go off and start a company,
• 50:09
  SPEAKER_01: if that's the best way to take it forward,
• 50:12
  SPEAKER_01: generally speaking,
• 50:14
  SPEAKER_01: we can support that.
• 50:15
  SPEAKER_01: So in general,
• 50:16
  SPEAKER_01: we are looking for,
• 50:18
  SPEAKER_01: we think of it as sort of at scale or scaled revolutionary outcomes,
• 50:23
  SPEAKER_01: so scaled revolutions.
• 50:26
  SPEAKER_01: that could take the form of commercial breakout, right?
• 50:30
  SPEAKER_01: And sometimes it needs to in order to actually have that impact.
• 50:33
  SPEAKER_01: I mean,
• 50:34
  SPEAKER_01: the DNA sequencers that everyone uses that made by Illumina,
• 50:36
  SPEAKER_01: which is a multi-billion dollar company,
• 50:38
  or you can think about OpenAI or other things in the AI space
• 50:42
  needing this compute
• 50:43
  SPEAKER_01: scale.
• 50:44
  SPEAKER_01: And so sometimes it makes a ton of sense.
• 50:46
  SPEAKER_01: And the fro would basically get it to the point where,
• 50:51
  SPEAKER_01: whereas initially you would have to just raise philanthropically.
• 50:54
  SPEAKER_01: But if it works,
• 50:57
  you now have such a powerful technology and you have such an
• 50:59
  ecosystem that you've
• 50:59
  SPEAKER_01: built around it,
• 51:00
  SPEAKER_01: partly by being very generous and very open.
• 51:04
  SPEAKER_01: that now the world is going to interact with you differently.
• 51:08
  SPEAKER_01: You're in a different world, essentially.
• 51:09
  SPEAKER_01: You've shifted the difficulty of now doing companies in that space.
• 51:13
  SPEAKER_01: But maybe it's not through a company.
• 51:15
  Maybe you've shifted the difficulty of partnering with the
• 51:18
  government or partnering
• 51:19
  with another large institute or creating a long-lived nonprofit
• 51:22
  in that space.
• 51:25
  SPEAKER_01: Sometimes we think about it.
• 51:27
  SPEAKER_01: Let me give you just,
• 51:29
  SPEAKER_01: I know we'll come back to neuroscience,
• 51:30
  SPEAKER_01: but if we could get brain mapping 100 times cheaper,
• 51:34
  SPEAKER_01: that's going to make it a lot easier for philanthropic actors,
• 51:37
  SPEAKER_01: government actors,
• 51:37
  SPEAKER_01: and commercial actors to take connectomics brain mapping as a forward,
• 51:45
  SPEAKER_01: right?
• 51:45
  SPEAKER_01: Isn't that not a multi-billion dollar endeavor to map a mouse brain?
• 51:49
  SPEAKER_01: Let's say it's now a millions of dollars question.
• 51:52
  SPEAKER_01: Now you've changed.
• 51:54
  that landscape the donors didn't you know profit from that but
• 51:58
  they enabled that to
• 52:00
  happen and somebody else can take it yeah so what are the
• 52:03
  trade-offs i mean so if
• 52:05
  if um your donors do does convergent own ultimately the
• 52:09
  intellectual property that
• 52:10
  comes out of this and then also you know for the people working
• 52:13
  on this technology
• 52:14
  like you said i mean i see all the pro cases i could see how in
• 52:18
  their head um
• 52:19
  a lot of what you guys do is open sourced you're sharing your
• 52:22
  data um that's sort
• 52:24
  of the point of all this so you could make an argument that yes
• 52:27
  you gotta have a
• 52:27
  head start and and maybe that that's and that is great but
• 52:30
  you're also kind of
• 52:31
  handing opportunities to other people to um compete with you so
• 52:37
  yeah i mean just
• 52:38
  SPEAKER_01: like a university
• 52:39
  SPEAKER_01: Yeah.
• 52:39
  SPEAKER_01: Essentially,
• 52:40
  SPEAKER_01: the FROs,
• 52:41
  SPEAKER_01: which are part of our,
• 52:42
  SPEAKER_01: in the US projects are part of our US nonprofit umbrella.
• 52:45
  We're doing some projects in the UK and that have a slightly
• 52:48
  different type of
• 52:49
  SPEAKER_01: structure where they're a bit more independent of the parent.
• 52:51
  SPEAKER_01: But they roll up into our nonprofit structure.
• 52:54
  SPEAKER_01: And so just like a university can license IP out to a spin out company,
• 52:57
  SPEAKER_01: I mean,
• 52:57
  SPEAKER_01: Stanford owns some IP on what became Google,
• 53:00
  SPEAKER_01: for example.
• 53:02
  SPEAKER_01: We can license out to a new company, for example.
• 53:06
  SPEAKER_01: And yeah, what you're saying, I mean, I think is quite right.
• 53:08
  SPEAKER_01: The team,
• 53:09
  And the focus and the organization and the know-how of that is
• 53:12
  tremendous to
• 53:15
  SPEAKER_01: actually go and do something versus the abstract IP as well.
• 53:18
  But in any case,
• 53:18
  we can license the IP at fair market value as a nonprofit to...
• 53:24
  SPEAKER_01: two, two spin outs.
• 53:25
  SPEAKER_02: Okay.
• 53:26
  I feel like it'd be doing a disservice if we didn't talk brains
• 53:28
  because you're both
• 53:29
  SPEAKER_02: such big thinkers on this area and have lived through it.
• 53:31
  SPEAKER_02: Um,
• 53:33
  SPEAKER_02: again,
• 53:33
  SPEAKER_02: not sure if this is an uncomfortable question or not,
• 53:36
  SPEAKER_02: but,
• 53:36
  SPEAKER_02: um,
• 53:36
  SPEAKER_02: as you,
• 53:37
  SPEAKER_02: you both know,
• 53:37
  SPEAKER_02: I know Elon and Brian Johnson pretty well.
• 53:40
  SPEAKER_02: And so,
• 53:40
  SPEAKER_02: you know,
• 53:41
  SPEAKER_02: they were kind of doing this kernel,
• 53:42
  SPEAKER_02: um,
• 53:43
  SPEAKER_02: people probably don't know kernel as well,
• 53:45
  SPEAKER_02: but it was trying to make these helmets that,
• 53:47
  SPEAKER_02: that,
• 53:48
  SPEAKER_02: Well,
• 53:48
  it started out wanting to go kind of full Neuralink with a more
• 53:51
  invasive kind of
• 53:52
  thing and then ended up in this place trying to read neuronal
• 53:56
  and blood activity
• 53:57
  SPEAKER_02: from outside the skull without doing an implant.
• 54:00
  SPEAKER_02: But, you know, I've been with both those guys.
• 54:03
  SPEAKER_02: I mean...
• 54:04
  SPEAKER_02: I felt like you might have,
• 54:05
  SPEAKER_02: just what I was reading last night,
• 54:06
  I felt like you must have been in the middle of this moment
• 54:08
  that I've talked with
• 54:09
  SPEAKER_02: both of them about,
• 54:10
  which is I kind of think Brian was there first and then was
• 54:14
  trying to spin Elon up
• 54:15
  SPEAKER_02: on this idea.
• 54:16
  SPEAKER_01: The brain-computer interface.
• 54:17
  SPEAKER_01: Of doing some brain-computer interface stuff.
• 54:19
  SPEAKER_01: Mini-boom of 2016.
• 54:20
  SPEAKER_01: Right, right.
• 54:20
  SPEAKER_01: I was there, yeah.
• 54:22
  SPEAKER_01: Yeah,
• 54:22
  SPEAKER_02: and then as I've seen happen with Elon and some other things,
• 54:26
  SPEAKER_02: he kind of decided to do it himself.
• 54:27
  But so,
• 54:28
  yeah,
• 54:28
  were you,
• 54:29
  I forget what I read,
• 54:30
  but it sounded like you were,
• 54:31
  I was involved in some early discussions with the team that
• 54:34
  became the Neuralink founding team.
• 54:36
  SPEAKER_01: I didn't end up joining that team,
• 54:37
  but there was a very interesting early discussions prior to it
• 54:40
  existing as a
• 54:41
  SPEAKER_01: company.
• 54:42
  And then I very shortly after that joined Brian as chief
• 54:46
  strategy officer of Kernel
• 54:47
  SPEAKER_01: and helped them start the work,
• 54:49
  SPEAKER_01: particularly the work that went on the non-invasive direction.
• 54:51
  Do you want to settle the debate as far as there is one as to
• 54:54
  who got there first?
• 54:57
  I honestly think there was two things going on and they both
• 55:02
  were just discussing
• 55:03
  SPEAKER_01: with each other and ended up not working together.
• 55:05
  SPEAKER_01: They ended up taking extremely different directions.
• 55:08
  SPEAKER_01: I think it's great actually that they took two different directions.
• 55:11
  SPEAKER_02: And so why then?
• 55:12
  SPEAKER_02: And we've got this Utah array that's around for decades.
• 55:16
  SPEAKER_02: It's a device that is invasive,
• 55:19
  SPEAKER_02: that is accomplishing amazing science,
• 55:22
  SPEAKER_02: but it's trapped in universities.
• 55:24
  SPEAKER_02: And then,
• 55:25
  SPEAKER_02: as you said,
• 55:26
  SPEAKER_02: like 2016,
• 55:27
  SPEAKER_02: all of a sudden,
• 55:29
  SPEAKER_02: people decide,
• 55:30
  SPEAKER_02: okay,
• 55:31
  SPEAKER_02: we're going to move this into startup land,
• 55:33
  SPEAKER_02: and then that has only dramatically changed.
• 55:36
  SPEAKER_01: It was very inspirational,
• 55:38
  SPEAKER_01: certainly,
• 55:38
  SPEAKER_01: for the idea of the FRO,
• 55:39
  SPEAKER_01: is to see what can you do in an academic lab.
• 55:42
  SPEAKER_01: The brain-computer interface problem,
• 55:44
  SPEAKER_01: right,
• 55:45
  it requires things like chip design and microfabrication and
• 55:49
  medical device,
• 55:50
  SPEAKER_01: you know,
• 55:50
  SPEAKER_01: hermetic packaging and FDA regulation.
• 55:54
  SPEAKER_01: if not really a neuroscience problem,
• 55:56
  but it was mostly left upon the neuroscientists to mostly make
• 55:59
  their own tools,
• 56:00
  SPEAKER_01: right?
• 56:01
  And so being able to shift it into this cross-disciplinary deep
• 56:04
  tech startup type
• 56:04
  SPEAKER_01: structure,
• 56:05
  SPEAKER_01: it felt very enabling.
• 56:08
  SPEAKER_01: That doesn't make it not an incredibly hard problem to do that,
• 56:12
  SPEAKER_01: but it was a major shift,
• 56:14
  SPEAKER_01: and I actually think it was a positive shift.
• 56:16
  SPEAKER_02: Is that the shift that took place, though?
• 56:17
  It just happened to be that two very wealthy individuals
• 56:20
  happened to get...
• 56:22
  SPEAKER_02: interested in the brain at that particular moment.
• 56:25
  SPEAKER_01: I can't speak for them,
• 56:26
  SPEAKER_01: obviously,
• 56:26
  SPEAKER_01: but I think part of what was happening,
• 56:28
  SPEAKER_01: what year was OpenAI founded?
• 56:30
  SPEAKER_02: OpenAI, it was around 2015, 2016.
• 56:31
  SPEAKER_01: Yeah, what year was AlphaGo or people were seeing AI, therefore...
• 56:38
  SPEAKER_01: we have a renewed interest in the brain.
• 56:40
  SPEAKER_01: And I think we actually share that.
• 56:42
  We think our neuroscience projects are even more important in
• 56:44
  proportion to how
• 56:45
  SPEAKER_01: fast AI goes actually.
• 56:47
  SPEAKER_02: Okay.
• 56:47
  SPEAKER_02: This is where I want to go too.
• 56:48
  SPEAKER_02: So, which is interesting.
• 56:50
  SPEAKER_02: And we've talked about this before,
• 56:51
  SPEAKER_02: I think,
• 56:52
  SPEAKER_02: and I don't,
• 56:53
  SPEAKER_02: I'm usually not very eloquent in how I explain this.
• 56:56
  SPEAKER_02: But there is this...
• 56:58
  SPEAKER_02: There's a fascinating thing where we are,
• 57:00
  SPEAKER_02: we,
• 57:00
  SPEAKER_02: okay,
• 57:00
  SPEAKER_02: yes,
• 57:00
  SPEAKER_02: we have brain computer interface startups now.
• 57:03
  SPEAKER_02: So there is money going there.
• 57:04
  SPEAKER_02: It's like,
• 57:05
  SPEAKER_02: and it's large,
• 57:06
  but it dwarfs what people are spending on artificial
• 57:10
  intelligence and creating
• 57:12
  SPEAKER_02: brains in the,
• 57:13
  SPEAKER_02: in the,
• 57:14
  SPEAKER_02: in,
• 57:16
  SPEAKER_02: We've talked about the connectome in the past.
• 57:20
  SPEAKER_02: So that's an effort to map the whole inner workings of the brain.
• 57:24
  I would argue relatively little money overall is kind of spent
• 57:27
  on what seems like
• 57:29
  the thing that defines our entire existence that we should
• 57:32
  probably know about.
• 57:33
  SPEAKER_02: I can see there's an obvious answer to my question, which is, well...
• 57:38
  AI has the potential to make money and knowing how our brain
• 57:41
  works like may or may not.
• 57:42
  And,
• 57:43
  and,
• 57:43
  um,
• 57:43
  so I get it,
• 57:44
  I get why there's this divergence of resources.
• 57:46
  SPEAKER_02: I also find this like an interesting philosophical question,
• 57:49
  SPEAKER_02: especially in this moment,
• 57:51
  SPEAKER_02: you would think that,
• 57:53
  SPEAKER_02: um,
• 57:54
  SPEAKER_02: rich people and rich countries would devote just,
• 58:00
  SPEAKER_02: I would think so many resources to understanding the most important organ
• 58:06
  imagine the thing that defines everything yeah but we're
• 58:10
  clearly spending way more
• 58:11
  inventing the artificial version of that it's also it's also
• 58:14
  hard to find the right
• 58:16
  focality of those efforts so there's obviously a lot of money
• 58:20
  spent for example on
• 58:21
  SPEAKER_01: alzheimer's whether that be the research or or
• 58:24
  SPEAKER_01: failed often therapeutics, right?
• 58:28
  SPEAKER_01: That's one piece of what's going on with the brain.
• 58:32
  SPEAKER_01: What we're trying to do is say, okay, let's step up on the tech tree.
• 58:35
  SPEAKER_01: The problem is connectomics is too brittle and too expensive.
• 58:38
  SPEAKER_01: If we can first do a fro that makes it less brittle and less expensive,
• 58:42
  SPEAKER_01: can that unlock what really should be?
• 58:45
  SPEAKER_01: Not even that expensive, but it should be.
• 58:48
  SPEAKER_01: We need to map entire mammal brains, right?
• 58:51
  SPEAKER_01: It's crazy that we haven't mapped entire mammal brains.
• 58:52
  SPEAKER_02: Yeah, we've done a worm and a fly, right?
• 58:54
  SPEAKER_02: Yeah.
• 58:55
  SPEAKER_00: I think that there's a thing...
• 58:57
  SPEAKER_00: that physicists have it easy in that you can,
• 59:00
  SPEAKER_00: from first principles,
• 59:02
  SPEAKER_00: basically all agree on the right experiment to run.
• 59:07
  SPEAKER_00: We have to build a collider.
• 59:09
  SPEAKER_00: Everyone knows we have to figure out how these particles interact.
• 59:12
  SPEAKER_01: At least in a few key cases, yeah.
• 59:14
  SPEAKER_01: The accelerator has to get bigger, so you can't see smaller particles.
• 59:17
  SPEAKER_00: You've got to make a bigger one.
• 59:19
  I think that in things like biology and neuroscience,
• 59:22
  that's much tougher.
• 59:24
  How do you get everyone to agree on what is the right rig to
• 59:28
  buy or to build that
• 59:30
  SPEAKER_00: enables the field?
• 59:31
  Because it's much harder from first principles just to do math
• 59:34
  that leads you there.
• 59:35
  Should we be studying how the mouse brain does spatial
• 59:38
  cognition or should we be
• 59:39
  SPEAKER_01: studying how the monkey brain neurodegenerates?
• 59:43
  SPEAKER_01: There's all these different possibilities.
• 59:45
  Now,
• 59:45
  I think if you have a cheap enough technology,
• 59:48
  you don't have to choose.
• 59:50
  SPEAKER_01: You can do 100 mouse brains that have different diseases and normal.
• 59:53
  SPEAKER_01: And so this is why we think neuroscience is so technology driven.
• 59:59
  SPEAKER_01: because it is hard actually to agree in a scientific megaproject.
• 1:00:04
  SPEAKER_02: Right.
• 1:00:04
  SPEAKER_01: Okay.
• 1:00:05
  SPEAKER_02: Okay.
• 1:00:06
  SPEAKER_02: I'm like excited in this moment because it feels like we are,
• 1:00:11
  it seems impossible that over the next 10 years we do not
• 1:00:16
  gather more data about
• 1:00:17
  SPEAKER_02: the brain on like some...
• 1:00:19
  SPEAKER_02: just insanely dramatic level than we've ever had before.
• 1:00:23
  SPEAKER_02: I mean,
• 1:00:23
  SPEAKER_02: we've got,
• 1:00:24
  SPEAKER_02: from all these startups,
• 1:00:25
  we're just going to have probes sitting on brains in ways that
• 1:00:28
  we never could
• 1:00:28
  SPEAKER_02: before.
• 1:00:29
  SPEAKER_02: So that seems to happen.
• 1:00:31
  SPEAKER_02: I am like a slightly, I've spent a lot of time with these BCI startups.
• 1:00:35
  SPEAKER_02: It's like my favorite thing to cover and write about.
• 1:00:38
  SPEAKER_02: I'm concerned though, you know, because they, Neuralink,
• 1:00:43
  the elon of it all neuralink has done great things they've
• 1:00:45
  brought all this money
• 1:00:46
  and attention into this field but we just talked about that
• 1:00:49
  these companies
• 1:00:50
  probably wouldn't exist um at least like 20 of them if it
• 1:00:53
  wasn't for that they just
• 1:00:54
  got this massive new valuation when i spend time with these
• 1:00:58
  companies it's so clear
• 1:00:59
  to me that in the near term this is all about helping people in
• 1:01:02
  quite dire
• 1:01:03
  SPEAKER_02: circumstances um
• 1:01:05
  SPEAKER_02: To me is a limited market.
• 1:01:07
  SPEAKER_02: You're talking about people who are paralyzed, have ALS.
• 1:01:10
  SPEAKER_02: Maybe if we get to like mental disorders, maybe it gets big enough.
• 1:01:15
  SPEAKER_02: I just, I worry that there's this gap between what I see.
• 1:01:19
  These companies are being valued at what are almost like
• 1:01:22
  consumer electronics.
• 1:01:25
  Which is the vision for Neuralink,
• 1:01:26
  and it's vaguely the vision for some of the others.
• 1:01:29
  SPEAKER_02: But,
• 1:01:29
  SPEAKER_02: man,
• 1:01:30
  SPEAKER_02: we're on a timeline because if you – I just worry that if we don't,
• 1:01:34
  SPEAKER_02: I think it will take longer than some people think.
• 1:01:37
  And then I think you could just have the rug pulled out of this
• 1:01:41
  whole thing if
• 1:01:42
  SPEAKER_02: you're not able to hit these numbers.
• 1:01:45
  SPEAKER_00: What about – no, I want you to tell.
• 1:01:47
  SPEAKER_01: Well,
• 1:01:48
  one of the reasons that we got excited about Forrest Neurotech
• 1:01:56
  is that I think if
• 1:01:58
  you look at Neuralink has gone in a direction that has invasive
• 1:02:02
  probes that can be
• 1:02:03
  SPEAKER_01: very fine details.
• 1:02:04
  SPEAKER_01: But in return for getting those fine details around any one link,
• 1:02:10
  SPEAKER_01: you can only get a certain number of channels.
• 1:02:12
  One link is not going to see things that are extremely far from
• 1:02:14
  that link.
• 1:02:15
  SPEAKER_02: You're like in the motor cortex.
• 1:02:17
  SPEAKER_02: Yeah,
• 1:02:17
  it's a local stuff in one part of the motor cortex and then you
• 1:02:20
  need to go,
• 1:02:20
  SPEAKER_01: if you want to look at some other part or especially deep brain areas,
• 1:02:23
  SPEAKER_01: you need to add on something else,
• 1:02:25
  SPEAKER_01: another link or something.
• 1:02:29
  SPEAKER_01: On the other hand,
• 1:02:29
  SPEAKER_01: you have these non-invasive methods,
• 1:02:30
  SPEAKER_01: and that's sort of where Kernel has gone,
• 1:02:33
  SPEAKER_01: using light,
• 1:02:34
  SPEAKER_01: for example,
• 1:02:35
  to try to go in and the light enters and interacts with the
• 1:02:38
  neurons and then comes
• 1:02:39
  SPEAKER_01: back.
• 1:02:40
  But that's kind of fuzzy and coarse because you have to go
• 1:02:43
  through the skull,
• 1:02:44
  SPEAKER_01: and as things go through the skull,
• 1:02:45
  SPEAKER_01: they kind of spread out,
• 1:02:46
  SPEAKER_01: and it's hard to localize something.
• 1:02:49
  SPEAKER_01: So ultrasound is a really, really interesting trade-off point.
• 1:02:54
  SPEAKER_01: where it's a kind of nice property of sound,
• 1:02:58
  SPEAKER_01: basically,
• 1:02:59
  SPEAKER_01: that the wavelengths of ultrasound,
• 1:03:02
  SPEAKER_01: okay,
• 1:03:03
  SPEAKER_01: it's a wave,
• 1:03:04
  SPEAKER_01: sound is a wave,
• 1:03:05
  and the wavelengths of ultrasound that can penetrate relatively
• 1:03:10
  efficiently through
• 1:03:11
  SPEAKER_01: brain tissue,
• 1:03:11
  and therefore maybe you put your probe on one side of the brain
• 1:03:14
  that emits and
• 1:03:14
  SPEAKER_01: detects the ultrasound,
• 1:03:16
  SPEAKER_01: and then you send the ultrasound,
• 1:03:17
  SPEAKER_01: it could see all the way to the other end of the brain.
• 1:03:19
  SPEAKER_01: So the ultrasound propagates through the entire brain.
• 1:03:23
  SPEAKER_01: it can do that even at small enough wavelengths,
• 1:03:26
  SPEAKER_01: okay,
• 1:03:27
  SPEAKER_01: that it can also see fine details,
• 1:03:30
  SPEAKER_01: right?
• 1:03:30
  So it's a balance between how fine of details you can see and
• 1:03:33
  field of view.
• 1:03:35
  And so ultrasound has a unique trade-off between field of view
• 1:03:38
  and resolution.
• 1:03:41
  SPEAKER_01: And so if you can,
• 1:03:42
  SPEAKER_01: especially if you can put an ultrasound chip,
• 1:03:45
  miniaturize it so that the emitter detector chip of ultrasound
• 1:03:48
  is on the inside of
• 1:03:49
  SPEAKER_01: the skull,
• 1:03:50
  SPEAKER_01: but it can do that without having to touch the brain,
• 1:03:52
  SPEAKER_01: right?
• 1:03:53
  SPEAKER_01: So that's very powerful, right?
• 1:03:54
  That could be a safer,
• 1:03:56
  but also more general because it allows you to see the whole
• 1:03:58
  brain.
• 1:03:59
  SPEAKER_01: Right, absolutely.
• 1:04:00
  SPEAKER_00: So there's this resolution scale trade-off.
• 1:04:04
  SPEAKER_00: There's this author, how invasive, how dangerous, what are the downsides?
• 1:04:11
  I think that we are a foreign place from a healthy individual
• 1:04:17
  opting into a highly
• 1:04:20
  SPEAKER_00: invasive,
• 1:04:21
  SPEAKER_00: potentially very dangerous surgery.
• 1:04:23
  If you talk to neurosurgeons,
• 1:04:25
  and this is,
• 1:04:25
  well,
• 1:04:25
  it's like,
• 1:04:26
  just a crinectomy.
• 1:04:27
  SPEAKER_00: We're just taking off.
• 1:04:29
  Weirdly importantly,
• 1:04:30
  a non-circular part of the skull because of how billing works.
• 1:04:34
  SPEAKER_00: They only draw a circle.
• 1:04:36
  SPEAKER_00: It's really bad for the neurosurgeons.
• 1:04:39
  SPEAKER_00: It's not a circle.
• 1:04:39
  SPEAKER_02: Because of the way the tool works?
• 1:04:41
  SPEAKER_02: Okay.
• 1:04:43
  SPEAKER_01: The procedure.
• 1:04:43
  SPEAKER_01: The procedure.
• 1:04:45
  In terms of reimbursement of the procedure,
• 1:04:46
  it's better if it's not a circle.
• 1:04:47
  SPEAKER_01: Oh.
• 1:04:48
  SPEAKER_01: That's a detail.
• 1:04:49
  SPEAKER_01: Yeah.
• 1:04:50
  SPEAKER_01: Okay.
• 1:04:51
  SPEAKER_01: It's a small, small, small...
• 1:04:54
  SPEAKER_01: Removal of part of the skull.
• 1:04:55
  SPEAKER_02: Oh, you said because of the way the billing works.
• 1:04:57
  SPEAKER_02: Yeah.
• 1:04:57
  SPEAKER_02: Okay.
• 1:04:58
  SPEAKER_02: Okay.
• 1:04:59
  SPEAKER_02: Should we explain that?
• 1:05:02
  SPEAKER_00: The way that surgeons get billed is per procedure.
• 1:05:05
  SPEAKER_00: And the procedure that drills a round hole is less...
• 1:05:10
  lucrative for the hospital and for the surgeon than a procedure
• 1:05:14
  that doesn't drill
• 1:05:15
  SPEAKER_00: a round hole.
• 1:05:15
  SPEAKER_01: Okay, great.
• 1:05:16
  SPEAKER_01: A neurosurgeon just has to drill with a circular drill.
• 1:05:19
  SPEAKER_01: They don't have to do any work.
• 1:05:20
  SPEAKER_01: They just have to drill something.
• 1:05:22
  SPEAKER_00: A neurosurgeon saying,
• 1:05:24
  SPEAKER_00: oh,
• 1:05:24
  SPEAKER_00: just the skull versus past the dura into the brain,
• 1:05:29
  SPEAKER_00: they're just radically different.
• 1:05:30
  SPEAKER_00: The chance of infection is radically different.
• 1:05:33
  SPEAKER_00: The chance of negative side effects is radically different.
• 1:05:37
  SPEAKER_00: They're just like...
• 1:05:38
  To me,
• 1:05:39
  it's all brain surgery,
• 1:05:40
  but to a neurosurgeon,
• 1:05:41
  they're vastly different things.
• 1:05:42
  SPEAKER_00: And so ultrasound has this ability to both be fairly fine detail,
• 1:05:49
  SPEAKER_00: fairly large scope,
• 1:05:50
  SPEAKER_00: can see most of the brain,
• 1:05:52
  and is not nearly as dangerous as things that actually have to
• 1:05:57
  enter the brain.
• 1:05:58
  SPEAKER_02: And I don't think I'm revealing anything I'm not supposed to,
• 1:06:01
  SPEAKER_02: but with Forrest,
• 1:06:02
  SPEAKER_02: you do this implant,
• 1:06:05
  SPEAKER_02: like you're saying,
• 1:06:05
  SPEAKER_02: I think it sits on top of the dura,
• 1:06:07
  SPEAKER_02: that's what we're talking about,
• 1:06:08
  SPEAKER_02: and then the idea is that
• 1:06:10
  SPEAKER_02: Um,
• 1:06:10
  SPEAKER_02: the patients can have this in when they're typically,
• 1:06:14
  SPEAKER_02: if we're doing stuff like this is like,
• 1:06:15
  you're trying to measure mental illness on someone while
• 1:06:18
  they're in a hospital in
• 1:06:20
  SPEAKER_02: unusual circumstances for just a few hours.
• 1:06:22
  SPEAKER_02: And it's like,
• 1:06:23
  SPEAKER_02: what are you,
• 1:06:23
  SPEAKER_02: you're not like,
• 1:06:24
  SPEAKER_02: what is the baseline you're even seeing from this data?
• 1:06:26
  SPEAKER_02: Whereas with this, you could have it for a month for six weeks.
• 1:06:29
  SPEAKER_02: You could be at home forever.
• 1:06:32
  SPEAKER_01: That's another thing overall that these companies,
• 1:06:34
  SPEAKER_01: I think Neuralink deserves a lot of credit for.
• 1:06:36
  SPEAKER_01: This is sort of making a really miniaturized,
• 1:06:38
  really good piece of electronics so that it's not connected to
• 1:06:41
  this giant mainframe
• 1:06:42
  SPEAKER_01: kind of machine.
• 1:06:44
  SPEAKER_02: People do not get this.
• 1:06:45
  SPEAKER_01: That means you can really, this context of use really changes.
• 1:06:49
  SPEAKER_01: And so that's, I mean, that's powerful.
• 1:06:50
  SPEAKER_01: And of course, Forrest would have a similar
• 1:06:52
  property of being able to be fully wireless so then you can
• 1:06:55
  actually see when
• 1:06:56
  you're depressed in your normal environment and get something
• 1:06:59
  that's like possibly
• 1:07:00
  real data yeah people do not appreciate this even though it's
• 1:07:02
  like i get that lots
• 1:07:03
  of people hate elon and stuff and neural link seems weird but
• 1:07:07
  when you're with
• 1:07:07
  these patients it's like they're at their house low power
• 1:07:10
  electronics involved in
• 1:07:11
  SPEAKER_01: this is incredible
• 1:07:12
  And some of the other startups,
• 1:07:14
  I kind of,
• 1:07:15
  I'm interested in all these companies.
• 1:07:16
  SPEAKER_02: I'm really not trying to besmirch anyone,
• 1:07:18
  SPEAKER_02: but they always just tell me like,
• 1:07:19
  SPEAKER_02: oh,
• 1:07:19
  SPEAKER_02: it's like,
• 1:07:20
  SPEAKER_02: you know,
• 1:07:20
  SPEAKER_02: well,
• 1:07:20
  SPEAKER_02: Neuralink has these problems.
• 1:07:21
  And then,
• 1:07:22
  um,
• 1:07:23
  we're also going to make this compact device,
• 1:07:26
  like in nine months and stuff.
• 1:07:28
  And I'm always like,
• 1:07:28
  well,
• 1:07:28
  it takes,
• 1:07:29
  you know,
• 1:07:29
  you guys might be underestimating me.
• 1:07:33
  SPEAKER_02: That part,
• 1:07:33
  because I think they don't get enough credit for some of that
• 1:07:36
  because it's very
• 1:07:37
  SPEAKER_02: hard.
• 1:07:38
  SPEAKER_01: It's just engineering, but it's incredibly hard.
• 1:07:40
  Just the manufacturing line just has to not leak anything and
• 1:07:44
  last for a really
• 1:07:45
  SPEAKER_02: long time and be tiny and be wireless.
• 1:07:49
  SPEAKER_02: Many steps.
• 1:07:51
  SPEAKER_02: Okay, I'm not going to keep you guys too, too long.
• 1:07:52
  SPEAKER_02: Oh, wait.
• 1:07:54
  SPEAKER_02: So I knew 2017 Brian Johnson.
• 1:07:58
  SPEAKER_02: Did you imagine him becoming...
• 1:08:00
  SPEAKER_02: worldwide longevity guru.
• 1:08:02
  SPEAKER_01: I love the guy.
• 1:08:07
  I think that there's some shared elements of his zeitgeist over
• 1:08:12
  that time.
• 1:08:12
  SPEAKER_01: I didn't imagine the specific thing.
• 1:08:14
  SPEAKER_02: There were hints, for sure.
• 1:08:17
  SPEAKER_01: I think his overall personality stayed pretty similar.
• 1:08:20
  SPEAKER_01: He was great to work with, yeah.
• 1:08:22
  SPEAKER_02: And when you guys, we talked about AI and how it's changing your thing.
• 1:08:29
  SPEAKER_02: What are you most excited about?
• 1:08:31
  SPEAKER_02: Since you've done these 10 areas,
• 1:08:33
  SPEAKER_02: what's like the coolest thing that you're hunting for right now?
• 1:08:38
  SPEAKER_01: I mean, a lot of it is actually just
• 1:08:41
  SPEAKER_01: AI-related safety security and biosecurity.
• 1:08:46
  SPEAKER_01: I mean, those are two areas where we really want to push more.
• 1:08:48
  SPEAKER_01: Are you doomers?
• 1:08:49
  SPEAKER_01: This is reducing the downside risk.
• 1:08:51
  SPEAKER_01: Not really,
• 1:08:52
  SPEAKER_01: but I think that there is such a thing as...
• 1:08:54
  SPEAKER_01: I mean,
• 1:08:54
  SPEAKER_01: I think we really like this defensive acceleration framework.
• 1:08:57
  Can you accelerate a technology which makes the civilization
• 1:09:00
  more robust to these
• 1:09:02
  SPEAKER_01: problems,
• 1:09:02
  SPEAKER_01: right?
• 1:09:02
  So the formal methods thing I mentioned with provably secure
• 1:09:05
  software is one
• 1:09:06
  SPEAKER_01: example of that,
• 1:09:07
  SPEAKER_01: yeah.
• 1:09:07
  SPEAKER_00: I actually like...
• 1:09:08
  SPEAKER_00: pretty concretely were not doomers.
• 1:09:11
  SPEAKER_00: We have a shared... Admit it.
• 1:09:14
  SPEAKER_02: No, I'm just kidding.
• 1:09:15
  SPEAKER_00: I have some concerns.
• 1:09:16
  SPEAKER_00: I'm not a doomer, but I'm not a doomer.
• 1:09:20
  I think that we see a version of a world...
• 1:09:24
  I think a lot of people are either...
• 1:09:27
  SPEAKER_00: AI bad or AI will solve everything black box.
• 1:09:30
  SPEAKER_00: And I think that we spend a lot of time saying,
• 1:09:33
  what is the version of the world or versions of the world that
• 1:09:37
  we can get to where
• 1:09:37
  SPEAKER_00: this has gone extremely well?
• 1:09:39
  And so what are the downsides to this defensive acceleration
• 1:09:42
  point where we're
• 1:09:42
  SPEAKER_00: like,
• 1:09:43
  SPEAKER_00: OK,
• 1:09:43
  SPEAKER_00: we should mitigate biosecurity risks.
• 1:09:46
  SPEAKER_00: We should mitigate some cybersecurity risks,
• 1:09:48
  SPEAKER_00: whether that be on the hardware side or formal verification,
• 1:09:52
  SPEAKER_00: but also enable
• 1:09:54
  SPEAKER_00: I don't know, why don't we have replicators?
• 1:09:57
  What are the fundamental material science questions to this
• 1:10:00
  tech tree problem?
• 1:10:02
  SPEAKER_00: I'm not building a replicator anytime soon.
• 1:10:04
  What are the fundamental material science questions that
• 1:10:07
  underlie where we could go
• 1:10:09
  SPEAKER_00: if AI could help us advance it quickly?
• 1:10:11
  SPEAKER_00: And so how do we work on those technologies?
• 1:10:15
  SPEAKER_00: on the neuroscience side,
• 1:10:16
  if we think that neuroscience really deeply understanding the
• 1:10:20
  brain will help both
• 1:10:22
  with what happens if we create other intelligence and or how do
• 1:10:27
  we make better
• 1:10:27
  SPEAKER_00: versions of ourselves with new capabilities.
• 1:10:31
  SPEAKER_00: And so picking these things with these broadly optimistic visions of,
• 1:10:35
  SPEAKER_00: you know,
• 1:10:36
  mitigating our downside and kind of creating as much
• 1:10:39
  possibility for the upside as
• 1:10:41
  SPEAKER_00: possible.
• 1:10:41
  AI is also even just thinking a little bit out on the horizon
• 1:10:44
  of what we might do.
• 1:10:45
  SPEAKER_01: I mean,
• 1:10:46
  SPEAKER_01: it's caused us to think a little bit more about social sciences,
• 1:10:48
  SPEAKER_01: actually,
• 1:10:49
  SPEAKER_01: than we had previously.
• 1:10:50
  Most of these things we've been talking about are microscopes
• 1:10:53
  or something like that.
• 1:10:53
  SPEAKER_01: But are there techno-social questions, right?
• 1:10:58
  SPEAKER_01: So can you use AI to improve...
• 1:11:01
  SPEAKER_01: democratic deliberation processes?
• 1:11:03
  Can you use AI to improve human thinking and epistemics even
• 1:11:06
  without a brain interface?
• 1:11:08
  Those types of questions are areas where we want to understand
• 1:11:12
  if you want
• 1:11:12
  Fro-shaped problems or the directed development of fundamental
• 1:11:15
  capabilities.
• 1:11:16
  There's lots of questions about the ideas and we're not trying
• 1:11:18
  to solve necessarily
• 1:11:20
  SPEAKER_01: deep theoretical issues like how to do alignment,
• 1:11:22
  SPEAKER_01: for example,
• 1:11:23
  SPEAKER_01: although I think some neuroscience may contribute to that.
• 1:11:27
  We're trying to find the actionable engineering or directed
• 1:11:30
  research problems,
• 1:11:31
  SPEAKER_01: including in those areas of social sciences.
• 1:11:33
  SPEAKER_01: How do you find those people?
• 1:11:34
  I hope you go on podcasts and you talk about how you want to
• 1:11:37
  find these people,
• 1:11:38
  SPEAKER_00: then they reach out to you.
• 1:11:40
  SPEAKER_02: Are those sociologists?
• 1:11:41
  SPEAKER_02: Are they philosophers?
• 1:11:43
  SPEAKER_01: There's a mix.
• 1:11:43
  SPEAKER_01: There's a set of people thinking about democracy.
• 1:11:47
  SPEAKER_01: that maybe are using some software.
• 1:11:50
  There's software called Polis,
• 1:11:51
  which is sort of about scaling more deliberative processes.
• 1:11:53
  SPEAKER_01: OK, how does that change with AI?
• 1:11:56
  SPEAKER_01: There's been some research from DeepMind,
• 1:11:57
  SPEAKER_01: for example,
• 1:11:57
  SPEAKER_01: on using AI for democratic deliberation purposes.
• 1:12:01
  SPEAKER_01: There's a set of other people that think about recommender systems,
• 1:12:04
  SPEAKER_01: or think about prediction markets,
• 1:12:05
  SPEAKER_01: or think about other types of epistemics,
• 1:12:08
  SPEAKER_01: human thinking type questions.
• 1:12:10
  SPEAKER_01: We go talk to those people,
• 1:12:12
  SPEAKER_01: and we say,
• 1:12:12
  SPEAKER_01: OK,
• 1:12:12
  SPEAKER_01: well,
• 1:12:13
  SPEAKER_01: can you do what you're doing now in your startup in your lab?
• 1:12:17
  SPEAKER_01: Do you need some sort of fundamental change to society for this to work,
• 1:12:20
  SPEAKER_01: in which case we may not be able to help?
• 1:12:22
  SPEAKER_01: But do you need like an engineering project?
• 1:12:24
  SPEAKER_01: Yeah, let's talk, you know?
• 1:12:26
  SPEAKER_02: Yeah.
• 1:12:27
  SPEAKER_02: You nodded when we were talking about this before pretty emphatically,
• 1:12:30
  SPEAKER_02: but I still have my mind just talking to you today.
• 1:12:36
  SPEAKER_02: You know, why are these nerd billionaires not just dumping...
• 1:12:40
  SPEAKER_02: I don't understand.
• 1:12:40
  SPEAKER_01: Even though it's a finite size, they don't have to dump everything.
• 1:12:45
  SPEAKER_01: They have to dump a meaningful amount.
• 1:12:47
  SPEAKER_02: Well, I interviewed Larry Page many years ago.
• 1:12:50
  SPEAKER_02: And I think at the time,
• 1:12:52
  SPEAKER_02: I'm doing this from memory,
• 1:12:53
  SPEAKER_02: and I think he was in the top 10 richest people in the world.
• 1:12:57
  SPEAKER_02: I think he probably had roughly like 25 billion.
• 1:12:59
  SPEAKER_02: And at the time, because Elon was not Elon, these companies were like,
• 1:13:03
  doing okay but but kind of not anywhere near where they were
• 1:13:07
  all his wealth was
• 1:13:08
  tied up in Tesla see I think he was worth like 400 million or
• 1:13:11
  something like that
• 1:13:11
  which is great but you know Larry Larry was like oh I'm so
• 1:13:14
  jealous he told me he's
• 1:13:16
  SPEAKER_02: like you know I wish I could do what Elon does and and I
• 1:13:20
  it's like i kind of got some of it was that part of it i think
• 1:13:23
  was like he was ceo
• 1:13:24
  of a public company and there's baggage that comes with that
• 1:13:27
  and elon could be
• 1:13:28
  that's free radical um to some degree and then that if larry
• 1:13:34
  does a flying car
• 1:13:34
  company and it doesn't work well it's sort of like reflects
• 1:13:38
  kind of badly on him
• 1:13:39
  and on google and it was like why you know who was funding what
• 1:13:43
  and there were
• 1:13:44
  SPEAKER_02: extra questions however you know
• 1:13:47
  SPEAKER_02: When I look at the list of like the world's 50 richest people,
• 1:13:51
  SPEAKER_02: and now we're talking about Larry's not worth,
• 1:13:54
  SPEAKER_02: and I'm not just,
• 1:13:55
  SPEAKER_02: he's a very nice guy.
• 1:13:55
  SPEAKER_02: I ran into it the other day.
• 1:13:56
  SPEAKER_02: I'm not trying to like totally besmirch him,
• 1:13:59
  SPEAKER_02: but you know,
• 1:14:00
  SPEAKER_02: these people are worth like 80 billion now,
• 1:14:02
  SPEAKER_02: 120 billion.
• 1:14:03
  SPEAKER_02: Yeah.
• 1:14:03
  SPEAKER_02: It just, it's crazy.
• 1:14:04
  SPEAKER_01: They can fund the catalytic activity to unlock the entire gap map.
• 1:14:08
  SPEAKER_01: For example, I think with private, you could do it.
• 1:14:10
  SPEAKER_02: And they're all, they're all science minded.
• 1:14:14
  SPEAKER_01: I think partly they're busy.
• 1:14:16
  SPEAKER_01: Partly they're running companies and busy and stuff.
• 1:14:19
  SPEAKER_01: Partly they want to make very high-quality decisions,
• 1:14:21
  SPEAKER_01: and the marketplace of philanthropy is very opaque.
• 1:14:22
  SPEAKER_01: How do you know who is the right scientist?
• 1:14:24
  You can go to Stanford,
• 1:14:25
  go to another university,
• 1:14:27
  and there's lots and lots of options.
• 1:14:29
  SPEAKER_01: If your money's being spent well.
• 1:14:31
  SPEAKER_01: How do you know you're doing it well?
• 1:14:34
  SPEAKER_00: In some ways, I have a lot of empathy or sympathy for philanthropists.
• 1:14:38
  SPEAKER_00: If I'm running a venture firm, I'm making a pretty clear deal.
• 1:14:42
  SPEAKER_00: I'm optimizing for dollars in the door later.
• 1:14:45
  SPEAKER_00: If I put out dollars now...
• 1:14:47
  SPEAKER_00: Will this or will this not return multiple X in 10 years?
• 1:14:51
  SPEAKER_00: And it's like a very clear trade-off.
• 1:14:54
  SPEAKER_00: If you're a philanthropist, you're like, well, do I donate to the Met?
• 1:14:57
  SPEAKER_00: Do I fund new artists?
• 1:15:00
  SPEAKER_00: Do I help save lives in sub-Saharan Africa?
• 1:15:04
  SPEAKER_00: Science storytellers.
• 1:15:06
  SPEAKER_00: Do you fund science storytellers?
• 1:15:08
  SPEAKER_00: Do you fund scientists?
• 1:15:11
  SPEAKER_00: Do you fund the existing scientists who already have a network?
• 1:15:15
  SPEAKER_00: Do you fund new scientists?
• 1:15:16
  SPEAKER_00: It's a...
• 1:15:17
  SPEAKER_00: There's no money back later on the other side of that experiment.
• 1:15:20
  SPEAKER_00: You have to be like, do I think that worked?
• 1:15:22
  SPEAKER_00: And I think it's a complicated thing.
• 1:15:24
  And I think that many philanthropists then build organizations
• 1:15:27
  to answer the complicated thing.
• 1:15:29
  And then those organizations become beholden to the things that
• 1:15:32
  all organizations
• 1:15:33
  SPEAKER_00: are beholden to of internal decision-making and internal processes.
• 1:15:37
  SPEAKER_00: And I think that it's complicated.
• 1:15:39
  SPEAKER_00: I think that our hope...
• 1:15:41
  is to build convergent research into an organization that can
• 1:15:46
  start to build things
• 1:15:47
  like the gap map and a list of what we think are the highest
• 1:15:53
  leverage bets you can
• 1:15:54
  SPEAKER_00: make in science and technology.
• 1:15:57
  SPEAKER_00: And say,
• 1:15:57
  SPEAKER_00: actually,
• 1:15:57
  SPEAKER_00: here is a bet that is $2 million to unlock a seedling that would,
• 1:16:03
  SPEAKER_00: then if that proved out,
• 1:16:04
  be a $40 million bet that then would transform this field and
• 1:16:08
  make a series of
• 1:16:09
  SPEAKER_00: things
• 1:16:10
  fairly robust roadmaps and conclusions to help people navigate
• 1:16:15
  this decision.
• 1:16:17
  SPEAKER_00: Conversion research, much like science storytellers, also has to go.
• 1:16:20
  So figuring out how to build an organization where that is a
• 1:16:25
  valued thing from the
• 1:16:26
  philanthropic community of really doing the deep dive into all
• 1:16:30
  of the different
• 1:16:31
  areas of science and across the board saying what would be
• 1:16:34
  highly impactful.
• 1:16:35
  SPEAKER_00: Okay.
• 1:16:36
  All right,
• 1:16:37
  let's end on a completely unfair,
• 1:16:41
  open-ended,
• 1:16:42
  impossible question.
• 1:16:43
  SPEAKER_02: Just because I feel like you're both so into deep science,
• 1:16:47
  SPEAKER_02: you're kind of sci-fi type thinkers and everything.
• 1:16:50
  SPEAKER_02: I mean, give me, and look, again, I just said it's unfair.
• 1:16:54
  SPEAKER_02: There's many, many wild cards.
• 1:16:57
  Like,
• 1:16:57
  give me,
• 1:16:57
  what's your vision of the next,
• 1:17:00
  where are we in like 100 years?
• 1:17:02
  SPEAKER_02: I mean,
• 1:17:02
  SPEAKER_02: are we about to unlock,
• 1:17:04
  SPEAKER_02: you know,
• 1:17:04
  SPEAKER_02: some great flourishing,
• 1:17:07
  SPEAKER_02: given the caveats I've already put in place?
• 1:17:09
  SPEAKER_02: When you go to sleep at night, what are you guys picturing?
• 1:17:14
  SPEAKER_01: I mean, it seems very volatile right now.
• 1:17:18
  SPEAKER_01: It does.
• 1:17:18
  SPEAKER_01: I mean,
• 1:17:19
  SPEAKER_01: the question doesn't mean I'm a quote-unquote doomer,
• 1:17:22
  SPEAKER_01: but I think it does feel very volatile,
• 1:17:23
  SPEAKER_01: so I think we have to navigate it carefully.
• 1:17:27
  SPEAKER_01: I mean, one of the things, I do think the neuroscience thing is...
• 1:17:30
  SPEAKER_01: is a big question.
• 1:17:33
  SPEAKER_01: And I think maybe in the years shortly after we have super advanced AI,
• 1:17:37
  SPEAKER_01: we need to spend a lot of time on neuroscience.
• 1:17:40
  SPEAKER_01: We may have super advanced AI and not know if it's conscious.
• 1:17:42
  SPEAKER_01: We may not know what determines whether we're conscious.
• 1:17:47
  SPEAKER_01: We may have super advanced AI and not know how to retune our own brains.
• 1:17:51
  So basically the tunability and understanding of the human
• 1:17:55
  brain,
• 1:17:56
  I think that's kind of the
• 1:17:58
  It's actually really hard for me to know what 2075 looks like
• 1:18:03
  without that.
• 1:18:04
  SPEAKER_01: So I think you can say, oh, we'll have rockets in space.
• 1:18:06
  SPEAKER_01: And we'll have a lot of low energy compute capacity.
• 1:18:09
  SPEAKER_01: And we'll have secured the world against bio threats and stuff like that.
• 1:18:13
  SPEAKER_01: And we'll be able to then do incredible
• 1:18:16
  SPEAKER_01: you know, accelerated, you know, research on aging and other diseases.
• 1:18:19
  SPEAKER_01: I mean, this is, this is all great stuff.
• 1:18:22
  SPEAKER_01: But like, what does 2745 actually look like?
• 1:18:24
  SPEAKER_01: I mean, I'm not sure.
• 1:18:25
  SPEAKER_01: Greg Egan,
• 1:18:26
  SPEAKER_01: maybe as a sci-fi author,
• 1:18:27
  SPEAKER_01: he's tried to go out there and very few have tried to go out.
• 1:18:31
  SPEAKER_02: I mean,
• 1:18:31
  SPEAKER_02: you make the,
• 1:18:33
  SPEAKER_02: you make the point that I'm,
• 1:18:35
  SPEAKER_02: that I'm betting on for,
• 1:18:36
  SPEAKER_02: I think,
• 1:18:37
  SPEAKER_02: I don't know,
• 1:18:37
  SPEAKER_02: who knows what it is.
• 1:18:38
  SPEAKER_02: Let's call it like the next 20 years of my life.
• 1:18:40
  SPEAKER_02: I mean, I think it's the, yeah, it's this brain, brain on brain,
• 1:18:45
  SPEAKER_02: Yeah.
• 1:18:45
  SPEAKER_02: Interplay questions, knowledge, I just think that part, I think.
• 1:18:49
  SPEAKER_01: I think that's,
• 1:18:50
  SPEAKER_01: there's still science in between here and there actually,
• 1:18:53
  SPEAKER_01: which would,
• 1:18:54
  I think our concepts in 2075 of what are we trying to do may be
• 1:18:57
  different because
• 1:18:58
  SPEAKER_01: of neuroscience.
• 1:18:58
  SPEAKER_01: Yeah.
• 1:18:59
  SPEAKER_02: All right, Adam took us back to 2075.
• 1:19:00
  SPEAKER_02: I did 100 years.
• 1:19:01
  SPEAKER_02: It was probably kind of gross and not really unfair.
• 1:19:03
  SPEAKER_02: You pick any timeline you want.
• 1:19:06
  SPEAKER_00: So I actually think I have high hopes for 100 years.
• 1:19:09
  SPEAKER_00: I mean, I really think they're at a steep part of the curve.
• 1:19:12
  SPEAKER_00: In 100 years to now, there's so much that we can even accomplish.
• 1:19:16
  SPEAKER_00: I think the next 20 years will be weird, man.
• 1:19:18
  SPEAKER_03: Weird, yeah.
• 1:19:19
  SPEAKER_00: Yeah.
• 1:19:19
  And so I think that in 100 years,
• 1:19:22
  I don't think that the replicator is impossible,
• 1:19:25
  right?
• 1:19:25
  SPEAKER_00: I don't think that, you know...
• 1:19:28
  SPEAKER_00: true high bandwidth BCIs are impossible in 100 years.
• 1:19:31
  SPEAKER_00: Like, I'm not sure about escape velocity for longevity.
• 1:19:34
  SPEAKER_00: I'm,
• 1:19:35
  SPEAKER_00: like,
• 1:19:35
  SPEAKER_00: not convinced that our meat sacks are meant to do that,
• 1:19:38
  SPEAKER_00: but,
• 1:19:38
  SPEAKER_00: like,
• 1:19:39
  SPEAKER_00: I think there's a lot possible...
• 1:19:40
  SPEAKER_02: I'm optimistic about that.
• 1:19:41
  SPEAKER_02: You are.
• 1:19:41
  SPEAKER_02: Yeah.
• 1:19:42
  SPEAKER_02: That's good to hang out with Brian, man.
• 1:19:44
  SPEAKER_00: Anyway, I'm sorry.
• 1:19:45
  SPEAKER_00: But, like, I think, like, 100 years, a lot of these things are possible.
• 1:19:48
  SPEAKER_00: I think that building...
• 1:19:50
  What are the bets that we can make in these next three years
• 1:19:54
  that make the
• 1:19:55
  SPEAKER_00: following 20
• 1:19:57
  SPEAKER_00: the best version is, like, a thing I'm thinking a lot about.
• 1:20:00
  SPEAKER_00: Okay.
• 1:20:00
  SPEAKER_02: Okay.
• 1:20:01
  SPEAKER_02: All right.
• 1:20:01
  SPEAKER_02: Next time you guys are on,
• 1:20:02
  SPEAKER_02: we're talking partial cell reprogramming,
• 1:20:05
  SPEAKER_02: IPSCs,
• 1:20:06
  SPEAKER_02: and NeuroPixels,
• 1:20:07
  SPEAKER_02: which are three things I didn't have time to get to.
• 1:20:12
  SPEAKER_02: Thank you so much for coming on.
• 1:20:15
  SPEAKER_02: Are you going back to Edge, Esmeralda?
• 1:20:17
  SPEAKER_00: Yeah.
• 1:20:17
  SPEAKER_02: For a little bit, yeah.
• 1:20:18
  SPEAKER_00: Actually, Hot Edge, if you go, Avery, whose last name?
• 1:20:21
  SPEAKER_00: I don't remember.
• 1:20:22
  SPEAKER_01: Avery Krieger.
• 1:20:23
  SPEAKER_00: Avery Krieger is there.
• 1:20:24
  SPEAKER_00: Krieger.
• 1:20:26
  collecting a bunch of neurodata and phenotypic behavioral data
• 1:20:31
  to build a giant
• 1:20:32
  SPEAKER_00: data set all month.
• 1:20:33
  So we're just getting like hundreds of hours of data as a pilot
• 1:20:37
  project for this crazy thing.
• 1:20:39
  SPEAKER_00: And this is the kind of stuff people are building at it.
• 1:20:41
  SPEAKER_01: With like an EEG or?
• 1:20:42
  SPEAKER_01: A lot of it is about the behavioral capture.
• 1:20:45
  SPEAKER_01: Okay.
• 1:20:45
  SPEAKER_01: Yeah, it's interesting.
• 1:20:47
  SPEAKER_01: So you don't get a neuropixel at the end.
• 1:20:48
  SPEAKER_01: I mean,
• 1:20:48
  SPEAKER_01: they ultimately want to say,
• 1:20:49
  SPEAKER_01: you know,
• 1:20:50
  SPEAKER_01: neuropsychiatry or neurosurgery or so on neurology.
• 1:20:52
  SPEAKER_01: They want to be able to just capture a richer human.
• 1:20:55
  SPEAKER_00: He has an EEG there too.
• 1:20:57
  SPEAKER_00: He got a bunch of equipment donated.
• 1:20:58
  So he's doing like an hour of neurodata for 10 hours of
• 1:21:02
  behavioral data and just
• 1:21:04
  SPEAKER_00: building a big data set.
• 1:21:06
  SPEAKER_02: Okay.
• 1:21:06
  SPEAKER_02: Okay.
• 1:21:07
  Do you guys,
• 1:21:07
  you must get bombarded with like,
• 1:21:10
  should people send you their wild ideas?
• 1:21:14
  SPEAKER_01: Abstracts at convergentresearch.org.
• 1:21:16
  SPEAKER_01: We have a particular one pager format.
• 1:21:22
  SPEAKER_01: And we can canvas those.
• 1:21:23
  And I think hopefully from this discussion,
• 1:21:25
  you understand a little bit what the fro shape is.
• 1:21:27
  SPEAKER_01: Don't send us a one person project or a $1 trillion project or, you know.
• 1:21:32
  SPEAKER_02: What's the sweet spot?
• 1:21:33
  SPEAKER_02: It's 20 to 50?
• 1:21:34
  SPEAKER_01: Yeah,
• 1:21:35
  SPEAKER_01: think about what you do with,
• 1:21:36
  SPEAKER_01: let's say,
• 1:21:37
  15 to 30 people over five years when those people can be
• 1:21:40
  professionals in the same
• 1:21:42
  SPEAKER_01: way that deep tech company hires professional operators,
• 1:21:45
  SPEAKER_01: professional chip designers,
• 1:21:46
  SPEAKER_01: professional biologists,
• 1:21:47
  SPEAKER_01: whatever.
• 1:21:48
  okay well there you go if somebody pitches you guys and it gets
• 1:21:51
  turned into a thing
• 1:21:52
  we get the exclusive and claiming claiming rights on anything
• 1:21:56
  derived from this
• 1:21:58
  this podcast to to uh not on the technology but on the story on
• 1:22:01
  the story um uh
• 1:22:03
  thank you guys so much for visiting i really appreciate it two
• 1:22:06
  of my favorite most
• 1:22:07
  SPEAKER_02: interesting people um thank you so much all right cheers
• 1:22:14
  SPEAKER_02: Core Memory is hosted by me, Ashley Vance.
• 1:22:17
  SPEAKER_02: The show is produced by David Nicholson and me.
• 1:22:21
  SPEAKER_02: Our theme music is by James Mercer and John Sortland.
• 1:22:25
  SPEAKER_02: And the show is edited by the John Sortland.
• 1:22:30
  SPEAKER_02: Thank you all for listening.
• 1:22:32
  SPEAKER_02: As mentioned, the show is sponsored by Ewan Ventures.
• 1:22:36
  SPEAKER_02: Thanks.