คำบรรยาย (194)
0:08Manoush Zomorodi: OK. So Amen,
I gave this shortest bit of your bio.
0:12But tell us the story
of how you got to be working with Bezos,
0:16your sort of trajectory to being here.
0:18ARM: Yeah. You still have me
blushing, nonetheless.
0:23It's really interesting
that you asked that question,
0:25because my pathway to the Bezos Earth Fund
0:28is almost polar opposite
to how we think about our pathway
0:34to adopting and using AI to accelerate
climate and nature solutions.
0:38And I'll explain why,
0:40really, in some quick points.
0:43I -- undergrad, master's, doctorate,
computer science, computer scientist,
0:48research labs -- did the whole thing.
0:50I was one of those computer scientists
that believed in computer science,
0:54you know, algorithm optimization.
0:57Through a couple
of personal things that took place,
0:59I realized that that was only a mechanism
by which I could do other things,
1:05which is have an impact.
1:06So then I began to chase problems
you mentioned here.
1:10I was the chief analytics officer
for the City of New York.
1:13How do we solve problems here?
1:16And then, you know,
coming to the Bezos Earth Fund,
1:19how do we use AI, computer science
to solve climate and nature problems?
1:24And so I was AI in search of a problem.
1:29At the Bezos Earth Fund,
1:30we think about starting
with a problem first
1:34and understanding that problem,
1:35and then looking for ways to use modern AI
1:38in order to scale solutions in that space.
1:41MZ: OK, so let's go deeper into that.
1:43How are you looking at different
projects that are out there?
1:46What are sort of the big ideas
that you're using
1:49to sort of lead you to find
what you want to fund?
1:53ARM: So, internally,
we have a mental model
1:56that we use to really get there.
1:58We think about this difference
between inventions and discoveries.
2:03And the way you want to think about that
is a telescope is an invention,
2:08looking through the telescope
to notice that Jupiter has moons
2:12is the discovery, right?
2:14And so for us, when we look at it,
2:17it's how do we identify
big, big innovations,
2:22grand innovations that have an impact
such that you can have discoveries
2:27that then have an impact
in climate and nature.
2:29And so we look for projects
2:31and efforts that sort of
go across that mental model.
2:36MZ: So before we get
into the discovery part,
2:38let's talk about the tool.
2:40Where are we when it comes to AI?
2:43I know there are some people
who might think,
2:45"What do you mean? We're at ChatGPT 5."
2:47But like from your perspective,
much different,
2:52where do you think we are?
2:54ARM: So I could spend hours
talking about digital twins,
3:00Earth observation models,
edge AI and all of those things.
3:03But one of the things
that have resonated with me
3:07is this concept called move 37.
3:11So move 37 was this move that AlphaGo,
3:15when playing against Go champion,
early on in the game,
3:21did a move that was
counterintuitive to all experts.
3:25It made no sense to any Go expert,
3:28but it was the move
that ultimately won the game.
3:32And so where AI is, is these two places.
3:36Right now, it's at a place
where it answers questions
3:41based on what it knows, right?
3:43It takes an average of reality
and then gives you answers.
3:49Move 37 was this view
into how AI can be creative
3:53and actually come up with a move
that no one has ever thought of,
3:58and it was counterintuitive to use.
3:59And so we really want to get to a place
where in climate and nature,
4:04AI is actually offering solutions,
4:08creative solutions
that even the world's greatest experts
4:13find counterintuitive,
4:15but are actually really powerful.
4:17MZ: Do you have an example of something
that's maybe happening already
4:21that demonstrates that?
4:24one of the projects
that really goes across this mental model
4:30is Meta really came up
with this AI innovation, invention
4:38which is a computer vision model,
very powerful computer vision model.
4:41And then they matched it
with satellite data.
4:46And it's really powerful innovation.
4:50But what they did was partner with WRI
in its restoration efforts,
4:56such that you could actually
track the growth of trees
5:00to an 80 percent accuracy of field surveys
5:05at three percent of the cost.
5:07And so now you can actually unlock
5:12performance-based financing
with this technology.
5:15So it followed that mental model
of grand innovation and invention,
5:22and ultimately a discovery
that leads to an impact.
5:29the whole thing about that
is we haven't gotten there yet,
5:32and that's where we should be going,
5:34which is there are restoration solutions
that people are using.
5:40"Tell me some of the best ways to do
restoration in this particular area,"
5:44what it's going to do
is identify an average or interpolation
5:49of the existing good solutions.
5:52What we want is AI
to come up with something
5:54that no one in the room can come up with
when it comes to restoration,
5:58and that's the trajectory.
6:00MZ: What's the timeline look for that?
6:02How will we know when
we have sort of hit that tipping point?
6:05ARM: One, there has to be
trust by the experts
6:09and the experts are using it,
6:11but then also there has to be a mechanism
6:13by which everyday people
who are living their lives,
6:18who are living in these regions
that we're concerned about,
6:22who are doing the work on the ground,
6:25can trust and use these tools as well.
6:28There is anyone who gives
you an exact number.
6:34Does it know the number,
how long it's going to take?
6:37But that's where we have to get to,
that level of trust
6:39and that level of use
across a number of types of people.
6:44MZ: I want to be sure to ask you,
6:46because there are many people
who say that the same tech giants
6:53are also responsible
for a lot of the environmental harms,
6:57and that their climate initiatives
essentially amount to greenwashing.
7:01How do you respond to that?
7:03ARM: You know, at the Bezos Earth Fund,
we believe that on balance,
7:07AI is going to be a tool
and a force for good
7:11and a tool and a force
for saving the planet.
7:14We have to acknowledge
that AI does contribute to degradation
7:21and challenges when it comes
to the environment.
7:25There are many, many solutions
that a lot of these companies,
7:29a lot of NGOs,
a lot of academic institutions,
7:33and a lot of governments
are applying in this space.
7:36And we will continue
as the Bezos Earth Fund
7:39to support those type of efforts,
7:41such that we are deliberate
in meeting that broad statement
7:46that AI, on balance, will have
a positive impact on the planet.
7:51MZ: I mean, it makes me nervous
7:53because it’s like,
“Let’s hope it works” a little bit.
7:56What are some of the sort of milestones
that we need to be looking for
8:03ARM: So I was listening
to a panel the other day,
8:09and someone said something
along the lines of,
8:13"Every time you do a query on ChatGPT,
8:16It's like throwing away
a bottle of water on the ground."
8:21And as soon as they made that statement,
8:23they said, "You know,
I don't know if that's true,
8:26but it sounds, you know,
like it might be true."
8:32One of the things
that we need to begin to do
8:35is to have precise accuracy
and understanding
8:40of exactly the impact
that AI is having on our environment
8:45and a shared understanding
across the board,
8:49such that we can make statements
that we all agree on,
8:53such that we can identify the solutions.
8:55So the first milestone, which will
include a level of transparency,
9:01a lot of information and data,
9:05such that we can really
get to a place of agreeing
9:08on exactly what those challenges are.
9:11The next milestone is because,
as we speak -- as you mentioned,
9:14I came to the Bezos
Earth Fund from NVIDIA --
9:17as we speak, companies are shifting
9:21how they build technology to support AI.
9:26For instance, cooling is no longer --
9:30just cooling at the data center level
9:32has shifted to now there are mechanisms
where you can cool at the chip level,
9:36such that the burden on water
is not so great.
9:42So these are the types of milestones
that would have to be in place.
9:45MZ: So I guess I want to end by saying,
you know, it's an exciting time,
9:51What is getting you sort of --
what makes you most hopeful?
9:54What are you most excited about
when you get up every day
9:58to go figure out how
we're going to find solutions?
10:01ARM: So let me say this.
10:04We believe that we are in a space
10:06where the consequential decade
10:10meets the decisive decade.
10:13And so if you've heard that term before,
the consequential decade,
10:16it's what AI practitioners
use to talk about,
10:20this is the time in which we have to think
10:22about ethics, policy,
10:27regulation, technology,
10:29innovation, invention,
10:31because these are the decisions
that are going to decide,
10:35these are the things
that are going to decide
10:37what impact AI has
on the global community.
10:41And we all know here
what the decisive decade reference is.
10:45And so this is a place
10:47where the consequential decade
meets the decisive decade.
10:50And so it really has to be
all hands on deck.
10:54And a commitment
from communities in the AI space
10:58and communities in the climate
and nature space.
11:01And the Bezos Earth Fund,
11:02we see ourselves
as sitting right in the middle
11:05and being a leader in that space.
11:08MZ: OK, we'll have to leave it there.
11:10Amen Ra Mashariki, thank you so much.
