November 26, 2021 • 48 mins
New Episode on Spreaker: AI, Climate Change, and Innovation with Peetak Mitra In this episode of Data Stand-Up!, we speak with Peetak Mitra, Scientist at PARC and Core Team member at Climate Change AI. Peetak is a Member of Research Staff/Scientist-II at the renowned Palo Alto Research Center (formerly Xerox PARC), located in the heart of Silicon Valley. His work at PARC focuses on developing cutting-edge AI and machine learning algorithms, with impactful applications across various fields, including:
- Distributed sensor systems (DARPA Ocean-of-Things)
- Predictive maintenance using IoT (Novity)
- Climate modeling, policy, and geoengineering (MCB)
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:06):
Bedrop is an independent data science and aiphirm specialized in
data driven business change. In this podcast, our guests help
us spread knowledge and experience with our listeners.
Speaker 2 (00:28):
Good Morning, Peak, How are you doing today?
Speaker 3 (00:31):
I'm being good, good afternoon. Has very excited to be here.
Speaker 2 (00:36):
No, thank you, thanks for taking the time. We always
started with the question of where are you, and I
think today is highly relevant because you are in the
place where many would like to be because of what
everything is happening around there. Could you tell the audience
where you are right now?
Speaker 3 (00:56):
Sure? So, I'm in the San Francisco be area, Silicon Valley,
to be exact, And you're absolutely right. This place is
fascinating in terms of the technology that it's building, the
kind of school work that's ongoing here, as well as
the really smart people that live here. So I totally
totally agree with you on that front.
Speaker 2 (01:16):
Well, the thing is, after the boom of Silicon Valley,
I feel that there's been many ecosystems growing on different
sides of the world. I'm trying to learn from this
experience and everything that happened in Silicon Valley. I think
it's been a reference for a few years and a
(01:37):
lot of amazing companies were born in this area. So
for junk professional I think it's a very cool and
place to be saying absolutely, Okay, I know we serve
a similar background, but before we get into that, I
like to know how your agenda looks like and what
(01:58):
you have ahead for the day, like we usually do
in these kind of calls.
Speaker 3 (02:03):
Sure, so typically every day at work is different, and
that's the fascinating bit of it. Typically it involves a
few different meetings sometimes we eat too many for my liking,
but also has some dedicated window of time that I
write hands on code with for my projects. In the
last few months, I have been focusing a lot on
(02:24):
writing proposals, very fascinating proposals, I might add for external
funding agencies such as DARPA and NASA, as well as
internal funding from my employer at Park. So every day,
you know, as I said, looks different, depends on the priorities.
Definitely some meetings and time to work.
Speaker 2 (02:42):
Okay, And as I were saying, we do share some
background as we well, you complete the PhD mechanical engineering.
I only go to the bachelor and then a transition
to our space. But it's curious that we both indeed
doing they that relates stuff on different fronts and on
(03:03):
different kind of use cases. If you want to put
it like that, How did you realize you wanted to
be immersed in the AI or data science field?
Speaker 3 (03:13):
Yeah, and yes, I remember having a chat with you
about your prelevant background in this area. I can't say
you missed much not having gotten a PhD, so jokes
apart as as you said, I did receive a PhD
in mechanical engineering from the University of Massachusetts AAST. My
research focus during this time was on developing radio startup
(03:35):
models for engineering systems relevant to Internet combustion engines. In
plain English, this means I worked on developing mathematical models
that are cheap to use but provide a pretticod approximation
of the system state and its evolution over time. Typically,
the industry in academia uses these models to perform design
space exploration as well as optimization. In the final two
(03:57):
years of my PhD, I saw an uptick in the
use of AI for engineering modeling. University of Massachusetts, as
you may know, is one of the best schools for
learning AI in the US. In fact, the entire field
of reinforcement learning was first proposed here by a doctoral student,
so I bit the bullet and took some courses from
the Computer science department. They were extremely challenging, as you
(04:18):
can imagine, but very rewarding and I learned a lot.
Following this, I had no funding to pursue this data
science avenue in my research. My funding came from elsewhere,
so I convinced my advisor to use his contacts in
the industry to start a data science consortium that would
focus on how AI can be used to accelerate engineering,
design and research in the area of internal combustion engines.
(04:41):
We raised an impressive sum of money and in the
end had pretty amazing partners such as NVDR, MathWorks, Siemens
comments join Us that helped me focus on using machine
learning for modeling onlinear complex systems such as fluid turbulence,
I field that you probably are aware of, and combustion
in the last few years of my PhD. During these
(05:02):
last three years, I also spent some time at the
famous Wassalamus National Laboratory, which is really well known for
a lot of things, including the Manhattan Project during World
War youth, and I spent time there during twenty nineteen
and early twenty twenty, and I learned a great deal
in this area. Lenel is one of the thought leaders
in the space of data driven fluid modeling and it's
(05:25):
definitely one of the most exciting places to do science
in the world. So, yeah, it was a bit of
circuitous route, but it worked out.
Speaker 2 (05:31):
I guess indeed you managed to do pretty well. I'm curious,
so you did or you started with those AIUS cases
focused on on the engineering beats, whether related to fuid
dynamics for different beats and pieces. But how did this
(05:52):
academic background, academic background in terms of, you know, the
engineering part and the experience using data and and modeling
led you to being a part of the core team
in Climate Change AI. Because for those that are listening,
you do come from an engineering background, but now, apart
(06:13):
from the work that you do at your current company,
you're also part of the core team of Climate Change AI.
So before discussing what climate chained AI, how did you
get into that and how did you think that was
relevant and a good step to take in your career.
Speaker 3 (06:32):
That's that's a great question, and that's a very important
question because I was definitely not in an area where
I was looking at climate very seriously. I was looking
at improving interm compression engines, but which had an indirect
impact on the climates as a result, but definitely not
looking at climate for you directly. So this actually happened
(06:54):
while I was at Las Alamos and attending a conference
called EUROPS, which is considered by many to be the
premier machine learning conference in the world, it's debatable. During
this time, I was organizing an informal get together for
friends and U for friends and people I knew with
similar interest AI and engineering as well as climate modeling
(07:14):
from Microsoft, Google, NVDA and academia, and as it turns out,
somebody from that group recommended I reached out to this
PhD student from Carnegie Mellon called Priadnd. Obviously she could
not make it, but we stayed in touch and I
attended a workshop she co organized studing the conference called
Tackling Climate Change with Machine Learning, and I saw the
(07:34):
saw the really large interest from the community, uh for
that for that particular event. In fact, I had to
wait in line for over an hour to get into
the room where the talks and posters were, so you know,
that was right. I was. I was so keen to
be a part of part of that movement. So finally
got to meet meet Priya during during the conference briefly
and we had a very nice chat and we exchanged emails.
(07:56):
But I, you know, subsequently moved on and forgot about it,
moved on with my own research. Then we got in
touch again in early twenty twenty and CCI asked if
I would like to co organize a part of their
ICLR workshop. So this is a workshop they were organizing
as part of a conference called ICLR, which I was
very thrilled about. I co organized it with doctor Kelly Koccenski,
(08:18):
who at the time was a PhD student at the
University of Colorado and currently is a scientist at McKinzie
doing some very impressive work. So basically, one thing led
to another and an opportunity opened up in their co
team for me. Subsequently, I interviewed with the team a
few different people, and in a few months I was
onboarded and it has been a great journey since then. CCI,
(08:40):
for those of you know, is an incredible organization with
a very fascinating and powerful mission at its core, and
I'm proud to be a part of it, and you know,
obviously doing a little bit in contributing to the.
Speaker 2 (08:52):
Overall gol No. I think it's exciting. Actually, the first
time I did heard about well, you call it c
c A I, which stands for climate change AI. Obviously
it was when we had when we had being Spatachia
here he briefly mentioned the work that this this organization does,
(09:18):
or your organization do. The thing is m we were
supposed to have Priya. And then I came across you,
and I think your profile was very interesting because you
also mixed the world that you do at Park. And
I think that's very relevant to the conversation because you
do very interesting thing things. And while you contribute to
(09:39):
the provano initiative of c c A, I think being
a scientist at Park comes with many interesting perks. Right.
I've read very briefly about an odd trait which stands
for otion of things. Of course, it was very similar
to Internet of Things. I was thinking of using sensors
(10:01):
and different types of IoT mechanisms to collect data in
or from the ocean. Right, And and you name this
initiative DARPA, which is great from the Department of Defense.
So I guess that you collect, as I said, some
some kind of environmental data moving across the ocean using
(10:24):
iote technology. But how do you use these data sets?
And I know, I know, because you've told me before
that this Brett is confidential. But anything that you could
say that's interesting to to anyone listening to this conversation,
to set an example for reference of work that you
do at part, that'd be very interesting.
Speaker 3 (10:43):
Indeed. Indeed, so yes, as you rightly said, the Ocean
of Things project is a play on the Internet of
Things in terms that that the industry uses. So so indeed,
this project is funded by an agency called DARPA for
the benefit of your audience, DARKAST and for the Defense
Advanced Research Projects Agency. It is a research and development
(11:05):
agency of the United States Department of Defense responsible for
the development of emerging technologies for use by the US military.
Some of the notable successes from dart PA program have
yielded technologies that we really take for granted today, including
the Internet that we are using to chat, the GPS
that we use very much daily, language translation, automated a
(11:25):
vice recognition, so and so forth. As you mentioned this,
since this project has sparked proprietary technology that we will
commercialize one day hopefully and is of strategic importance to
the US. I cannot go into the details of the
hardware platforms, sensors, software algorithms and share details about where
this is deployed in the field, but I can definitely
(11:46):
share some high level objectives that are already in the
public domain that you know, hopefully some of your users
can actually look at and be inspired by. So the
Dart Position of Things program, you know, at the high level,
seeks to enable persistent Maritimes situational awareness over large oceans
by deploying thousands of low cost, environmental friendly intelligent floats
(12:06):
that drift in a distributed sensor network. So the idea
is that oceans cover seventy percent of the Earth service,
but we know very very little about them. The grand
idea from this program is to use the data generated
via this program to support the Department of Defense missions
obviously that are of strategic importance to the US, as
(12:27):
well as public oceanographic research and commercial applications. So basically,
this modern platform is built to support a dynamic and
more efficient ocean management tool and as an improvement over
terrestrial based management techniques that lack real time data about
changes in the ocean environment. This is very important because
(12:48):
a large effect, to a large degree, the effect of
climate change and a warming planet is actually seen in
the oceans. Oceans absorbed disproportionately higher amounts of carbon and
the heat that that we have as a result of
global warming because of which it is it is. It
is a fast changing environment and we need to understand
it from the perspective of understanding how oceans are changing
(13:12):
and how that might impact whether it's feather systems down
the road for for us in especially in the warming planet.
So some of the benefits of this Ocean of Things
program is that it passively collects all of these information
such that there's no supervision needed. And for example, one
idea is in benefiting marine animals in using the acoustic
(13:34):
data collected at a very very fine spatio temporal scale,
so in at at a quick interval UH and that
is able to detect, identifying track marine animals and quantify
the background noise scapes. This helps in their preservation, This
helps in ensuring that these animals have a very habitat
(13:56):
that they can thrive in. This also this information, you know,
thely can be used to improve understanding of even marine
animal behavior in regional reasons such as such as the
Atlantic or the specific. The program also helps to develop
algorithms to automatically detect, track, and identify nearby vessels, including
(14:16):
new indicators of anomalous marietime maritime activities such as illegal
fishing behaviors. If somebody is interested, they can definitely find
more information about this program from the darka website. It's
called Ocean of Things dot dark dot dot mill Male
stands of the military part. So yeah, I mean so
it's a it's a pretty fascinating program, a remarkable feed
(14:39):
of engineering for sure.
Speaker 1 (14:40):
No.
Speaker 2 (14:41):
Actually, I'm very interested in learning, learning more. I'm very
passionate about everything in regards to the sea. I love
Ian Yes, snor Clean, I love that, and I really
understand that. Well, I know where you come from when
you say, well, the ocean can be that weakness that
(15:01):
tells us how climate change and global warming is affecting everything. Yes,
because you can pay attention to how mammals are moving
across different oceans, maybe paying attention to some of the
metrics which relate to the environmental state you mentioned carbon,
But probably you are also paying attention to others. So
(15:23):
I think it's a very clever initiative. So so thanks
for that, because I guess that goes for all of
the society, or it goes for the benefit of everyone
listening to this conversation. And can you name or explain
a few other initiatives that you know are relevant for
the conversation here, which may relate to climate change that
(15:45):
you are launching, living or somehow participating in under the
umbrella of park, I would.
Speaker 3 (15:53):
Love to talk about some initiatives currently in the very
park in this in the space of clean technology, I
would like to talk about a project that focuses on
emissions monitoring. As you know, methane is one of the
most potent greenhouse gases and responsible for a third of
current bombing from human activities, and there has been a
growing focus on methane on a curbing methane as a
(16:16):
way of buying extra time to tackle vanta change. Although,
for example, there there is more cootwo in the atmosphere
and it checks around for longer than a methane molecule,
but individual methane molecules have a more powerful bombing effect
on the atmosphere than single COTWO molecules. I'm sure most
of your audience have recently heard about the pledge to
(16:38):
cut methane emissions by thirty percent by twenty twenty thirty
at the most recent COP twenty six conference in Glasgow.
At Park this is this is an area of focus
for our clean tech portfolio and our research teams are
actively developing domain in fumed algorithms for institute and remote
sensing data assimilation for detecting methane leakage. So that is
(17:01):
one aspect of the work done at Park. Another project
that the researchers at Park have done a fabulous job
is in reimagining the HBAG system. It is again part
of our Gleentic portfolio and the idea here is that
in the challenge as well, here is that in a
warming worlds, as you mean global temperature rises and as
people become more well off, the demand for cooling will
(17:23):
only increase substantially. Now, if you look at HVAC systems
are a major cause of greenhouse gas emissions. For example,
ninety percent of US households use air conditioning. That accounts
for six percent of the country's residual energy use. In
terms of numbers, that translates to about one hundred million
tons of carbon dioxide every year. That's a lot, and
these statistics are probably getting updated every single day. Most
(17:47):
HBAG systems use harmful hydrofluoro carbon refergence. Although these are
very short lived compounds, they are thousand times more potent
and siatio in terms of absorbing heat. Without going into
the technical details of the proprietary technologies developed at Park,
researchers here have developed next generation systems that are refrigive
and free and need eighty percent less energy to operate
(18:09):
compared to a standard h BAG system. This is truly revolutionary,
honestly in terms of the impact this will have on
the human experience, human comfort, as well as in the
fight against climate change. Apart from building these novel technologies,
a lot of teams at Park, including MIND, have been
working on developing climate models that improve our understanding of
the Earth and planetary systems. One of the projects I'm
(18:31):
involved in uses AI to discover relationships between the cloud
reflectivity or albedo as the community calls it, and the
atmospheric circulations, which is very poorly understood. The technological impact
of this would be able to identify potential sites per
radiative forcing, which means that you could artificially bride the
(18:53):
marine clouds and to characterize its impact on the global
energy and precipitation cycles. There are many of the projects
at Park, including carbon capture and so many other cool
stuff that researchers are working on that are focused on
counteracting the negative effects of climate change, and honestly, we
would need a full day to discuss all of them
(19:13):
in detail. But hopefully this gives a sense to your
listeners on kind of work we do here at Park.
Speaker 2 (19:21):
No, it actually does. I'm just curious how have you
structured the team in terms of well, you surely have
data scientists, you surely have data engineers, but how did
you build that structure? Because it's helpful for me to know,
you know, I'm leaving bedrock here, and I don't think
(19:46):
there is some exact answer to how you structure the
science team around professional services, right, because you do have
people that are client facing, and then you have people
that play the client facing role that also have to
do innovation throughout the week, right and the decade sometime too,
are in the some cues to know how have you
(20:07):
build these these corporate structure at park to the extent
that you can share sure.
Speaker 3 (20:13):
Indeed. So you know, park is has been around for
fifty years and as I mentioned at the start, it
has been you know, at the at the core of
developing a lot of important technological breakthroughs such as the
Internet or the traffic user interface, the personal computing things
like that. So it has had a history of people
(20:35):
working with different people with different skills, working on very
interdisciplinary problems and bringing solutions to market. So so right now,
climate is a major area of focused within park where
we are looking at different avenues and trying to attack
it from different different different sites. For example, we have
something called us the Clean Tech Tower, which looks at
(20:58):
technologies that can be comingalized in the next few years
within the clean technology realm uh and some of these
projects are within within that scope. So and we do
have folks in the in the in the company that
are dedicated from the business side that that aim to
uh commercialize certain ventures uh some some some some folks
(21:19):
at me for example, that are more from the technical side.
We id eate and we we actually execute on these
ideas and we have a team of as you said,
data scientists, software engineers, domain experts among others to kind
of contribute to this multi disciplinary effort. So so it's
it's a pretty interesting dynamic something I have not seen
(21:39):
anywhere replicated anywhere else, to be honest, But but yeah,
I mean it definitely works.
Speaker 2 (21:47):
Yeah. Well, well it's really interesting also to me and
to us as Petruck. I think we should find some
time to talk because you have really interesting initiatives here
and there be so much much work to be done
also here in Spain. I think this is very good
learnings that you that you've made. Okay, now living park
(22:11):
a side and also trying to dedicate some time to
climate change. He during this conversation. Actually I learned as
I said about c c A when I spoke to Beings,
But after that I did out of reading around the
cases that you that you build around some very interesting
(22:32):
names that you have in the advisory board. I saw
Joshua Benjo and you n G. I mean, those those
are very popular names that you know provide an overview
of the They mentioned that this initiative has taken probably
over the previous months and years. I don't know how
did this initiative start from prea If you have that information.
(22:55):
I guess it's global already. Yeah, well yes, I mean
it little bit about the beginning of this and how
I grew up so quickly.
Speaker 3 (23:06):
Absolutely, I would love to so I quickly introduced climate
Change AI for the benefit of your audience who are
not aware of the group or have never heard of
have never heard of us. Climate Change AI is a
globally initiative to catalyze important, impactful work at the intersection
of climate change and machine learning. That's a main mission.
I'll come back to how we do it in a
(23:28):
bit and provide some examples for your audience, But now
I'll kind of try to answer the first question about
how this began. So the idea to build this organization
started from an informal discussion on how AI can tackle
climate change that was led by current CCI coachs David Ralik,
Pria Donti, and Linn Cock, alongside Europe's twenty eighteen with
(23:49):
that was organized in Montreal. The founding members realized that
if they are the large amount large number of people
who attended an informal session that the coach is organized
are so interested in working out how they could make
an impact, then probably the best way to do this
was to externalize this effort and put a list of
(24:09):
recommendations for how people could use their skills to create
impact for this challenging problem. CCI officially launched just after
several founding members co authored the foundational paper called Tackling
Climate Change with Machine Learning. I would strongly encourage your
audience to take a look at that. This paper provides
a detailed overview of how machine learning can be used
(24:30):
impactfully for climate change mitigation and adaptation across different sectors.
Since its early days, CCI has grown to become a
global community as you said, and we have representations from
all parts of the globe. The main goals of CCI
are in building a community. For example, CCI heads emal researchers,
climate researchers, climate related businesses for example, as well as
(24:52):
other stakeholders connect with each other via regular workshops you organize.
These workshops are very well attended and typically have many
thousand footfalls or digital footballs as we like to call
them these days. As well as our social media that
we have recently launched it's called the CCI Community Hub,
(25:14):
which is going very fast to reduce the barriers for participation.
We also provide funding for students to attend our workshops
when we used to have them physically or even now
the digital workshops, we have made them free for people,
and we also organize mentorship programs for folks interested in
breaking into this space to help individuals who are interested
in person climate research or do their bit for climate
(25:35):
change medication and adaptation. We also provide online resources for
them to learn more about the intersection between their fields
via programs like Summer School. The registration service which are
actually open right now, so I'd encourage your audience to
take a look at that if they're interested, peace feel
free to sign up, as well as regular webinars which
features leaders from the industry, academia, and national laboratories. In
(25:58):
addition to that, we release monthly newsletter that actually went
up today the day of the recording for this podcast
that currently is delivered to over eight thousand people worldwide
and it's growing every every month and it contains a
highly curated list of opportunities including jobs, events, calls for
funding in this area of climate change and ai CCI
(26:20):
also guides research and the area of climate change and
ai pro facilitating grants programs. Our most recent grant program
involved a total award money part of one point eight
million dollars that was that will be awarded to a
dozen or so projects starting in twenty twenty two. This
was a pretty well received program by the community. We
(26:43):
received a lot of entries for this grand program called Finally,
CCI also provides formal and informal advice to researchers, companies,
policymakers from around the globe. Most recently, CCI coachair LINKAK
provided feedback on eus AAR regulations to include impact and
climate that were discussed actually at the European Parliament in
(27:04):
Europe last week. You say, also participated in a panel
on Governance in AI and climate change within German pavilion
at TOP twenty six conference in Glasgow. So again, this
is only a small fraction of all that CCI is doing.
It's growing really fast thanks to the efforts of my
colleagues who are not here, who are not here in
present today. For anyone in the audience that is interested
(27:26):
to learn more about us, please check out our website
climate Change that AI and sign up for a monthly
newsletter to keep updated about our activities.
Speaker 2 (27:35):
Yes, no, I'm actually recommending that as well, because when
I took a look everything that you were doing. Some
of the names that you had in there, I mean
people from stam for the mind and everything. It's surely
we'll cut a lot of people's attention. I didn't know
that you also participated in the new EU AI legal
(27:59):
framework is coming and I think is pretty much needed
on different fronts. I guess that your area of specialization
relates to environmental AI, so climate change AI, but there
are different fronts that you could also contribute as well.
I mean, if you pay attention to the names that
(28:22):
you have in the team, your scope of advisory could
go well beyond climate change. I guess it's relevant because
it's climate change AI, but I'm sure that they would
appreciate that help across other domains as well. Okay, I
did read that two of the founders of CCI were
(28:44):
named among mi T Technology Reviews words. I think it's
something like thirty five innovators and they're thirty five, well
thirty howl these two because I do not remember for
their names, so probably you can you can mention they
were recognized for doing research on different topics around optimizing
(29:10):
power greeds, even monitoring biodiversity or something like. You mentioned.
Also what you do at park. Maybe it's a relevant here,
so it's part of the learnings that you can use
in this space. But can you share more insights on
how they did this? I mean, I've mentioned research on
optimizing power greeds and monitoring biodiversity, but I'm curious to
(29:33):
know the data collection technology, data infrastructure mL algorithms that
you've used, any insights that you can share on these
research pieces that led you, I mean, these co founders
of CCA to get in this list.
Speaker 3 (29:51):
Absolutely so we were super excited when the scheme up.
You're very proud of the work and impact made by
to CCI coaches. The TUCCI coaches that you were talking
about are Priya Donti and David Rolneck that were named
in the Technology Reviews thirty five and A thirty five list,
as well as Lin Khak, who is our third CC
(30:13):
culture and when we're very proud of their contributions, active
contributions to the scientific community. They publish, you know, regularly
as well as they do this. This this really stellar
work by CCI will not While I'm not entirely privy
to the work done by them or what the reviewers
that Tech Review were looking at, I know and David
(30:36):
have both looked extensively in the constrained learning space, especially
for problems involving optimization that that have hard constraints, such
as in the energy and the power sector. So I
would defer you to their papers that are publicly available
because I'm not fully familiar with with what they exactly used.
(30:57):
But you know, they they did a lot of work
in the area of constrain learning, physical conscreen learning, or
with heart constraints. Uh. And they have an excellent paper
that came out recently called DC three uh And And
you know, so, so if if anybody is interested in
the audience, the papers are available publicly. Yeah, we're curious.
Speaker 2 (31:18):
Yes, I will take a look at them. Apart from
what we've mentioned, and well, you've already mentioned a few
or we've mentioned a few breakthroughs as a global movement,
but have we missed any that you think will be
(31:39):
impacting our society in the short medium term that the
audience would be aware of.
Speaker 3 (31:45):
Sure.
Speaker 4 (31:47):
So you know, first I kind of talk about some
other interesting things that you mentioned about, like how c
C I can can you leverage some of our advisory
boats to kind of make a bigger impact. So since
you talked about US being a global movement and the
achievements of CCI as a global movement. Let me talk
(32:08):
about the recently released recommendation on Climate Change and AI.
So these recommendations are for government action and this was
released along the lines of COP twenty six conference. This
new report, developed by CCAI and Center for AI and
Climate for the Global Partnership on AI, offers forty eight
specific policy recommendations for governments on how to support work
(32:30):
at the intersection of AI and climate change and how
to better align the use of AI overall with climate goals.
Some of the use cases highlighted.
Speaker 3 (32:38):
In the report include the National Grid ESO, which has
used AI to double the accuracy of its forecast for
US electricity UK elliptricity demand radically improving the forecast of
electricity demand and renewable energy generation is definitely a critical
aspect in enabling greater proportions of renewable energy on electric
(32:58):
grids because you have been lable energy all the time
in the day. So if you can able to if
you're able to forecast the demand, you can you can
plan for that. The UN Satellite Center has developed for
these example is the UN Satellite Center has developed this
flood AI system that delivers a high frequency flat report
that is actually used UH and has improved the disaster response,
(33:20):
the climate related disaster response in Asia and Africa. The
third example that that's this report talks about UH is climatories.
It is a global coalition of organizations that is actually
working to radically improve the transparency and accuracy of emissions
monitoring by leveraging AI algorithms and data from more than
(33:43):
three hundred satellites and eleven thousand sensors. So so these
are some of the like you know, other impacts that
Climate change AI has made in the space, especially making
these specific policy recommendations to governments and people who have
the power to impact impact all of our lights on
a very large scale.
Speaker 2 (34:05):
I was standing a fair yes it was like a
fair yesterday imagrid mainly related to renewal energies. There were
many stands about solar energy on different fronts. Some of
the engineering companies doing or advancing their work around the
(34:28):
movement of the panels to ensure that you get the
maximum energy from the sun. I mean different initiatives, many
interesting ones, just because as you as you already know,
the the solar field is a world in itself, there
is so much to be learned. I mean, data and
analytics for EI is a big thing with many different
(34:51):
sup fields, but solar is another one. So I'm curious
to know if there is anything that you've done in
regards to solar paneling, specifically from climate chain to AI
that we could highlight to the audience.
Speaker 3 (35:08):
Indeed, so, as as I mentioned earlier, climacy c I
is a group of very highly multi disciplinary people who
actively contribute to the scientific community as well as we
organize these workshops alongside major machine learning conferences that sees
a that sees an impressive list of participants, that that
(35:28):
that share their work with us, has a very regreous
review process to ensure quality. Uh and I would highly
you know, recommend it to your audience to kind of
come check out some of these papers that are archived
on our website. So so for example, you talked about
solar panels and and and all these important things. So
I'll give you an example of a paper that was
(35:50):
presented in our in our workshop in the past that
use machine learning to deduct solar installations globally was actually
published in Nature, you know, the one of the world's
most prestigious journals recently. Well, I'm not trying to take
credit for their work or suggest that our review process
helped them, but it is just commenting on the high
quality impact people are making using our resources. So so so,
(36:15):
there are so many other papers where or other projects
that you will probably see featured in our workshops that
actually use machine learning to identify optimal sites for installation
of solar panels, for example, lars scale solar farms that
that can generate reliable electricity for for for the community.
(36:39):
So these are some of the high back areas as
well as there is a lot of interest and work
by the community that is looking at developing better materials
for you know, improving the efficiency of the solar panels
that is actually improving by the day. As well as
how you can use machine learning to detect fault in
(37:00):
solar panels, for example, if there's a crack, or if
there is certain anomalies due to manufacturing or other things,
how can you use machine learning to detect all of
that to ensure your production you know, doesn't suffer because
of certain certain effects that that you know happen due
to certain way andeer that happens during the use of
(37:22):
an engineering platform. So there are multiple ways that machine
learning is looking into that, and a lot of these
are actually reflected in our workshops, you know, community members
kind of present their work in our workshops and you
can see all of these papers on on our website.
They are archived for the benefit of of of the
(37:44):
audience such as yourself.
Speaker 2 (37:46):
I mean we we are already doing some work in
these field. So I think, as I said before, we
need to set some dim side to learn about the
advancements that you've made in this field. Yes, because some
of the players in the smallar industry, well they are
busy in their day today, right, and not everyone in
(38:07):
these companies have dedicated enough time to pay attention to
how mL could help them around different meds. They've looked
into predictive maintenance, but sometimes they haven't looked into panels efficiency,
using advanced analytics and things like that. So at some
point they will need to be supported by companies like
(38:31):
ours as at truck. So learning and exchanging knowledge with
an organization like yours as CCA, I think would be
would be an amazing starting point. Okay, so this is
the first probono initiative that I see around the world
(38:53):
in the intersection, as you said, of machine learning or
AI and climate change. Is there any other initiative, whether
national or global, that you know of that is also
helping you in this mission and that also serves the
same vision.
Speaker 3 (39:14):
Indeed, you know, there are so many other groups that
we that that that we follow data that are doing
excellent work in this area as well. For example, Climate
Informatics is a group that has been around for about
a decade or so, started by Professor Klaremant Montellioni, who
is actually one of the advice who is one of
(39:36):
the members of the CCS Advisory Board, and other people.
And the goal of climate informatics is is and not
to describe it in my own works, but the goal
of climate informatics is to apply statistics, machine learning and
data science to solve problems relevant to climate science, so
the science of climate, including climate modeling and how the
(39:57):
climate climate changes, so on, so forth. U. There is
also another group that aims to use quantum computing for
climate change. You know, obviously quantum computers, you know, they
are almost on the horizon. They would be extremely fast,
so they can actually help in help in detecting better
or or or understanding what materials could actually help improve
(40:20):
solar efficiencies for example, for these solar service or battery
research or accibertate, some of these areas that are extremely
critical for us to transition into a zero net, zero
carbon world. So that is super interesting. The quantum computing
for climate change, I you know, and and and honestly,
what if the tools the community uses it is It
(40:41):
is actually super exciting to see a lot of creative
folks working on these innovative solutions out there. So yeah,
there are certain these are certain groups that come to
my mind. There are obviously a lot more, and I
apologize for missing.
Speaker 2 (40:54):
Out on them. You probably know if you've heard or
re listened to the previous podcast that that we've made
with with well many interesting guests. But now that we
come to the to the end of this core I
always ask for someone to join us in this data
stand up podcast. And you've mentioned a few initiatives and
(41:17):
probably you have some other professionals in your radar that
you think would be you know, with guests, So I
don't know, could you give me a couple of names
and then make an introduction for us to have him here.
Speaker 3 (41:32):
Sure, indeed, So I think I think you know, because
I'm from the industry in the US and in the
b Area, I'm probably a little bit biased towards, you know,
the kind of impactful technologies that that these companies are building,
especially because the climate problem is so big, we need
we need technologies that can be developed fast and at scale. So,
(41:54):
for example, Albert recommend you know some some folks like
Steve can grasp from Climate AI. Climate e I is
separate from Climate change. AI is actually a startup in
the Bay Area that is using AI for precision agriculture,
so agriculture in a changing climate, because you know, the
predictions are that as as the global warming increases, there
(42:19):
will be more more and more crop failures and stuff
like that. So so Climate I is actually using AI
to help farmers get a better yield depending on local
conditions and so on and so forth. So so Stephen
rasp would be an excellent candidate to talk about his work.
Stephen Heyer, So he's from Google AI. They are working
(42:39):
on He and his his team has worked on near
term precipitation. So how can you know how can AI
and generative models use UH be used for for predicting
precipitation in a very short time window. And and they
have done excellent work in that space. Shak Mohammu from
(43:00):
a deep Mind. Once again it's a Google company. Now, uh,
they have done some excellent work in the same area.
I think they have built even better model than Google
that does the snow casting work. And and Shakira is
super super active in this area of of you know,
climate change machine learning as has has has been a
(43:24):
part of our previous workshops as well. So yeah, these
these names are its strongly recommend for you to to
look into happy to make introductions as as you see
for it.
Speaker 2 (43:34):
Yeah, well that's amazing. I mean I usually get one
or two names, but there, that's really good. Thanks for that.
And yes, before we close the call of I also
ask for you know, a book, a newsletter or some
reading material that you highly recommend. And it doesn't have
to be specifically around climate change and AI or even
(43:56):
around data. It can be around anything because in the
past they had really good segestion so on. I don't
know how you structure your day to day to be
more effective and productive things like that, So anything that
you can think of, yes, yes, one piece of reading
material that you recommend.
Speaker 3 (44:16):
Indeed, so you know, well you can call me a biased,
but i'd highly recommend your audience to sign up for
our c c AI newsletter if you haven't already jokes
about since you said newsletter as well in your question, uh,
newsletters from organizations like Climate Piece, Climate AI, which is
(44:37):
separate from c c A. I batch some of the
pretty interesting news letters in this area. About books, you know,
there are a lot of interesting books I've I've read,
but I think one book that I'm currently reading, you know,
catches my attention and I think it's it's an important
book for all of us to read and understand. Is
It's a book called Speed and Scale by on Tour.
(45:01):
It is a book about series of action plans for
solving the climate problem right now using technologies that exist today,
are solutions that can be developed fast and scaled up
to massive magnitude of climate change. I think I would
highly recommend it for everybody who has you know, has
has has an interest in solving this problem, not just
(45:22):
standing by the wayside, but actually contributing to any of
these solutions. And once again, I would highly recommend your
audience to come join us at t C I uh
t C I is an actually open community and very
welcoming community. So we would love to have some of
your audience members join us in this, in this in
this great emission of making making making an impact in this,
(45:46):
in this big fight against climate change.
Speaker 2 (45:49):
Not for sure, Well, I'm going to tell my colleagues
here and I think again jokes aside what we've discussed
during this call, both around sett in some time aside
to talk about different learnings and changing knowledge on different
pieces of work. I think bedrocs should be in some
(46:10):
sort of way supporting these initiatives. And I don't know
if you've previously established any connections with any Spanish companies
that operates in the data science in aifield, but we
would love to be a part of this. So I'm
saying this now while we are alive, but I think
it's something that we should look into because the work
that you do is highly relevant to all of us
(46:31):
as a society, and I think that this can be
articulated in different fronts. So the same way you are
supporting the EU for those legal frameworks, there is some
hands on work that requires time and if we can
put some resources to use and learn together with you. Well,
(46:52):
that that partnership would be would make us proud for sure. Anyhow,
bit that really thanks for taking time. We and I appreciate, uh,
the time that you've taken in in in thinking about this,
this this podcast, and the time that you've dedicated today
taken my call. I think you have a bright future
(47:14):
ahead of you, well yourself as Speedup, but also as
as part of park in climate change. I so congratulations
to that and and good luck to you.
Speaker 3 (47:24):
Well yes, yeah, thanks, thank you for the very kind words.
You know, we would you know, definitely love to chat
with you offline. It was a pleasure to be on
this podcast and I cannot wait to see the great
things you and your colleagues will do in the future.
Thank you very much
Speaker 2 (47:40):
As this Thank you half a good one Pickers Productststructus
Bedrop is an independent data science and aiphirm specialized in
data driven business change. In this podcast, our guests help
us spread knowledge and experience with our listeners.
Speaker 2 (00:28):
Good Morning, Peak, How are you doing today?
Speaker 3 (00:31):
I'm being good, good afternoon. Has very excited to be here.
Speaker 2 (00:36):
No, thank you, thanks for taking the time. We always
started with the question of where are you, and I
think today is highly relevant because you are in the
place where many would like to be because of what
everything is happening around there. Could you tell the audience
where you are right now?
Speaker 3 (00:56):
Sure? So, I'm in the San Francisco be area, Silicon Valley,
to be exact, And you're absolutely right. This place is
fascinating in terms of the technology that it's building, the
kind of school work that's ongoing here, as well as
the really smart people that live here. So I totally
totally agree with you on that front.
Speaker 2 (01:16):
Well, the thing is, after the boom of Silicon Valley,
I feel that there's been many ecosystems growing on different
sides of the world. I'm trying to learn from this
experience and everything that happened in Silicon Valley. I think
it's been a reference for a few years and a
(01:37):
lot of amazing companies were born in this area. So
for junk professional I think it's a very cool and
place to be saying absolutely, Okay, I know we serve
a similar background, but before we get into that, I
like to know how your agenda looks like and what
(01:58):
you have ahead for the day, like we usually do
in these kind of calls.
Speaker 3 (02:03):
Sure, so typically every day at work is different, and
that's the fascinating bit of it. Typically it involves a
few different meetings sometimes we eat too many for my liking,
but also has some dedicated window of time that I
write hands on code with for my projects. In the
last few months, I have been focusing a lot on
(02:24):
writing proposals, very fascinating proposals, I might add for external
funding agencies such as DARPA and NASA, as well as
internal funding from my employer at Park. So every day,
you know, as I said, looks different, depends on the priorities.
Definitely some meetings and time to work.
Speaker 2 (02:42):
Okay, And as I were saying, we do share some
background as we well, you complete the PhD mechanical engineering.
I only go to the bachelor and then a transition
to our space. But it's curious that we both indeed
doing they that relates stuff on different fronts and on
(03:03):
different kind of use cases. If you want to put
it like that, How did you realize you wanted to
be immersed in the AI or data science field?
Speaker 3 (03:13):
Yeah, and yes, I remember having a chat with you
about your prelevant background in this area. I can't say
you missed much not having gotten a PhD, so jokes
apart as as you said, I did receive a PhD
in mechanical engineering from the University of Massachusetts AAST. My
research focus during this time was on developing radio startup
(03:35):
models for engineering systems relevant to Internet combustion engines. In
plain English, this means I worked on developing mathematical models
that are cheap to use but provide a pretticod approximation
of the system state and its evolution over time. Typically,
the industry in academia uses these models to perform design
space exploration as well as optimization. In the final two
(03:57):
years of my PhD, I saw an uptick in the
use of AI for engineering modeling. University of Massachusetts, as
you may know, is one of the best schools for
learning AI in the US. In fact, the entire field
of reinforcement learning was first proposed here by a doctoral student,
so I bit the bullet and took some courses from
the Computer science department. They were extremely challenging, as you
(04:18):
can imagine, but very rewarding and I learned a lot.
Following this, I had no funding to pursue this data
science avenue in my research. My funding came from elsewhere,
so I convinced my advisor to use his contacts in
the industry to start a data science consortium that would
focus on how AI can be used to accelerate engineering,
design and research in the area of internal combustion engines.
(04:41):
We raised an impressive sum of money and in the
end had pretty amazing partners such as NVDR, MathWorks, Siemens
comments join Us that helped me focus on using machine
learning for modeling onlinear complex systems such as fluid turbulence,
I field that you probably are aware of, and combustion
in the last few years of my PhD. During these
(05:02):
last three years, I also spent some time at the
famous Wassalamus National Laboratory, which is really well known for
a lot of things, including the Manhattan Project during World
War youth, and I spent time there during twenty nineteen
and early twenty twenty, and I learned a great deal
in this area. Lenel is one of the thought leaders
in the space of data driven fluid modeling and it's
(05:25):
definitely one of the most exciting places to do science
in the world. So, yeah, it was a bit of
circuitous route, but it worked out.
Speaker 2 (05:31):
I guess indeed you managed to do pretty well. I'm curious,
so you did or you started with those AIUS cases
focused on on the engineering beats, whether related to fuid
dynamics for different beats and pieces. But how did this
(05:52):
academic background, academic background in terms of, you know, the
engineering part and the experience using data and and modeling
led you to being a part of the core team
in Climate Change AI. Because for those that are listening,
you do come from an engineering background, but now, apart
(06:13):
from the work that you do at your current company,
you're also part of the core team of Climate Change AI.
So before discussing what climate chained AI, how did you
get into that and how did you think that was
relevant and a good step to take in your career.
Speaker 3 (06:32):
That's that's a great question, and that's a very important
question because I was definitely not in an area where
I was looking at climate very seriously. I was looking
at improving interm compression engines, but which had an indirect
impact on the climates as a result, but definitely not
looking at climate for you directly. So this actually happened
(06:54):
while I was at Las Alamos and attending a conference
called EUROPS, which is considered by many to be the
premier machine learning conference in the world, it's debatable. During
this time, I was organizing an informal get together for
friends and U for friends and people I knew with
similar interest AI and engineering as well as climate modeling
(07:14):
from Microsoft, Google, NVDA and academia, and as it turns out,
somebody from that group recommended I reached out to this
PhD student from Carnegie Mellon called Priadnd. Obviously she could
not make it, but we stayed in touch and I
attended a workshop she co organized studing the conference called
Tackling Climate Change with Machine Learning, and I saw the
(07:34):
saw the really large interest from the community, uh for
that for that particular event. In fact, I had to
wait in line for over an hour to get into
the room where the talks and posters were, so you know,
that was right. I was. I was so keen to
be a part of part of that movement. So finally
got to meet meet Priya during during the conference briefly
and we had a very nice chat and we exchanged emails.
(07:56):
But I, you know, subsequently moved on and forgot about it,
moved on with my own research. Then we got in
touch again in early twenty twenty and CCI asked if
I would like to co organize a part of their
ICLR workshop. So this is a workshop they were organizing
as part of a conference called ICLR, which I was
very thrilled about. I co organized it with doctor Kelly Koccenski,
(08:18):
who at the time was a PhD student at the
University of Colorado and currently is a scientist at McKinzie
doing some very impressive work. So basically, one thing led
to another and an opportunity opened up in their co
team for me. Subsequently, I interviewed with the team a
few different people, and in a few months I was
onboarded and it has been a great journey since then. CCI,
(08:40):
for those of you know, is an incredible organization with
a very fascinating and powerful mission at its core, and
I'm proud to be a part of it, and you know,
obviously doing a little bit in contributing to the.
Speaker 2 (08:52):
Overall gol No. I think it's exciting. Actually, the first
time I did heard about well, you call it c
c A I, which stands for climate change AI. Obviously
it was when we had when we had being Spatachia
here he briefly mentioned the work that this this organization does,
(09:18):
or your organization do. The thing is m we were
supposed to have Priya. And then I came across you,
and I think your profile was very interesting because you
also mixed the world that you do at Park. And
I think that's very relevant to the conversation because you
do very interesting thing things. And while you contribute to
(09:39):
the provano initiative of c c A, I think being
a scientist at Park comes with many interesting perks. Right.
I've read very briefly about an odd trait which stands
for otion of things. Of course, it was very similar
to Internet of Things. I was thinking of using sensors
(10:01):
and different types of IoT mechanisms to collect data in
or from the ocean. Right, And and you name this
initiative DARPA, which is great from the Department of Defense.
So I guess that you collect, as I said, some
some kind of environmental data moving across the ocean using
(10:24):
iote technology. But how do you use these data sets?
And I know, I know, because you've told me before
that this Brett is confidential. But anything that you could
say that's interesting to to anyone listening to this conversation,
to set an example for reference of work that you
do at part, that'd be very interesting.
Speaker 3 (10:43):
Indeed. Indeed, so yes, as you rightly said, the Ocean
of Things project is a play on the Internet of
Things in terms that that the industry uses. So so indeed,
this project is funded by an agency called DARPA for
the benefit of your audience, DARKAST and for the Defense
Advanced Research Projects Agency. It is a research and development
(11:05):
agency of the United States Department of Defense responsible for
the development of emerging technologies for use by the US military.
Some of the notable successes from dart PA program have
yielded technologies that we really take for granted today, including
the Internet that we are using to chat, the GPS
that we use very much daily, language translation, automated a
(11:25):
vice recognition, so and so forth. As you mentioned this,
since this project has sparked proprietary technology that we will
commercialize one day hopefully and is of strategic importance to
the US. I cannot go into the details of the
hardware platforms, sensors, software algorithms and share details about where
this is deployed in the field, but I can definitely
(11:46):
share some high level objectives that are already in the
public domain that you know, hopefully some of your users
can actually look at and be inspired by. So the
Dart Position of Things program, you know, at the high level,
seeks to enable persistent Maritimes situational awareness over large oceans
by deploying thousands of low cost, environmental friendly intelligent floats
(12:06):
that drift in a distributed sensor network. So the idea
is that oceans cover seventy percent of the Earth service,
but we know very very little about them. The grand
idea from this program is to use the data generated
via this program to support the Department of Defense missions
obviously that are of strategic importance to the US, as
(12:27):
well as public oceanographic research and commercial applications. So basically,
this modern platform is built to support a dynamic and
more efficient ocean management tool and as an improvement over
terrestrial based management techniques that lack real time data about
changes in the ocean environment. This is very important because
(12:48):
a large effect, to a large degree, the effect of
climate change and a warming planet is actually seen in
the oceans. Oceans absorbed disproportionately higher amounts of carbon and
the heat that that we have as a result of
global warming because of which it is it is. It
is a fast changing environment and we need to understand
it from the perspective of understanding how oceans are changing
(13:12):
and how that might impact whether it's feather systems down
the road for for us in especially in the warming planet.
So some of the benefits of this Ocean of Things
program is that it passively collects all of these information
such that there's no supervision needed. And for example, one
idea is in benefiting marine animals in using the acoustic
(13:34):
data collected at a very very fine spatio temporal scale,
so in at at a quick interval UH and that
is able to detect, identifying track marine animals and quantify
the background noise scapes. This helps in their preservation, This
helps in ensuring that these animals have a very habitat
(13:56):
that they can thrive in. This also this information, you know,
thely can be used to improve understanding of even marine
animal behavior in regional reasons such as such as the
Atlantic or the specific. The program also helps to develop
algorithms to automatically detect, track, and identify nearby vessels, including
(14:16):
new indicators of anomalous marietime maritime activities such as illegal
fishing behaviors. If somebody is interested, they can definitely find
more information about this program from the darka website. It's
called Ocean of Things dot dark dot dot mill Male
stands of the military part. So yeah, I mean so
it's a it's a pretty fascinating program, a remarkable feed
(14:39):
of engineering for sure.
Speaker 1 (14:40):
No.
Speaker 2 (14:41):
Actually, I'm very interested in learning, learning more. I'm very
passionate about everything in regards to the sea. I love
Ian Yes, snor Clean, I love that, and I really
understand that. Well, I know where you come from when
you say, well, the ocean can be that weakness that
(15:01):
tells us how climate change and global warming is affecting everything. Yes,
because you can pay attention to how mammals are moving
across different oceans, maybe paying attention to some of the
metrics which relate to the environmental state you mentioned carbon,
But probably you are also paying attention to others. So
(15:23):
I think it's a very clever initiative. So so thanks
for that, because I guess that goes for all of
the society, or it goes for the benefit of everyone
listening to this conversation. And can you name or explain
a few other initiatives that you know are relevant for
the conversation here, which may relate to climate change that
(15:45):
you are launching, living or somehow participating in under the
umbrella of park, I would.
Speaker 3 (15:53):
Love to talk about some initiatives currently in the very
park in this in the space of clean technology, I
would like to talk about a project that focuses on
emissions monitoring. As you know, methane is one of the
most potent greenhouse gases and responsible for a third of
current bombing from human activities, and there has been a
growing focus on methane on a curbing methane as a
(16:16):
way of buying extra time to tackle vanta change. Although,
for example, there there is more cootwo in the atmosphere
and it checks around for longer than a methane molecule,
but individual methane molecules have a more powerful bombing effect
on the atmosphere than single COTWO molecules. I'm sure most
of your audience have recently heard about the pledge to
(16:38):
cut methane emissions by thirty percent by twenty twenty thirty
at the most recent COP twenty six conference in Glasgow.
At Park this is this is an area of focus
for our clean tech portfolio and our research teams are
actively developing domain in fumed algorithms for institute and remote
sensing data assimilation for detecting methane leakage. So that is
(17:01):
one aspect of the work done at Park. Another project
that the researchers at Park have done a fabulous job
is in reimagining the HBAG system. It is again part
of our Gleentic portfolio and the idea here is that
in the challenge as well, here is that in a
warming worlds, as you mean global temperature rises and as
people become more well off, the demand for cooling will
(17:23):
only increase substantially. Now, if you look at HVAC systems
are a major cause of greenhouse gas emissions. For example,
ninety percent of US households use air conditioning. That accounts
for six percent of the country's residual energy use. In
terms of numbers, that translates to about one hundred million
tons of carbon dioxide every year. That's a lot, and
these statistics are probably getting updated every single day. Most
(17:47):
HBAG systems use harmful hydrofluoro carbon refergence. Although these are
very short lived compounds, they are thousand times more potent
and siatio in terms of absorbing heat. Without going into
the technical details of the proprietary technologies developed at Park,
researchers here have developed next generation systems that are refrigive
and free and need eighty percent less energy to operate
(18:09):
compared to a standard h BAG system. This is truly revolutionary,
honestly in terms of the impact this will have on
the human experience, human comfort, as well as in the
fight against climate change. Apart from building these novel technologies,
a lot of teams at Park, including MIND, have been
working on developing climate models that improve our understanding of
the Earth and planetary systems. One of the projects I'm
(18:31):
involved in uses AI to discover relationships between the cloud
reflectivity or albedo as the community calls it, and the
atmospheric circulations, which is very poorly understood. The technological impact
of this would be able to identify potential sites per
radiative forcing, which means that you could artificially bride the
(18:53):
marine clouds and to characterize its impact on the global
energy and precipitation cycles. There are many of the projects
at Park, including carbon capture and so many other cool
stuff that researchers are working on that are focused on
counteracting the negative effects of climate change, and honestly, we
would need a full day to discuss all of them
(19:13):
in detail. But hopefully this gives a sense to your
listeners on kind of work we do here at Park.
Speaker 2 (19:21):
No, it actually does. I'm just curious how have you
structured the team in terms of well, you surely have
data scientists, you surely have data engineers, but how did
you build that structure? Because it's helpful for me to know,
you know, I'm leaving bedrock here, and I don't think
(19:46):
there is some exact answer to how you structure the
science team around professional services, right, because you do have
people that are client facing, and then you have people
that play the client facing role that also have to
do innovation throughout the week, right and the decade sometime too,
are in the some cues to know how have you
(20:07):
build these these corporate structure at park to the extent
that you can share sure.
Speaker 3 (20:13):
Indeed. So you know, park is has been around for
fifty years and as I mentioned at the start, it
has been you know, at the at the core of
developing a lot of important technological breakthroughs such as the
Internet or the traffic user interface, the personal computing things
like that. So it has had a history of people
(20:35):
working with different people with different skills, working on very
interdisciplinary problems and bringing solutions to market. So so right now,
climate is a major area of focused within park where
we are looking at different avenues and trying to attack
it from different different different sites. For example, we have
something called us the Clean Tech Tower, which looks at
(20:58):
technologies that can be comingalized in the next few years
within the clean technology realm uh and some of these
projects are within within that scope. So and we do
have folks in the in the in the company that
are dedicated from the business side that that aim to
uh commercialize certain ventures uh some some some some folks
(21:19):
at me for example, that are more from the technical side.
We id eate and we we actually execute on these
ideas and we have a team of as you said,
data scientists, software engineers, domain experts among others to kind
of contribute to this multi disciplinary effort. So so it's
it's a pretty interesting dynamic something I have not seen
(21:39):
anywhere replicated anywhere else, to be honest, But but yeah,
I mean it definitely works.
Speaker 2 (21:47):
Yeah. Well, well it's really interesting also to me and
to us as Petruck. I think we should find some
time to talk because you have really interesting initiatives here
and there be so much much work to be done
also here in Spain. I think this is very good
learnings that you that you've made. Okay, now living park
(22:11):
a side and also trying to dedicate some time to
climate change. He during this conversation. Actually I learned as
I said about c c A when I spoke to Beings,
But after that I did out of reading around the
cases that you that you build around some very interesting
(22:32):
names that you have in the advisory board. I saw
Joshua Benjo and you n G. I mean, those those
are very popular names that you know provide an overview
of the They mentioned that this initiative has taken probably
over the previous months and years. I don't know how
did this initiative start from prea If you have that information.
(22:55):
I guess it's global already. Yeah, well yes, I mean
it little bit about the beginning of this and how
I grew up so quickly.
Speaker 3 (23:06):
Absolutely, I would love to so I quickly introduced climate
Change AI for the benefit of your audience who are
not aware of the group or have never heard of
have never heard of us. Climate Change AI is a
globally initiative to catalyze important, impactful work at the intersection
of climate change and machine learning. That's a main mission.
I'll come back to how we do it in a
(23:28):
bit and provide some examples for your audience, But now
I'll kind of try to answer the first question about
how this began. So the idea to build this organization
started from an informal discussion on how AI can tackle
climate change that was led by current CCI coachs David Ralik,
Pria Donti, and Linn Cock, alongside Europe's twenty eighteen with
(23:49):
that was organized in Montreal. The founding members realized that
if they are the large amount large number of people
who attended an informal session that the coach is organized
are so interested in working out how they could make
an impact, then probably the best way to do this
was to externalize this effort and put a list of
(24:09):
recommendations for how people could use their skills to create
impact for this challenging problem. CCI officially launched just after
several founding members co authored the foundational paper called Tackling
Climate Change with Machine Learning. I would strongly encourage your
audience to take a look at that. This paper provides
a detailed overview of how machine learning can be used
(24:30):
impactfully for climate change mitigation and adaptation across different sectors.
Since its early days, CCI has grown to become a
global community as you said, and we have representations from
all parts of the globe. The main goals of CCI
are in building a community. For example, CCI heads emal researchers,
climate researchers, climate related businesses for example, as well as
(24:52):
other stakeholders connect with each other via regular workshops you organize.
These workshops are very well attended and typically have many
thousand footfalls or digital footballs as we like to call
them these days. As well as our social media that
we have recently launched it's called the CCI Community Hub,
(25:14):
which is going very fast to reduce the barriers for participation.
We also provide funding for students to attend our workshops
when we used to have them physically or even now
the digital workshops, we have made them free for people,
and we also organize mentorship programs for folks interested in
breaking into this space to help individuals who are interested
in person climate research or do their bit for climate
(25:35):
change medication and adaptation. We also provide online resources for
them to learn more about the intersection between their fields
via programs like Summer School. The registration service which are
actually open right now, so I'd encourage your audience to
take a look at that if they're interested, peace feel
free to sign up, as well as regular webinars which
features leaders from the industry, academia, and national laboratories. In
(25:58):
addition to that, we release monthly newsletter that actually went
up today the day of the recording for this podcast
that currently is delivered to over eight thousand people worldwide
and it's growing every every month and it contains a
highly curated list of opportunities including jobs, events, calls for
funding in this area of climate change and ai CCI
(26:20):
also guides research and the area of climate change and
ai pro facilitating grants programs. Our most recent grant program
involved a total award money part of one point eight
million dollars that was that will be awarded to a
dozen or so projects starting in twenty twenty two. This
was a pretty well received program by the community. We
(26:43):
received a lot of entries for this grand program called Finally,
CCI also provides formal and informal advice to researchers, companies,
policymakers from around the globe. Most recently, CCI coachair LINKAK
provided feedback on eus AAR regulations to include impact and
climate that were discussed actually at the European Parliament in
(27:04):
Europe last week. You say, also participated in a panel
on Governance in AI and climate change within German pavilion
at TOP twenty six conference in Glasgow. So again, this
is only a small fraction of all that CCI is doing.
It's growing really fast thanks to the efforts of my
colleagues who are not here, who are not here in
present today. For anyone in the audience that is interested
(27:26):
to learn more about us, please check out our website
climate Change that AI and sign up for a monthly
newsletter to keep updated about our activities.
Speaker 2 (27:35):
Yes, no, I'm actually recommending that as well, because when
I took a look everything that you were doing. Some
of the names that you had in there, I mean
people from stam for the mind and everything. It's surely
we'll cut a lot of people's attention. I didn't know
that you also participated in the new EU AI legal
(27:59):
framework is coming and I think is pretty much needed
on different fronts. I guess that your area of specialization
relates to environmental AI, so climate change AI, but there
are different fronts that you could also contribute as well.
I mean, if you pay attention to the names that
(28:22):
you have in the team, your scope of advisory could
go well beyond climate change. I guess it's relevant because
it's climate change AI, but I'm sure that they would
appreciate that help across other domains as well. Okay, I
did read that two of the founders of CCI were
(28:44):
named among mi T Technology Reviews words. I think it's
something like thirty five innovators and they're thirty five, well
thirty howl these two because I do not remember for
their names, so probably you can you can mention they
were recognized for doing research on different topics around optimizing
(29:10):
power greeds, even monitoring biodiversity or something like. You mentioned.
Also what you do at park. Maybe it's a relevant here,
so it's part of the learnings that you can use
in this space. But can you share more insights on
how they did this? I mean, I've mentioned research on
optimizing power greeds and monitoring biodiversity, but I'm curious to
(29:33):
know the data collection technology, data infrastructure mL algorithms that
you've used, any insights that you can share on these
research pieces that led you, I mean, these co founders
of CCA to get in this list.
Speaker 3 (29:51):
Absolutely so we were super excited when the scheme up.
You're very proud of the work and impact made by
to CCI coaches. The TUCCI coaches that you were talking
about are Priya Donti and David Rolneck that were named
in the Technology Reviews thirty five and A thirty five list,
as well as Lin Khak, who is our third CC
(30:13):
culture and when we're very proud of their contributions, active
contributions to the scientific community. They publish, you know, regularly
as well as they do this. This this really stellar
work by CCI will not While I'm not entirely privy
to the work done by them or what the reviewers
that Tech Review were looking at, I know and David
(30:36):
have both looked extensively in the constrained learning space, especially
for problems involving optimization that that have hard constraints, such
as in the energy and the power sector. So I
would defer you to their papers that are publicly available
because I'm not fully familiar with with what they exactly used.
(30:57):
But you know, they they did a lot of work
in the area of constrain learning, physical conscreen learning, or
with heart constraints. Uh. And they have an excellent paper
that came out recently called DC three uh And And
you know, so, so if if anybody is interested in
the audience, the papers are available publicly. Yeah, we're curious.
Speaker 2 (31:18):
Yes, I will take a look at them. Apart from
what we've mentioned, and well, you've already mentioned a few
or we've mentioned a few breakthroughs as a global movement,
but have we missed any that you think will be
(31:39):
impacting our society in the short medium term that the
audience would be aware of.
Speaker 3 (31:45):
Sure.
Speaker 4 (31:47):
So you know, first I kind of talk about some
other interesting things that you mentioned about, like how c
C I can can you leverage some of our advisory
boats to kind of make a bigger impact. So since
you talked about US being a global movement and the
achievements of CCI as a global movement. Let me talk
(32:08):
about the recently released recommendation on Climate Change and AI.
So these recommendations are for government action and this was
released along the lines of COP twenty six conference. This
new report, developed by CCAI and Center for AI and
Climate for the Global Partnership on AI, offers forty eight
specific policy recommendations for governments on how to support work
(32:30):
at the intersection of AI and climate change and how
to better align the use of AI overall with climate goals.
Some of the use cases highlighted.
Speaker 3 (32:38):
In the report include the National Grid ESO, which has
used AI to double the accuracy of its forecast for
US electricity UK elliptricity demand radically improving the forecast of
electricity demand and renewable energy generation is definitely a critical
aspect in enabling greater proportions of renewable energy on electric
(32:58):
grids because you have been lable energy all the time
in the day. So if you can able to if
you're able to forecast the demand, you can you can
plan for that. The UN Satellite Center has developed for
these example is the UN Satellite Center has developed this
flood AI system that delivers a high frequency flat report
that is actually used UH and has improved the disaster response,
(33:20):
the climate related disaster response in Asia and Africa. The
third example that that's this report talks about UH is climatories.
It is a global coalition of organizations that is actually
working to radically improve the transparency and accuracy of emissions
monitoring by leveraging AI algorithms and data from more than
(33:43):
three hundred satellites and eleven thousand sensors. So so these
are some of the like you know, other impacts that
Climate change AI has made in the space, especially making
these specific policy recommendations to governments and people who have
the power to impact impact all of our lights on
a very large scale.
Speaker 2 (34:05):
I was standing a fair yes it was like a
fair yesterday imagrid mainly related to renewal energies. There were
many stands about solar energy on different fronts. Some of
the engineering companies doing or advancing their work around the
(34:28):
movement of the panels to ensure that you get the
maximum energy from the sun. I mean different initiatives, many
interesting ones, just because as you as you already know,
the the solar field is a world in itself, there
is so much to be learned. I mean, data and
analytics for EI is a big thing with many different
(34:51):
sup fields, but solar is another one. So I'm curious
to know if there is anything that you've done in
regards to solar paneling, specifically from climate chain to AI
that we could highlight to the audience.
Speaker 3 (35:08):
Indeed, so, as as I mentioned earlier, climacy c I
is a group of very highly multi disciplinary people who
actively contribute to the scientific community as well as we
organize these workshops alongside major machine learning conferences that sees
a that sees an impressive list of participants, that that
(35:28):
that share their work with us, has a very regreous
review process to ensure quality. Uh and I would highly
you know, recommend it to your audience to kind of
come check out some of these papers that are archived
on our website. So so for example, you talked about
solar panels and and and all these important things. So
I'll give you an example of a paper that was
(35:50):
presented in our in our workshop in the past that
use machine learning to deduct solar installations globally was actually
published in Nature, you know, the one of the world's
most prestigious journals recently. Well, I'm not trying to take
credit for their work or suggest that our review process
helped them, but it is just commenting on the high
quality impact people are making using our resources. So so so,
(36:15):
there are so many other papers where or other projects
that you will probably see featured in our workshops that
actually use machine learning to identify optimal sites for installation
of solar panels, for example, lars scale solar farms that
that can generate reliable electricity for for for the community.
(36:39):
So these are some of the high back areas as
well as there is a lot of interest and work
by the community that is looking at developing better materials
for you know, improving the efficiency of the solar panels
that is actually improving by the day. As well as
how you can use machine learning to detect fault in
(37:00):
solar panels, for example, if there's a crack, or if
there is certain anomalies due to manufacturing or other things,
how can you use machine learning to detect all of
that to ensure your production you know, doesn't suffer because
of certain certain effects that that you know happen due
to certain way andeer that happens during the use of
(37:22):
an engineering platform. So there are multiple ways that machine
learning is looking into that, and a lot of these
are actually reflected in our workshops, you know, community members
kind of present their work in our workshops and you
can see all of these papers on on our website.
They are archived for the benefit of of of the
(37:44):
audience such as yourself.
Speaker 2 (37:46):
I mean we we are already doing some work in
these field. So I think, as I said before, we
need to set some dim side to learn about the
advancements that you've made in this field. Yes, because some
of the players in the smallar industry, well they are
busy in their day today, right, and not everyone in
(38:07):
these companies have dedicated enough time to pay attention to
how mL could help them around different meds. They've looked
into predictive maintenance, but sometimes they haven't looked into panels efficiency,
using advanced analytics and things like that. So at some
point they will need to be supported by companies like
(38:31):
ours as at truck. So learning and exchanging knowledge with
an organization like yours as CCA, I think would be
would be an amazing starting point. Okay, so this is
the first probono initiative that I see around the world
(38:53):
in the intersection, as you said, of machine learning or
AI and climate change. Is there any other initiative, whether
national or global, that you know of that is also
helping you in this mission and that also serves the
same vision.
Speaker 3 (39:14):
Indeed, you know, there are so many other groups that
we that that that we follow data that are doing
excellent work in this area as well. For example, Climate
Informatics is a group that has been around for about
a decade or so, started by Professor Klaremant Montellioni, who
is actually one of the advice who is one of
(39:36):
the members of the CCS Advisory Board, and other people.
And the goal of climate informatics is is and not
to describe it in my own works, but the goal
of climate informatics is to apply statistics, machine learning and
data science to solve problems relevant to climate science, so
the science of climate, including climate modeling and how the
(39:57):
climate climate changes, so on, so forth. U. There is
also another group that aims to use quantum computing for
climate change. You know, obviously quantum computers, you know, they
are almost on the horizon. They would be extremely fast,
so they can actually help in help in detecting better
or or or understanding what materials could actually help improve
(40:20):
solar efficiencies for example, for these solar service or battery
research or accibertate, some of these areas that are extremely
critical for us to transition into a zero net, zero
carbon world. So that is super interesting. The quantum computing
for climate change, I you know, and and and honestly,
what if the tools the community uses it is It
(40:41):
is actually super exciting to see a lot of creative
folks working on these innovative solutions out there. So yeah,
there are certain these are certain groups that come to
my mind. There are obviously a lot more, and I
apologize for missing.
Speaker 2 (40:54):
Out on them. You probably know if you've heard or
re listened to the previous podcast that that we've made
with with well many interesting guests. But now that we
come to the to the end of this core I
always ask for someone to join us in this data
stand up podcast. And you've mentioned a few initiatives and
(41:17):
probably you have some other professionals in your radar that
you think would be you know, with guests, So I
don't know, could you give me a couple of names
and then make an introduction for us to have him here.
Speaker 3 (41:32):
Sure, indeed, So I think I think you know, because
I'm from the industry in the US and in the
b Area, I'm probably a little bit biased towards, you know,
the kind of impactful technologies that that these companies are building,
especially because the climate problem is so big, we need
we need technologies that can be developed fast and at scale. So,
(41:54):
for example, Albert recommend you know some some folks like
Steve can grasp from Climate AI. Climate e I is
separate from Climate change. AI is actually a startup in
the Bay Area that is using AI for precision agriculture,
so agriculture in a changing climate, because you know, the
predictions are that as as the global warming increases, there
(42:19):
will be more more and more crop failures and stuff
like that. So so Climate I is actually using AI
to help farmers get a better yield depending on local
conditions and so on and so forth. So so Stephen
rasp would be an excellent candidate to talk about his work.
Stephen Heyer, So he's from Google AI. They are working
(42:39):
on He and his his team has worked on near
term precipitation. So how can you know how can AI
and generative models use UH be used for for predicting
precipitation in a very short time window. And and they
have done excellent work in that space. Shak Mohammu from
(43:00):
a deep Mind. Once again it's a Google company. Now, uh,
they have done some excellent work in the same area.
I think they have built even better model than Google
that does the snow casting work. And and Shakira is
super super active in this area of of you know,
climate change machine learning as has has has been a
(43:24):
part of our previous workshops as well. So yeah, these
these names are its strongly recommend for you to to
look into happy to make introductions as as you see
for it.
Speaker 2 (43:34):
Yeah, well that's amazing. I mean I usually get one
or two names, but there, that's really good. Thanks for that.
And yes, before we close the call of I also
ask for you know, a book, a newsletter or some
reading material that you highly recommend. And it doesn't have
to be specifically around climate change and AI or even
(43:56):
around data. It can be around anything because in the
past they had really good segestion so on. I don't
know how you structure your day to day to be
more effective and productive things like that, So anything that
you can think of, yes, yes, one piece of reading
material that you recommend.
Speaker 3 (44:16):
Indeed, so you know, well you can call me a biased,
but i'd highly recommend your audience to sign up for
our c c AI newsletter if you haven't already jokes
about since you said newsletter as well in your question, uh,
newsletters from organizations like Climate Piece, Climate AI, which is
(44:37):
separate from c c A. I batch some of the
pretty interesting news letters in this area. About books, you know,
there are a lot of interesting books I've I've read,
but I think one book that I'm currently reading, you know,
catches my attention and I think it's it's an important
book for all of us to read and understand. Is
It's a book called Speed and Scale by on Tour.
(45:01):
It is a book about series of action plans for
solving the climate problem right now using technologies that exist today,
are solutions that can be developed fast and scaled up
to massive magnitude of climate change. I think I would
highly recommend it for everybody who has you know, has
has has an interest in solving this problem, not just
(45:22):
standing by the wayside, but actually contributing to any of
these solutions. And once again, I would highly recommend your
audience to come join us at t C I uh
t C I is an actually open community and very
welcoming community. So we would love to have some of
your audience members join us in this, in this in
this great emission of making making making an impact in this,
(45:46):
in this big fight against climate change.
Speaker 2 (45:49):
Not for sure, Well, I'm going to tell my colleagues
here and I think again jokes aside what we've discussed
during this call, both around sett in some time aside
to talk about different learnings and changing knowledge on different
pieces of work. I think bedrocs should be in some
(46:10):
sort of way supporting these initiatives. And I don't know
if you've previously established any connections with any Spanish companies
that operates in the data science in aifield, but we
would love to be a part of this. So I'm
saying this now while we are alive, but I think
it's something that we should look into because the work
that you do is highly relevant to all of us
(46:31):
as a society, and I think that this can be
articulated in different fronts. So the same way you are
supporting the EU for those legal frameworks, there is some
hands on work that requires time and if we can
put some resources to use and learn together with you. Well,
(46:52):
that that partnership would be would make us proud for sure. Anyhow,
bit that really thanks for taking time. We and I appreciate, uh,
the time that you've taken in in in thinking about this,
this this podcast, and the time that you've dedicated today
taken my call. I think you have a bright future
(47:14):
ahead of you, well yourself as Speedup, but also as
as part of park in climate change. I so congratulations
to that and and good luck to you.
Speaker 3 (47:24):
Well yes, yeah, thanks, thank you for the very kind words.
You know, we would you know, definitely love to chat
with you offline. It was a pleasure to be on
this podcast and I cannot wait to see the great
things you and your colleagues will do in the future.
Thank you very much
Speaker 2 (47:40):
As this Thank you half a good one Pickers Productststructus
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
The Breakfast Club
The World's Most Dangerous Morning Show, The Breakfast Club, With DJ Envy, Jess Hilarious, And Charlamagne Tha God!