Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:02
Latitude Media, podcast at the
0:04
frontier of climate technology. Hey
0:08
Chad, are you there? Hey, yeah, I'm here.
0:10
What's up? Back in May, OpenAI
0:12
released GPT-4-0, and it was a big
0:14
leap forward for voice and video interaction
0:16
with a chatbot. Hey, let's do, let's
0:18
have some fun. There are a
0:21
bunch of demos from OpenAI like this one,
0:23
where the chatbot is mimicking human emotion in
0:25
ways we really haven't seen outside science fiction.
0:28
Uh, I'd like you to be super sarcastic. Everything you
0:30
say from now on is just going to be dripping
0:32
in sarcasm. How does that sound? Oh,
0:36
that sounds just amazing. Being
0:39
sarcastic all the time is exhausting
0:41
or anything. I'm so
0:43
excited for this. Nope,
0:46
sarcasm. Let's get this party started.
0:48
Or whatever. Generative
0:52
artificial intelligence has come a long way
0:55
in just the last two years. Built,
0:57
of course, on many decades of research
0:59
before. A lot of the stuff
1:01
that is being leveraged today in the AI
1:03
world comes out of language models. And we
1:05
were doing very early days of speech recognition
1:07
and language models back 20, 25 years
1:09
ago. From the late 80s into
1:11
the mid 90s, Mike Phillips was a research scientist
1:13
at MIT. He worked on
1:16
speech recognition and natural language processing. Mike
1:18
founded multiple companies in the speech recognition space,
1:21
one of which built the voice assistant for
1:23
the Samsung Galaxy phone. In the early
1:25
days of these voice assistants, smartphones were
1:27
just coming out. We kind of
1:29
realized that the phones were getting connected to the
1:31
data networks and they were going to start
1:33
to be more like the internet and then
1:35
therefore become people's personal information,
1:37
communication, entertainment device. And we also
1:39
realized, well, you know, you
1:42
shouldn't have to like type on a little,
1:44
little keypad. You should be able to just
1:46
talk to these things. After selling both of
1:48
those companies, Mike turned to energy and he
1:51
asked what platform can unlock savings and decarbonization
1:53
that is analogous to the smartphone. He
1:55
co-founded Sense, which created an energy monitor
1:58
installed in an electrical panel. to
2:00
give real-time information on every device in a
2:02
home. Look, if smart devices were fully enabled
2:04
and every device out in your home told
2:06
us what it was doing, we'd be happy
2:08
to use that. But it's far
2:10
from being the case. And we realized that
2:12
if we could measure the power in a
2:15
detailed, left-right way, could we figure out what's
2:17
going on just from the power signatures? So
2:19
you can start to see where the similarities
2:21
come in, that we used to
2:23
do speech recognition on audio signals.
2:25
We are now doing electrical device
2:28
identification based on power signals. So
2:37
the company was started in 2013, and
2:39
then you had years of time deploying
2:41
devices in the field, and then you
2:43
discovered that suddenly you had
2:45
a lot of visibility beyond the home.
2:48
What kind of view of the grid
2:50
did desegregation uncover? The first thing we
2:52
did was collect some signals and realize,
2:54
oh, look, we can't do this unless
2:56
we have high-resolution data. Our
2:59
initial view is we had previously written
3:01
software for smartphones, we'd write software for
3:03
smart meters, and we'd be all set,
3:06
quickly found out that the existing smart meters just
3:08
did not have the data we needed. So that
3:10
let us down a long path that many of
3:12
you know that we started to build these little
3:15
orange boxes that go inside
3:17
electrical panels and collect data at super
3:19
high resolution, up to a million samples
3:21
a second in the little orange box.
3:23
So that high-resolution data was the key
3:25
to unlocking what happens in the home.
3:28
And now to your point, we realize
3:31
sometime afterwards that that same technical
3:33
capability, so high-resolution data, edge computing, real-time networking
3:35
that we use to interact with consumers on
3:38
what's happening in the home, we can look
3:40
the other direction and we can see what
3:42
the grid is doing in real time from
3:44
the edge. Are you saying that the first
3:46
class of smart meters weren't actually as smart
3:48
as we think they are? Hey, you said
3:50
it, not me. SENSE
3:52
is now focused on putting its machine-learning
3:54
technology into new generations of smart meters.
3:57
Mike is bullish on the role of advanced meters for
3:59
grid intelligence. but as utilities
4:01
start a new wave of rollouts,
4:03
he worries they aren't investing in
4:05
the right architecture. Most people think
4:07
of meters just as data collection
4:09
devices. You have to start
4:11
to change that mindset. And once you start
4:13
to think of this as a distributed platform,
4:16
not just a data collection device, this
4:18
entire world of making use of the data
4:21
at the edge and AI machine
4:23
learning at the edge starts to get opened up.
4:28
This is The Carbon Copy. I'm Stephen Lacy.
4:32
This week, a conversation with Sense CEO
4:35
Mike Phillips on what AMI 2.0 could
4:37
and should look like. Past
4:40
deployments of smart meters didn't bring the intelligence
4:42
promised. How do we avoid the same outcome?
4:51
America's green banks are preparing to unleash
4:54
a wave of capital for clean energy.
4:56
The greenhouse gas reduction fund invests a
4:58
historic $27 billion in projects nationwide. This
5:00
could mobilize up to $150 billion of
5:02
private capital for
5:04
solar, storage, efficiency, and electrification
5:06
in underserved communities. So
5:09
how do we deploy those billions quickly,
5:11
efficiently, and with the highest impact? On
5:13
July 18th, Latitude Media and Bandion Infrastructure
5:15
will host a virtual event exploring the
5:18
on-the-ground realities of making America's green banks
5:20
a success. Register for free by clicking
5:22
the link in the show notes or
5:25
go to latitudemedia.com/events. Clean
5:27
energy and climate tech are policy-driven industries
5:29
and anyone working in this field touches
5:31
local, state, and federal policy in a
5:33
very real way. And that's why you
5:36
should be listening to Political Climate, a
5:38
podcast from Latitude Media and Boundary Stone
5:40
Partners that delivers an insider's view on
5:42
climate policy and politics. Every other week,
5:45
co-hosts Julia Piper, Emily Dominić, and Brandon
5:47
Hurlbuck cover the nuances of government funding,
5:49
regulations, backroom negotiations, and the election, of
5:51
course. Political Climate is a
5:53
show for people who want authentic conversations
5:56
and strong opinions from voices across the
5:58
political spectrum. Listen at Latitude Media. or
6:01
subscribe to the show anywhere you get your
6:03
podcasts. So
6:09
as you said, the first
6:11
generation of smart meters were just
6:13
seen as data collection devices. And
6:16
most people who look at the
6:18
rollout of AMI 1.0
6:20
would say it did not live
6:23
up to expectations. Ultimately,
6:26
what do you think the limitations of the
6:28
first rollout of smart meters was? Was it just that
6:31
the devices themselves weren't powerful enough? Was
6:33
it that it was the wrong infrastructure?
6:36
Was it just as a data collection
6:38
device, the utilities didn't know what to
6:40
do with that data? And they had
6:43
an inability to use it properly. Give
6:45
us your broad view on what went
6:48
wrong with AMI 1.0, which
6:50
I think most people think wasn't that
6:52
successful. Yeah, look,
6:55
I think it was basically the wrong
6:57
architecture. Look, it made sense back in
6:59
2008 or whenever that happened, but it
7:01
hasn't transitioned since then. So
7:04
by wrong architecture, look back to my
7:06
telecom example, imagine if the way Google
7:09
Maps on your phone worked is
7:11
your phone would collect 15 minute interval
7:13
data of your location, send it up
7:16
in batches to your telecom
7:19
provider, who would then make it
7:21
available through what gray button instead of green button or
7:23
something like that. And then applications like
7:25
Google Maps could get that data a
7:27
day later and then
7:29
do something with it. Well, what would Google Maps on your
7:31
phone be? It would be like a static map and
7:34
maybe a monthly historical report of your
7:36
traffic on your route to work and
7:38
maybe compare your neighbors, how
7:40
you drive compared to your neighbors. But would
7:43
you use that app? You
7:45
might use it every now and then, but you would not engage
7:47
in it in the way you do with Google
7:49
Maps. And what I just
7:51
described exactly matches AMI 1.0
7:54
architecture, low resolution data sent up to
7:56
the service provider, made available later on
7:58
in the day. And look, there's some
8:01
things you can do with that, but you just can't
8:03
have a real-time consumer-facing
8:05
app, and then you can't
8:07
see the grid in real time from the edge either. So
8:10
we're at a point now where we're
8:13
rolling out a lot of new meters.
8:15
We've had a decade to develop the
8:17
architecture. What is your sense now for
8:19
the current architecture that utilities are reinstalling?
8:22
We're crossing the threshold now. Look,
8:25
there are meters available today on the market
8:27
that do all the stuff we want to
8:30
be able to do and unlocks all this
8:32
potential. And by unlocking potential,
8:34
let me circle back to something
8:36
that you all are quite into,
8:38
and everyone's talking about AI for
8:40
the grid. Well, look, we
8:42
know AI is driven mainly by
8:44
machine learning-based approaches these days, and
8:47
that's mainly driven by data. And
8:49
if you don't have the right data, you're
8:51
kind of stuck. And there's a lot of
8:54
talk in the industry about grid edge intelligence,
8:56
but people are mainly talking about taking data
8:58
from the edge, this 15-minute interval data, processing
9:01
it in the cloud. And there's some things you
9:03
can do with that. I'm not denying that. But
9:05
to fully unlock the potential for AI for the
9:07
grid, we need the right data.
9:09
And this is what we learned long ago
9:11
at Sense, to have a real-time consumer experience.
9:14
By real-time, I mean you turn on your
9:16
microwave and it shows up in the app
9:18
a second later. To have
9:20
that kind of experience, you can only do
9:22
it with high-resolution data. What we've learned since
9:24
then is that same high-resolution data lets us
9:26
see the grid from the
9:28
edge, lets us see transformers arcing, lets
9:30
us see vegetation hanging power lines in
9:33
real-time. That only happens through high-resolution data.
9:35
So sorry to keep going on
9:37
about this, but the number one step is to
9:39
get the right data in
9:41
the meters themselves and be able to
9:44
process it there. Well, I know you've
9:46
been somewhat disappointed with how some utilities
9:48
are reinvesting in their metering networks. Is
9:50
that just because they're essentially investing
9:53
in a technology that is not much
9:55
better than the first generation? Yeah,
9:57
we are certainly worrying about
9:59
this. that there's a big opportunity
10:01
now that, like I say, there are meters
10:03
available today that can provide a lot of
10:06
headroom for what happens in the future. And
10:09
there's still decisions being made for a
10:11
previous generation of meters. So we're trying
10:13
to help utilities and help others. And
10:16
in fact, we're just publishing up on
10:18
our website an AMI buyer's
10:20
guide. And look, we don't have
10:22
the full picture of all the things that a
10:24
utility needs to consider for AMI deployments, but we
10:26
do know a lot about how
10:29
to deploy data, intelligence,
10:31
AI at the edge of the grid. And
10:34
so we're putting together the point of
10:36
view of all the things you need.
10:38
High resolution data, enough computation, and
10:40
the ability to have real-time networking are the
10:42
three things you need, but you've got to
10:44
pay attention to the details and get it
10:46
right. So give us a sense for what
10:48
the generation of meters that you're working on
10:50
can give us in terms of data versus
10:54
the traditional meters. The
10:57
most basic thing is to get high
11:00
resolution data of voltage
11:02
and current in the grid. Let's
11:05
just see what's happening at home. Let's just
11:07
see the grid and buy that. Sorry to
11:09
go into some of the techie details, but
11:11
this means continuous sampling of these voltage and
11:13
current waveforms of at least, there's
11:15
meters on the market that are doing that at
11:17
15,000 times per second. So that's 50
11:20
million times more data than AMI 1.0. And
11:23
again, for those of you who get scared, we don't even know how to
11:25
deal with AMI 1.0 data, we don't
11:27
mean you send all that up to the network,
11:29
you're processing at the edge. So with 50 million
11:32
times more data, we see the first 100 harmonics
11:35
of signals at the edge, and that is what
11:38
we need to see what's happening inside the home, see what's
11:40
happening in the grid. So
11:42
those meters are available today. Utilities can
11:44
go get rate cases approved
11:46
and be able to deploy them, and
11:48
it's happening in pockets right now. And
11:52
look, I know there's been a lot of talk,
11:54
and we're involved in this too, about what comes
11:56
after that also. So we're starting to work with
11:58
meter makers to go... all the
12:00
way up to a megahertz sampling. So
12:03
there's progression and things are happening in this
12:05
world, but there's meters on the market today
12:07
that are sufficient for doing what we do.
12:10
And we wanna make sure that utilities are
12:13
fully equipped with the information about that so
12:15
they don't go by the previous generation. On
12:23
July 18th, join Latitude Media and Banion Infrastructure as
12:26
we take a deep dive into the next phase
12:28
of deploying the $27 billion greenhouse
12:31
gas reduction fund. We'll provide practical insights
12:33
for the state agencies and local lenders
12:35
at the front lines of dispensing these
12:38
funds. Where are the bottlenecks? How do
12:40
we balance speed with transparency? And can
12:42
America's green banks live up to the
12:44
expectations of both local communities and Wall
12:47
Street? Latitude Media's Stephen Lacy, Banion co-founder
12:49
Amanda Lee, and Clean Energy Fund of
12:51
Texas EVP Billy Brisco will answer these
12:54
questions and more on July 18th at
12:56
1 p.m. Eastern. This is a must-attent
12:58
for project developers and financiers. Register for
13:00
free at latitudemedia.com/events or click the link
13:03
in the show notes. I'm
13:06
Julia Piper. I'm Brandon Herbert. And I'm
13:08
Emily Dominich. A little over a year
13:10
ago, political climate took a break so
13:12
we could focus on the groundwork of
13:14
implementing America's biggest ever climate bill, the
13:16
Inflation Reduction Act. I'm excited to say
13:18
political climate is back. And I'll be
13:20
joined by my two co-hosts to riff
13:22
on the top political stories and insider
13:24
scoops from state houses to the halls
13:26
of Congress to regulatory agencies and even
13:28
international climate talks. We'll explain how those
13:30
developments are driving industry decisions today. Political
13:32
climate is a show for people who
13:34
want authentic conversations and to learn about
13:36
how energy and climate policy is shaped
13:38
within both political parties from the people
13:40
who have actually helped shape it. So
13:42
join me, Brandon and Emily every other
13:44
week starting in April for fresh episodes
13:46
of Political Climate. Subscribe to the show
13:48
wherever you listen to podcasts. So
13:51
GPU chip maker Nvidia is working with
16:00
people can see what's going on in their home and track
16:02
down energy hogs, we call them in the home. It
16:05
also is having a big
16:07
impact on people's participation in
16:09
demand flexibility, load flexibility. How
16:11
do you deploy time-infused rates or demand
16:13
charges if you can't let the users
16:16
see what's happening in real time in
16:18
their homes? And then we're
16:20
also using this high-resolution, detailed view of what's
16:22
happening in homes for helping the electrification, find
16:24
the homes that are best candidates for heat
16:26
pumps and so on. Okay, so
16:29
that's the consumer side. What about the grid side? Well,
16:31
this is actually a surprise for us.
16:34
We did not actually realize that the
16:36
edge of the grid is largely unmonitored.
16:40
And again, the technical capabilities in
16:42
these new meters, high-resolution data, edge
16:45
computing, real-time networking that
16:47
we used to see inside the home, we can see the
16:49
other direction to the extent that
16:51
when things happen on the grid, if it
16:53
shows up as power fluctuations in the home,
16:56
we see at the edge. And
16:58
when you just look at power once a second, which
17:00
is what was done even in the internals of
17:03
AMI 1.0, you don't
17:05
see that much. You see big
17:07
level shifts whether conservation
17:10
voltage reduction is going too far and so off.
17:13
You could do that with AMI 1.0. As you get
17:16
to high-resolution data, even tiny
17:18
little arc flashes that are happening on
17:20
the secondary side, so transformers that are
17:22
arcing problems in
17:25
the meter socket and stuff like this, we
17:27
see all that in detail. And
17:29
things like vegetation hitting primary
17:31
lines. We see
17:33
the entire, all the way up through the distribution
17:36
system and all the way into the transmission system.
17:38
So when things happen, we see it
17:40
at the edge. So what are
17:42
the consequences of getting these meter
17:44
rollouts wrong? These rollouts tend to
17:46
happen on around 15-year cycles. So
17:48
the first cycle was between 2008,
17:50
2011. And
17:54
so now we have a bunch of utilities that need to
17:56
reinvest, as we've said. What are
17:58
the consequences for... It's
22:00
got some very particular details that we've been
22:02
deep into. We
22:05
also are seeing the parallels of what's happened
22:07
in audio processing and so on. We
22:09
actually have someone coming from leading
22:12
these efforts at Google is going to be
22:14
joining us shortly as an advisor for the
22:16
company. He's really helping
22:19
us take all the latest
22:21
techniques of what's happening in the big
22:23
tech companies and apply it to this
22:25
particular problem. You also
22:27
have a new generation of meter that's coming
22:29
out as well, right? As
22:32
I said earlier, the meters on the market
22:35
today are great. They support,
22:38
like I say, we can process up to
22:40
the 100 harmonics of the power signals. It
22:42
supports the use cases that
22:44
we see out there today. Technology
22:48
is always advancing. We
22:50
are working with some of the meter makers to have
22:53
the next generations of meters that have even higher sampling
22:55
rates than that going all the way up to a
22:57
megahertz. That's unlocking some
22:59
more use cases around arc fault
23:01
detection. When there's an arching on
23:03
a transformer or something like that, it shows up in
23:05
these high frequencies. If you start to
23:08
worry about even a broader set of use cases, continue
23:11
to push the envelope on data
23:13
and, as you mentioned, more and
23:15
more computation. It's a tricky topic.
23:18
This happens in every technology field. We
23:21
don't think you should wait because the latest generation
23:23
is great. If you're making a meter decision now,
23:25
you need to replace your meters. There are meters
23:27
on the market that do what we need to
23:29
do. The next step
23:31
is coming here shortly. Mike Phillips,
23:33
co-founder and CEO of Sense, thank you so
23:36
much. Thank you, Steven. That's
23:45
going to do it for the show. If
23:47
you want to learn more about AI and
23:49
smart meters, we're covering that subject in particular
23:51
and AI generally at latitudemedia.com. You can subscribe
23:54
to our newsletter there. You'll get all our
23:56
editorial coverage in your inbox in the middle
23:58
of the week.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More