Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
AI: Enhancing The Possible in Building Automation
39 Plays6 months ago
In this episode, Deepinder Singh, founder and CEO of 75F, joins Aakash Bhasin, Managing Director at Alantra to delve into the transformative impact of AI on the building automation sector. Deepinder explains how AI-based machine learning makes it possible to optimize automation in complex environments like commercial buildings, contributes to decarbonization goals, and mitigates climate change impacts. He also discusses why large language models are helping democratize technology by making expert knowledge accessible globally.
Recommended
Transcript
Introduction to Deep Singh and 75F
00:00:06
Speaker
Good morning, everyone. Welcome to Alantra's podcast. Today we've got a very interesting speaker, Deep Singh. Deep is the founder and CEO of a company called 75F.
AI in Building Automation: Market Trends 2024
00:00:20
Speaker
75F leverages AI to make smart building automation systems and make them in a way that they can be leveraged not just by large buildings where traditional BMS go today, but also by small and mid-sized buildings. My name is Akash Pasin. I'm a managing director at Lantra. I lead our building technology practice based here in New York. And hopefully you'll find the the podcast helpful and interesting. The first question I really wanted to get into is where do you see the market for building automation
Why US Buildings Need Smart Systems
00:00:53
Speaker
solutions in 2024? And twenty twenty four what is your outlook for some of the key trends you expect in that market over the next decade?
00:01:00
Speaker
That's a hot one Akash, I think. Overall, in general, this building sector itself has been highly neglected. Most of the automation that has really occurred is in a very small sliver of buildings. So we keep looking at the places like the Burj Khalifa. We look at all the the skyscape, which looks in New York, looks really cool. But unfortunately, there's a whole lot of buildings which are outside core downtown Manhattan. It's interesting that basically, even in a country like the US, about 84% of the buildings have got absolutely no automation at all.
00:01:33
Speaker
The current systems are really more visible, but that doesn't mean that they have actually gained a huge amount of penetration. So what we are really seeing is a proliferation of these technologies, which would allow more automation. If we talk about the more renewable future, it is imperative that these buildings are automated and smart so that they can actually interact with the grid real time.
Tech Deployment in Small Buildings: Challenges & Trends
00:01:55
Speaker
And that's going to be very, very key from our perspective. So I think that is overall building stock needs to be retrofit the overall number of new buildings especially in developed countries is reducing so we'll see a lot more of the smaller buildings coming up which traditionally have not necessarily been automated so i think there's a huge opportunity to deploy solutions which are relatively easy to deploy which can be put in much faster.
00:02:21
Speaker
And more importantly they can be put in by people who are not necessarily controls experts ah we keep hearing about this acute shortage of train labor and that is a true. Thing that the skills which are required and the complexity which is growing it is unlikely that we're gonna be able to find enough train professionals to be able to truly go and cover. these buildings in a wide swath, but I think one of the saviors is how can we actually deploy technology to overcome some of these manpower shortages and how can we make things easier so that these solutions can be scaled across a much larger swath of buildings.
Rising Interest in Smart Energy Management
00:02:56
Speaker
Even with some of the data that we've looked at just in the US s alone, the six million commercial buildings,
00:03:02
Speaker
And vast majority of them are below 200,000 square feet, you know, small and minimize buildings. There has always been this talk about shortage of trained folks that can actually install BMS systems. I think some of the trends that you're talking about in terms of easy to use, easy to deploy, I think that's going to become very, very important on a goal board basis.
Global Adoption of Energy Management Systems
00:03:22
Speaker
Why is the client interest in smart energy management systems increasing? I think we've seen a fair degree of interest in people wanting to deploy these systems. And how do these systems really interact with broader BMS systems?
00:03:35
Speaker
I think primarily it's really being driven by the COP agreements. So one of the key things, what's called a global stock trade is to actually figure out where is each of the countries, the 200 countries who are signatory to this agreement, they're going to take stock of where they are in terms of the remissions and they've agreed to cut them down significantly, I think it's 46% from the 2019 levels. That's a very aggressive goal and this is by 2030. So we have a short window in which A, we have to assess where we are in terms of our emissions and then we actually have to cut it down by 46%. And unfortunately, we have actually still exceeding those ah let and carbon emissions from where we were in 2019. So it's not as if we have actually made any progress to these goals.
00:04:17
Speaker
So we have some very aggressive goals and the governments are recognizing that they need to have both the carrot and a stick. So you'll see a fair amount of legislation, but you also see a fair amount of incentives that the utilities of the governments are giving to make sure that the buildings are run more efficiently. From a more global perspective, are there like certain geographies that are likely to see greater deployments of these smart energy management systems versus others? Right now, globally, Europe has been one of the better citizens in terms of enforcing what's called EPBT, the European Directive around Building Performance. but And what they're doing is they're actually mandating that buildings over a certain size have a certain EUI energy use intensity. They're being far more aggressive than the other countries in the world.
00:05:03
Speaker
I think the US is kind of following suit now, specifically at the state level and at the federal level. But ultimately it boils down to where the money has been committed.
AI's Role in Democratizing Building Technology
00:05:11
Speaker
And that's been in Europe and the US specifically at this point. I think the Middle East is trying to be good citizens themselves. And Saudi Arabia, of course, is trying to also use some of their free cash flow that they have and pour that back into making sure that the new construction that is going on there is more efficient. Thank you, Deep. The main topic, which is what impact is AI having on the building automation market and how do you expect AI technologies to transform the individual experience in commercial buildings in particular and possibly other kinds of buildings? AI seems to be really coming into the forefront, so I would love to get your views on how it's impacting your broader ecosystem. I think this has been truly a great savior in terms of
00:05:56
Speaker
our ability to democratize and customize technology and knowledge. Machine learning, which is a specific form of AI, has been going on for a number of years. But I think with the coming of the large language models and open AI, it is in some ways given a voice and a face to AI. So it's far more recognizable. Well, it used to be working in the background, but now we can actually have AI converse with you. It's a little bit more real in some ways. And it's also kind of interesting in terms of just the use cases which are exploding, how you can use that to fill in some of the gaps that we have or shortages that we have in labor.
00:06:35
Speaker
there is just not going to be enough trained labor to run buildings more efficiently. Can we deploy a technology to basically help augment the people who are running those buildings? And can we actually do the things that would normally take a very highly trained, skilled person and basically offload those over to AI? That's definitely where I believe ah the industry is progressing. That's one of the key things that 75F is really working hard on. like One of the things that I'm really excited about is that AI has allowed us massively to democratize some of this technology and some of this knowledge. right What it means is that it is no longer resident in a very few number of highly trained, highly skilled individuals, but it can actually be shipped and it can be deployed in buildings which cannot actually afford these highly skilled professionals.
00:07:29
Speaker
The other thing that i'm really super excited about is that this is come in fact much easier than you would have normally expected is the amount of the last language model which have been trained by people like open ai. What they're allowing us to do is transfer this knowledge across multiple languages. And that's really making these technologies far more accessible to geographies where they would not necessarily have been deployed before.
Adapting to Renewable Energy Sources
00:07:54
Speaker
So even in places like Middle East, you can actually have customization. You can have the AI or you can have an AI that allows you to converse with your building. So that's one of the key things that 75F has done is the generalized AI that understands HVAC, understands energy and allows you to basically interact with the building.
00:08:13
Speaker
So there's a digital twin which powers all of this, but more importantly, this digital twin communicates seamlessly across multiple geographies and multiple languages. and So it really is super helpful for all the non-English speaking areas. Some of these technology innovations had not necessarily percolated before. I just wanted to tease out a couple of things you said. One, AI is allowing technology to be deployed more broadly and globally. And secondly, AI is helping people use that technology with greater ease. Maybe you can comment on both of those and maybe that's a good transition point also a little bit into 75F and what you and your company do. and I think that's a really good observation that you have. One part is the optimization piece, which is the machine learning piece.
00:09:03
Speaker
And that is the algorithmic part. That's the more geeky part in terms of how do you use AI to transform and run buildings more efficiently and do this, which is in some ways could be called very large scale data crunching. And the second part is the more conversational, more accessible. How do you make that data to be more accessible? And that's where the large language models have really come in. It is very fascinating to see. what used to be geek speak can now be put into words that people can really understand so it makes it more real for them. Maybe a little bit on some of the advances you guys have made specially over the last five years at seventy five f and how you're using it to create a customer experience that is much less.
00:09:47
Speaker
We've been big proponents of AI for a long period of time. We did a clean sheet design of what we thought the buildings in the future are going to look like. We were already planning for this concept around that there's going to be far more renewable energy sources. So there's going to be more solar, there's going to be more wind. But one of the things that people don't necessarily talk or recognize is that both solar and wind
75F's IoT System and Conversational AI
00:10:09
Speaker
tend to be more fluctuating in terms of their overall capacity to serve. So when wind is low, your grid is not going to be generating as much power. If there is occlusion or cloud cover, what you actually end up seeing is that overall power generation on the solar side is reduced.
00:10:26
Speaker
Up till now we want to be sticking it for granted that the grid is always reliable and always has the capacity to meet our needs unfortunately that is not going to be the case unless we massively deploy storage and it is at the level which is almost going to be an economical so what we have to do is now have a live in a more adaptable world where. The load itself, which is like a very large part of the building, is the HVAC. Buildings are the fourth largest emitter of greenhouse gases and a very large part, about 67% of that is on the HVAC side. yeah There's a very large load that can actually be more mutable.
00:11:04
Speaker
to the variations in the grid generation. Now the question is, how do you interact with the grid and how do you do it real time? And that can only be done by machines. So that's one of the key things that we decided is that we're going to create an out of box, what's called an IoT based building management system. It required a rethink of the current paradigm that we have, which is our direct digital controls. and Those DDC controls are very manual. They're very cumbersome. They're too slow. They will not operate and in the buildings of the future. So ultimately, you have to do a complete paradigm change. So we did that and we did a redesign of how we thought the buildings are going to be run and how a building management system is going to be. And we thought of that architecture completely coming up with this paradigm of IoT-based BML. So we'd been doing it on the machine learning side.
00:11:49
Speaker
We'd been using it for optimization and making sure that the buildings are running more comfortably. How do you look at a building more holistically in taking into account the number of people inside the building? So we've been thinking about how buildings are going to be interacting with the grid. How do you take into account the number of people inside the building? And then how do you actually combine all of that with indoor air quality and predict how the building is going to behave and how do you optimize it for my energy perspective? On top of that, we're able to leverage the AI and what we were able to do is now inspect and talk to the building. So we have ah what's called a conversational AI. We launched that last year. And that's actually been very unique and it's still unique in the market because no other company has been able to do that. And the the reason they have not been able to do it because they did not have the digital DNA to begin with.
00:12:38
Speaker
They were not looking at yeah machine learning to go and optimize. We were, so we were able to leverage that and feed that to AI. So we could actually have a generalized AI model that allows you to go and converse with your building in multiple different languages at any
Hardware-Software Integration for Accurate AI Data
00:12:52
Speaker
time. So the construct that we are going towards, and now you can actually go and inspect your building. You can ask it to do things on your behalf, like change temperatures, but also you can actually ask it, why is the temperature in my room not Warm enough as an example so to actually go and do a diagnosis it'll actually go and do what's called an internal monologue to actually figure out and train the people asking this question on what that analysis is and come up with a reason to answer just like you would have an expert.
00:13:20
Speaker
HVAC technician go through this. As a company, we've been preparing for this future for a number of years. So now what we did is we could feed this data back over to our AI, both in terms of our own internal disk articles, in terms of how we've responded to calls, also in terms of how the technical documentation that we have on the product, and along with the real-time data, this makes for a really, really powerful symbiotic relationship that allows for a very highly trained the equal equivalent of a highly trained expert for each and every building. So you're using AI in two ways. One is using AI to optimize energy usage. That's the algorithms and the data crunching aspect of it. But you're really also using AI in a completely different way. I know you pioneered this term of a zero interface BMS. Do you see that as a trend and do you see that catching on?
00:14:15
Speaker
Ultimately, if we look at what people are really interested in is an outcome. So if you go back and take a look at like an outcome based system, how the delivery is done is going to vanish into the background. And so if we were to take it on from the outcome perspective, what we are expecting is a world where These buildings are going to be running by themselves. They're going to be running more optimally. And the only reason you would interact with them is when you have exceptions or you have a change in direction that you want them to follow. Talking a little bit more about 75F, one of the things that kind of strikes me as interesting. There's a lot of companies out there that are currently developing or have developed smart energy management systems. They seem to be taking
00:15:00
Speaker
Two very different approaches, which is there are solutions that are software only. And then there are solutions such as yours that is a combination of software and hardware. Why did you choose that latter route and what advantages do you think that that gives you? Excellent question, Akash. We didn't choose this path. We decided originally we were really looking at making something simple and the simplest way is to actually go and leverage just software, which is what the bulk of the companies do. and So you'll see there's a bunch of companies who are already coming in.
00:15:35
Speaker
But as we embarked on a journey, one of the things we really found was that, A, the sensing technologies that we have from what would we call incumbents, those are just not scalable. So most of them are really using analog sensors, and those are highly susceptible to signal degradation, they're susceptible to noise, and they need to be recalibrated over time, what's called drift. What that really means is that if you're using AI, and if your thesis is that AI is the one that is actually going to be making these smart decisions, the more noisier your signal, The worst your outcomes are going to be so ultimately an underperforming sensor is always going to lead you to garbage in garbage out and this is specifically true in AI systems where the system itself becomes habituated and trained on the garbage that you're actually giving it. The first order of problem was can we make sure that the integrity of what we are measuring is very fiber so that is one of the reasons we choose to do this hard part of making sure that from the sensing technology that's ours.
00:16:35
Speaker
The second thing that we saw is issues is that the current paradigm that we have with what's called direct digital controls, they're highly dependent on the programming which is done by the people. So the actual algorithms and sequences of operations are really implemented by folks in the field. And those people typically will try to do the easiest, simplest way of getting the job done rather than what would be the most optimal way of running it. Another issue that exists with existing controllers was that once they've done the sequence of operation there's no way of updating remotely it requires on site visits that's highly expensive so we had really had to make a control system and a controller around a paradigm of what we call software defined hardware.
00:17:16
Speaker
And the software to-defined hardware necessitated us making our own hardware, which could be more amenable to software over-the-air updates. And that's just like your Tesla gets smarter. So now we make our own sensors that measure things like temperature, humidity, volatile organic compounds, CO2 levels, as you can see. And then we pair it with the software defined hardware controller. So think of it like your iPhone. You can use it to play video games. You can use it to browse the net. You can even use it to make phone calls. But ultimately, it's one single piece of hardware that takes on different personalities. So that's part of this part and parcel of this construct around an IoT based BMS that we had to really from a ground up construct.
00:17:54
Speaker
And that has given us a huge competitive advantage because it allows us to do these refresh cycles and make the technology better as we learn more things as the AI gets smarter. Ultimately, the AI does no good if the data was measured incorrectly in the first instance.
AI and Mass Customization in Building Automation
00:18:11
Speaker
How is AI ultimately impacting and going to impact some of the decarbonization goals and really kind of help mitigate some of the impacts from climate change that we've seen. AI is a big democratizer of technology. It allows the proliferation of what would have been considered expert knowledge and now it's made more accessible to people who would not necessarily need to be experts in that realm.
00:18:38
Speaker
That's one aspect. The second aspect of it is that AI allows for mass customization. One of the interesting things we've heard it in our industry is that every building is a snowflake. Why are some of these buildings not automated? It's primarily because it is too expensive, it's too cumbersome, and every single building is different than the next one. When you have this snowflake paradigm, what really means is that it requires a huge amount of customization going from building to building. Which is exactly what the current systems have been created for, is to allow for this customization, but it doesn't allow for mass customization. AI, once it's properly trained and when you have a generalized AI for a very specific vertical, what that allows us to do is that AI now has a very good depth of understanding and it can actually take into account and generate new sequences which are novel.
00:19:30
Speaker
It is not necessarily the same thing regurgitated again and again. And that's, I think, the fascinating part around AI is that you could actually have this mass customization occur for a large swath of the buildings, which not necessarily would not have actually normally benefited. It would have been cost prohibitive for humans to go in and deploy these solutions at scale. Got it. How do you see these AI-based BMS systems interacting with smart grid optimization? Do you see those two worlds ultimately coming together? I think it's inevitable. So if your building is smart, if you have an AI which is going to be controlling a building and that does require that you have the right infrastructure, that you have the right kind of BMS like 75F in there. So you will actually have an AI which is at the utility level, which is responsible for a macro
00:20:21
Speaker
overall all redistribution and asking for load shedding as an example. And it would go and interact with the AI, which is sitting in the building. And that AI would then be responsible for actually acting on that action. So that piece, I believe that's going to be a really important part of the future about all of these AI's talking to each other and working and acting on our behalf. Like adaptive demand response. It's not just adaptive demand response is not just that the utility sends a signal telling the building as an example when to shed off load, but more importantly that the building is telling the grid back how are the conditions changing within for the occupants inside.
00:21:01
Speaker
What that really means is that allows the grid to be more sensitive. One of the key things that we see right now is with what's called the unidirectional demand response. The utility will actually send a command over to the building and say that you need to cut down demand for a certain number of hours. But the problem is that the utility has no understanding of what in discomfort it's actually causing over to the occupants inside the building. So what we're seeing is in some areas like California where they've been having DR signals for some period of time, people have started opting out of DR.
00:21:33
Speaker
Because what you're doing is you're causing more discomfort to the people during what would normally be already the hottest or the warmest parts of the year. So the utilities have to be cognizant of the discomfort and what's going on and in their actual customer sites. And that's one of the key things that we're working with some of the utilities to enable. Reflecting on this conversation deep and all of all the comments you've made, The AI trends you mentioned and the democratization, is it fair to say that there is a very high likelihood that these systems will be ubiquitous, not just in the large buildings where they are today, but across the building spectrum, the commercial building spectrum in particular, and potentially in dezaa other building verticals like industrial and maybe even residential? I think they're going to be more ubiquitous in the smaller mid-sized buildings than they are in the larger buildings.
00:22:25
Speaker
I believe mid-sized buildings are actually more geared up to benefit from this in the future. Of course, there are some very large building like folks in Saudi Arabia who are doing places like the Neom where there is no way that a building like that is going to be ever comprehended by a human being. to truly run that building so you will need an ai for that piece but i believe that the value proposition might actually be stronger in the near term for the mid-sized buildings and you're absolutely right i think it's gonna be absolutely ubiquitous i see no other alternate i see no other future to the.
Future Vision: AI in Building Management and Autonomous Vehicles
00:22:57
Speaker
Thank you, Deep. This has been incredibly helpful, and it's really good to get your perspectives given you're at the cutting edge of this. Any final closing remarks from your side? If you imagine a world in the future where you are going to have a self-driving car, if you're going in a self-driving car, can you imagine that your buildings are going to be run the way they are? Of course not. The entire world is going to be a very different. We should take that for granted. The buildings are going to be smarter. They're going to be run by AI. There is no other way. Thank you Deep. It's really good to speak with you ah every time and really appreciate you making the time this morning.