Energy Transition Talks

From models to meshes: Building trust, balance and resilience in AI for energy and utilities

• CGI in Energy & Utilities • Season 4 • Episode 4

Share episode

Send us a text

Energy & utilities leaders: move beyond AI pilots. CGI’s Peter Warren and Frédéric Miskawi show how small, quantized on-device models, edge robotics, and digital triplets drive real-time decisions on rigs and grids—while cutting cost and carbon. Get a pragmatic playbook: multi-model architectures, “healthy paranoia” against bias, and continuous data governance to fight digital entropy and build trustworthy, resilient AI. 

Visit our Energy Transition Talks page

Peter Warren:

Hey everyone, welcome back to part two of our series uh talking about AI to ROI. This is part of our ongoing energy and transition talks uh here at uh CGI. Uh my guest today uh from part one and part two is Fred. I'll let you reintroduce yourself again. Fred.

Frederic Miskawi:

Hi, Peter. Fred Miskewi. I'm part of our global AI enablement team. I lead our AI innovation expert services globally, and that uh gives me the the deep luck of being able to work across geographies, across the world, across teams with clients, different industries as well. I've been involved with our artificial intelligence in one way or another since the 1990s.

Peter Warren:

Everybody thinks uh AI may have started right now, but of course this industry has been using machine learning for a long time. So we understand the the early days of it, and CGI has been part of that as well. In part one, we covered off a few things about is my data good enough? We we touched base on you know the use of the best algorithm for the job uh idea and concept. Don't don't necessarily use a sledgehammer to uh put in a uh put in a screw uh type idea about KPIs. Uh today we're gonna hit into a couple of interesting concepts. Uh the shift between everything needing to be a large language model, like one of the big superscale uh multi-scalers or uh hyperscaler systems versus small language uh models. When do you go to a small language model? When do you determine uh do things that are more deterministic and more control-based? How do you manage that uh decision process and what fits what?

Frederic Miskawi:

Great question, Peter. And that goes back to what I was mentioning in part one, which is the best algorithm for the job. We're talking about a multi-model ecosystem, multi-agent ecosystem. And what we're seeing and what we've seen evolved very quickly organically, is the cost and energy usage associated with these large language models can be quite substantial. So, how do you manage budgets? How do you manage the overall cost of the solution? How do you manage the also the response time? Uh that can performance can also influence what kind of algorithm you're leveraging. So we're seeing this shift towards smaller models, on-premise models, working with uh uh on-cloud or hyperscaler models. Uh, we're seeing quantization happen with those models so that you can get them as small as needed to work on a device. Um, even six months ago, the on-device models were not necessarily great, but you're seeing that evolve uh very quickly. You see the capabilities evolve very quickly to the point where now you've got Nvidia coming out with things like Jetson Jetson Thor, uh which will be powering a new generation of walking uh models, which are these uh bipedal robots that are coming out in the next 12 to 24 months.

Peter Warren:

Yeah, we were we just saw the uh I guess it was in Beijing they had the robot uh Olympics uh for the first time, and uh it was kind of a mixed uh mix of things. But uh I suppose that's a very dramatic example of edge computing. I mean, our industry, uh both everything from oil rigs to uh uh energy production right through to smart grids uses a lot of uh edge computing type of technology, not all of it being, you know, major computer systems, some of it even being a little legacy. How do you see those types of computer systems evolving as we move forward?

Frederic Miskawi:

Um you're gonna see an evolution of that ecosystem. We're already seeing it in the algorithms. Um, you're gonna have a wide variety of systems that are powering our enterprise, powering our networks, powering our various uh assets across the company. Uh, we're seeing it, even for us as consulting firms, we're starting to see that evolution occur. Um, as we're talking about the energy and utility industry, you're gonna see this deployment of bipedal robots increasingly increasingly powerful and capable. Um, what we're seeing in the lab today are robots that walk and talk like you and I, uh very fluid, able to do martial arts or able to dance, able to move very fine objects and operate equipment in a way that is uh a lot more deterministic than it was in the past. That level of ecosystem evolution is what we're seeing happen today. Um, and what you're not seeing, what's in the labs today that we get glimpses uh of with the work that we do is going to truly revolutionize the industry and the way that we operate. Uh you're gonna have very dangerous solutions and jobs that are handled increasingly by teams of humans and robots. And where if a if a robot gets crushed, uh you're gonna have a lot less uh heartaches than if a human co-equipe um co-worker getting hurt. So you're gonna see just by simple need these this evolution occur of new algorithms, new ways of of running these algorithms in our analog world occur and develop over the next two, two, three years.

Peter Warren:

The mining industry has been a big adopter of uh robotics and self-driving vehicles. Um this industry has been a big adopter of the quadrupeds, so the four-legged uh robots, um the robot dogs, as they're sometimes called. We see those in harsh environments. Um, even just looking at edge computing, let's say a static device, something on the edge, making a decision about do I turn this electricity on, do I open that dam, what do I do, that type of edge computing. Um, and you made an interesting comment. We were saying that some of these new models will even work on your very old Mac. Uh it's not that people need the latest and greatest uh NVIDIA technology in every case. How do you how do you see people moving forward with some of these smaller systems, maybe on some more affordable platforms?

Frederic Miskawi:

Yeah, and that that's uh I was mentioning earlier about quantized models, the ability to take a small model to begin with, and then kind of uh streamline it, filter it, remove some of the underlying parameters in order to get it as small as possible to run on a smaller, weaker device where we can run some of these more powerful models, even on CPUs. It might be a little slower, but it it works. Um, with that technology, you're always gonna be dealing with a statistical model. So you've got to be able to work with these smaller quantized models running on these older hardware. Um, and you've got to work with them in layering to make sure that you get a little bit more deterministic behavior out of it. And that's where agents come in. So if you have a very small model that can run on device, that is really the brain of something that's a little bit more deterministic in the in the body of what we call an AI agent, then you have the ability to run decisions, binary decisions, more complicated categorization on device. Uh so you've got very targeted needs. And for very targeted needs, you can do that on older machinery. And and that means that you don't have to wait. You don't have to capsulize a massive digital transformation in order to get the benefit of the technology.

Peter Warren:

And you've talked about layering, and I think layering, you know, you've explained the very s smaller, the edge, and that's going to probably continue to expand and improve, as you mentioned. Um if you start and you start to look in things, you referred to in uh a concept about enterprise neural meshes and the use of a digital twin and actually stacking those and making a digital triplet, which is a concept that uh uh Diane Gushu and yourself have brought forward and talked about quite a bit. That really is the layering. And when I explained that to a couple of executives, they've said, well, finally I see a value to me for AI because it's actually helping me versus maybe the people in the field 100% or the different layers. Do you want to explain that layering really right from the edge, right to uh helping the executive decide what to do?

Frederic Miskawi:

Yeah, our our vision and what we're working towards, at least internally from a client zero story perspective, is the enterprise neural mesh. It's the near real-time view of the enterprise. And we've been seeing inklings of that over time. But the idea is to be able to answer any and all questions that executives might have, stakeholders, investors, analysts on what's happening within a given uh enterprise. So for us, what that means is, for example, 21, 22%, I think, of our revenue comes from IP. It's to be able to see and have a visibility of the value that's being delivered, uh, how many people are working on it, what kind of value is being delivered, the quality that comes out, and see it in a near real-time basis. These models and this decentralization of intelligence, um, both on legacy hardware as well as new hardware, as well as upcoming bipedal robots that will become not breathing but moving data collection engines, we're feeding all that into this digital triplet that gives you a view and understanding of the enterprise so that you can make decisions. And also, most importantly, you can start running uh simulations where when you have that level and type of data collection and the layering that comes with it, you can start looking at, well, if I were to make this decision, what would occur with the enterprise? What would be the impact of that particular strategic decision? That's what we're seeing evolve. The technology enables you to do that, it's already there. And now it's becoming more of a human transformation story. The technology is moving and has moved beyond our ability to absorb it. Uh, that's what I see day in and day out. Um, we're just working from an organizational change management perspective with teams, individuals, clients, client teams, building that capability understanding of the technology so that we can absorb features, functions that were released several months ago, several maybe even years ago. So that's what we're seeing right now. It's that that human evolution of understanding of this technology, the absorption of the technology to work towards a enterprise neural mesh, a real-time, near-real-time view of the enterprise.

Peter Warren:

Yeah, it's interesting. You mentioned customer zero. So for those that don't know what we're referring to there, that's we're we're doing this to ourselves. Uh we're actually modifying how we operate internally and how we run. Um but simultaneously, as you mentioned, we're working with uh clients that are sort of first movers in those areas and touching base on it. Um wrap this up, where do you think, you know, if you used your crystal ball, uh we see a bunch of both good things and bad things today in the news? There was talking about disinformation uh from certain websites, uh specifically coming from Russia, trying to train train large language models, more the public ones on stuff. Um, you know, that's probably propaganda, that's uh their point of view. Uh how do you sort of manage all of these things as you move forward? When do you, you know, even your personal life, how do you manage AI and how do you see companies managing this as they go forward for, again, data quality and getting the right outcome to do the right action?

Frederic Miskawi:

Yeah, in two words, healthy paranoia. And and funnily enough, I had a similar conversation with with my oldest son this morning on healthy paranoia. These solutions are amplifiers, they accelerate access to knowledge information, whether that information is accurate or not, whether that information is intentionally inaccurate. And with healthy paranoia, you're building a filter. You're building filtering layers between yourself and this technology and potential actors that want to effectively brainwash you. And with healthy paranoia and understanding that this technology can be used, can be and abused for amplification. Uh, this technology is not necessarily accurate either, not always. It will get better, of course. But right now, what we're seeing is that we are dealing with statistical engines that may or may not always tell you facts. And with healthy paranoia, you can start asking questions, validating information, double checking, having multiple sources of information. So we're even doing that in our solutioning. We're bringing in multiple models, multiple hyperscaler models, as well as on-premise models within the confines of the same solution, again, to help with deterministic behavior and more accuracy, uh, dealing with biases and ensuring transparency. So we're going to continue to see these news stories come out, and this technology will be used to manipulate, will be used to steer uh groups of humans toward a particular uh realization. And it's up to all of us individually to realize that this technology is powerful. This technology has to be questioned, and we have to build healthy paranoia as we move forward.

Peter Warren:

Yeah, no, it's a good idea. Even when you're using uh just the data within, and we'll we'll wrap up here in a second talking about going back to the first question about data, data quality, and governance is that, you know, um in organizations uh maintaining the data, you said you did in the first part, we we didn't have to start with pristine data, but it's continuing to evolve. So that requires a bit of change management in the organization. So we see the companies that are being most successful with this type of technology are being very agile in the way they work, are being restructuring things, managing the use of this, they're responding to the data. Uh, but how do they go forward on maintaining this forest of data that is continually growing and uh getting weeds? How do they how do they deal with that on a day-to-day basis?

Frederic Miskawi:

Well, I think number one is embrace the idea that there is such a thing as digital entropy. So with digital entropy, the idea is that over time your data will continue to reduce the level of accuracy that it has, the level of usefulness that it has, and to the point where that data may actually be counterproductive to your business goals. So when you do that and you have that healthy paranoia with your systems and digital entropy, you're putting in layering in place to make sure that this data is being catered to in a more automated fashion. Uh, what we've been seeing over time through kind of historical anthropology of that data is that it gets old, it gets stored, it gets uh layered, uh, it may be abused, it may be reused, and all that were you know following human processes. Now, these systems need uh not just quality data, they need data. And the more data, the better. And they can infer patterns based on the data that gets ingested, but you want to be able to cater to that data to make sure that you have the ability to collect the data, you have the ability to cater to the quality of the data in case there is known sources of data that is uh not conducive to your business goals, to eliminate the data, archive it if needed. There's these processes, this layering that you put in place, all is there to manage digital entropy. So if you know and understand that there are natural organic slash digital processes in place, um, that's gonna make you understand that you have to have healthy paranoia and that uh you have to put these these solutions and layering in place to manage the ecosystem.

Peter Warren:

I think that's a great spot to stop, uh, Fred. Uh thank you very much for today's uh conversation in both part one and part two. Uh thank you everyone for listening, and uh, we will see you again in our podcast series uh on the ever evolving energy transition. Thanks very much. Bye bye.

Frederic Miskawi:

Thank you, everyone.

Head Shot

Peter Warren

Host
Head Shot

Peter Warren

Host

Podcasts we love

Check out these other fine podcasts recommended by us, not an algorithm.

Energy Transition Talks Artwork

Energy Transition Talks

CGI in Energy & Utilities