In a resource-constrained world, solutions for modelling and managing assets can be a boon to society and help to reduce our outsized carbon footprint. Many of today’s bright minds dedicate considerable time and effort to resolving the conundrum of balancing economic growth and community well-being against negative impacts on the environment. And over the past decades, they’ve turned to a powerful ally, artificial intelligence (AI), to aid this quest.
Imagine, for a moment, an urban environment, where different systems communicate with each other to achieve better outcomes. Rather than conforming to regular schedules, city buses would run where and when they’re needed, for example. Energy would be distributed not only based on demand – but also in consideration of the environmental impact, therefore maximizing clean power sources and enabling efficiency.
For Yoshua Bengio, professor in the Department of Computer Science and Operations Research at Université de Montréal and scientific director of both Mila, the Quebec AI Institute, and IVADO, an AI research and transfer institute, such a vision is less utopian than it may sound.
“AI is already contributing to a range of efforts that can help us tackle some of the most urgent problems of today, climate change among them,” he says. “We have systems that allow us to use energy more efficiently, whether that’s by controlling the temperature in buildings, managing the grid or predicting both the demand and availability of energy.”
Enabling the successful application of AI in the built environment is an abundance of readily available data, for example, from the operating systems in commercial buildings, says Jean-Simon Venne, co-founder and CTO of BrainBox AI, a provider of fully autonomous solutions for reducing energy use and carbon emissions in buildings.
“Most commercial buildings already have control systems in place that operate heating, cooling and ventilation [HVAC] in real time,” he says. “Over the past 30 or 40 years, building owners and operators have made significant investments into these digital control systems. However, most of them are still reactive, like the thermostat in your living room, that responds to higher or lower temperatures.”
Integrating the thermostats and other sensors in a building can present an opportunity to collect data on energy use – and this can then be leveraged to create a predictive model, powered by AI. “The challenge is to find a way to connect these control systems, extract the data, bring it to the cloud – and leverage it for deep learning,” Mr. Venne says. “These deep learning and neural networks can use the data to generate predictions that give us a picture of what happens in the spaces of a building over time.
“We basically forsee the future,” he adds. “We then apply operational research techniques to optimize the control strategies to change future outcomes.”
A better future, in regard to the built environment, means offering better comfort and spending less energy, hence reducing both energy cost and emissions, says Mr. Venne. “Depending on the building’s location, there may be times when energy is available from green sources or when prices go up during peak energy use. Factoring in these parameters could help guide more efficient energy use and offer some dynamic flexibility on the electrical grid distribution side.”
Yoshua Bengio Professor in the Department of Computer Science and Operations Research at Université de Montréal
" AI is already contributing to a range of efforts that can help us tackle some of the most urgent problems of today, climate change among them. We have systems that allow us to use energy more efficiently, whether that’s by controlling the temperature in buildings, managing the grid or predicting both the demand and availability of energy.
Commercial buildings represent a significant source of “untapped potential” – and residential buildings are also increasingly generating data through digital technology implementation. He suggests that “connecting them – and translating the different control languages into a language that is understandable by deep learning” – could help to make the whole system more resilient.
Deep learning isn’t just the capability to extract insights from a data set – it allows for the continuous discovery of new things. An AI building system will continue to learn through the seasons, not only from weather conditions but from the humans using the building, Mr. Venne explains. “The beauty of a neural network is that it keeps learning as it’s being exposed to more data. It’s not a model that is cast once and then it’s done.”
Learning from the human mind
While this continual learning – paired with the capacity to process vast quantities of data – is certainly a significant advantage of artificial neural networks, Dr. Bengio is especially interested in applying insights from neuroscience and cognitive science to making them perform even better.
“By looking at how computation is performed in the brain – and what role neurons play in this computation – we can see how humans achieve abilities beyond current deep learning,” he says. “One of the things we found is that human intelligence is much more adaptive compared to AI.
“For example, if you’ve been driving in Canada, and then you rent a car in England, you would probably be able to drive without an accident. Current AI would not only be prone to have accidents in such a circumstance – it would likely take thousands of accidents before it would learn to adapt.”
When navigating a foreign system, like driving in the streets in London, humans don’t rely only on habit, explains to Dr. Bengio. “You would use conscious attention to deliberately drive on the left, and constantly remind yourself of potential consequences. This attention and ability to reason is connected to conscious thought processing, which has been studied extensively by neuroscientists and cognitive scientists.”
This reasoning ability, which allows humans to consciously and sequentially process thought and then abstract insights, is something Dr. Bengio regards as an important part of the explanation of the current gap between human and artificial intelligence.
He hopes that efforts to reach a similar level of probabilistic inference in artificial neural networks will “guide the next wave of AI – and get us closer to human intelligence and some of its advantages in terms of robustness to changes in our environment.”
Jean-Simon Venne Co-founder and CTO of BrainBox AI
" Depending on the building’s location, there may be times when energy is available from green sources or when prices go up during peak energy use. Factoring in these parameters could help guide more efficient energy use and offer some dynamic flexibility on the electrical grid distribution side.
Creating impact for society
As computer scientists work to apply lessons from human thought processes to artificial neural networks – to enhance their reasoning abilities, Dr. Bengio suggests they team up with a range of multidisciplinary collaborators to explore the question, “How can we help society in a deep way?” He also urges funding agencies, including governments, to consider the potential long-term benefits that can be gained from these technological advances.
“Scientific discoveries don’t happen in a vacuum – they build on previous efforts and require support,” he says. “And AI has strong roots in Canada.”
AI is already making a difference in many fields; for example, by powering search engines for text, voice and images, or to translate text in web pages or your smart phone. Educational applications can keep track of learning outcomes and allow students to progress through course materials faster.
In health care, diagnostics have been improved through AI-assisted analytics of medical images. Models of antibiotic molecules, their use and drug resistance are allowing the discovery of new antibiotics and more targeted medical interventions, says Dr. Bengio. “AI is going to transform how we understand our body, and how we find cures that are much more personalized.”
Deep learning systems have been deployed to analyze satellite images to not only detect catastrophes due to extreme weather but also monitor some of the slow changes that come from a shifting climate, he explains. And our ability to create climate models, which require significant computational power, has also been enhanced by artificial neural nets to obtain faster simulations.
Reducing the climate impact of human settlements also goes beyond improving the sustainability of buildings, says Mr. Venne. As the power grid is changing from a linear “generation-to-consumption” pathway to a model that includes components like renewable and traditional energy generation, energy storage and feedback loops, long-distance and local distribution, and consumption that is subject to consumer behaviour and changing societal conditions, the question becomes how to balance such a system in real time.
For example, “the growing adoption of electrical vehicles, including school buses and delivery trucks, will bring substantial additional demand on the existing electricity grid,” and Mr. Venne says this complexity will require “AI agents located at the building level to start talking to each other – and start collaborating.
“When we look at a world where everybody has the goal to reach net zero, we need to consider all the parts in the equation, including how power grids can accommodate more electric vehicles,” he says. “We need to do a lot more to ensure our systems are capable of meeting these ambitious objectives.”
While ambition is essential for saving our planet, AI can help along the way, Mr. Venne suggests. “The tools are already there. The potential is enormous, and the field is moving fast.”
Imagine what can be accomplished when the best of human and artificial intelligence work together to support the vision of Dr. Bengio and Mr. Venne: to create a brighter future.
Advertising feature produced by Randall Anthony Communications. The Globe’s editorial department was not involved.