As CTO, one thing I’m frequently asked about is ChatGPT and what its impact could be on building management. The short answer is that AI in general could be transformative if applied to a building that has been connected to a modern data platform.
ChatGPT is a type of intelligence known as a Large Language Model (LLM), which, as the name suggests, displays an intrinsic understanding of a given language, be it a traditional human language or a computer language used to express software source code. LLMs can generate plausible-looking text in response to a question, which may include contextual information such as previous responses. They make ideal chatbots. Further, if you ask them to write a software script to answer a question about a dataset, and then run that code on the data, it’s likely that the answer will be of good quality – in fact, there are several ChatGPT plugins that will turn it into an AI data scientist.
How could all this help manage a building better? At learnd, our Remote Operations Center (ROC) provides customers four benefits: added security, reduced cost, increased visibility of their estate, and knowledgeable technical support. It’s easy to see how technical support could be enhanced with a ChatGPT chatbot – the data we collect from buildings could be queried by a ChatGPT-written script prompted by a customer question. Instead of waiting for a minute on the phone while a ROC operative investigates, you could have the answer in seconds.
That’s just the beginning
A few years ago, DeepMind demonstrated a whopping 40% reduction in cooling energy usage in already well-managed datacenters via the use of cutting-edge AI. This required specialists to research and design AI models, then perform some trial-and-error to produce these impressing savings. Instead, we can envisage a future where a non-specialist user asks ChatGPT to write a custom model for a given building, based on facts that ChatGPT has inferred about its energy usage, and a normalised data structure. This model would then be continuously tweaked over time to improve its results, possibly drawing on ideas and strategies from other buildings it observes. It may never save quite as much as a hand-designed human model, but the approach could be scaled much faster to a large estate of buildings – a greater impact for far less cost, and a network effect demonstrating the power of many buildings to reduce carbon emissions, fast.
The steps we are taking with our data platform today put us in prime position for this future. ChatGPT and its AI friends may seem smart, but what they cannot do is work in the absence of well-structured, granular data – something in short supply in buildings. A fragmented ecosystem of vendors and proprietary protocols mean that retrofitting any estate-wide software solution, let alone implementing AI, is likely to be a costly, monumental undertaking.
By combining our unique know-how of different BMS vendors with a standardised data model, we are creating a platform that can bridge this gap time-effectively without requiring expensive equipment. We’re talking to AI specialists who know how to make the models but don’t have access to buildings data, so what currently may seem out of reach could be coming to a building near you very soon – watch this space.
If you’re interested in finding out what steps you can take to be ready for this revolution, feel free to contact us today
