Artificial intelligence and machine learning are already here. But there are barriers preventing them from being fully exploited
You could be forgiven for reading the headlines and thinking that artificial intelligence (AI) and machine learning (ML) are still part of a future-facing set of technologies that are not yet having much impact. Or that, alternatively, we are on the verge of a jobless world where AI/ML will replace humans in a wide range of industrial and professional tasks.
But neither of these are true. In fact, AI/ML is already positively impacting many industries for the better and pushing forward significant advancements in areas as diverse as healthcare, finance, online services and agriculture.
‘The first nation to successfully digitise its built infrastructure, and thereby generate the data suitable for AI/ML, will reap huge benefits in improved infrastructure provision, in better public services, and in generating whole new areas of economic activity and enterprise’
This includes construction. AI and ML are rooted in data – of which there is more available now than ever before. More data can generate more intelligence; it means we now have the potential to truly see what’s going on within a piece of infrastructure. For example: how is a building really operating? How are its structural components doing over time? What are its real-time energy, lighting and water uses? How do they get to and from the building? How do people actually use the building in practice, and what do they really want? And what’s exciting is the notion of connecting these data and intelligences and linking them together – within a smart cities context, for instance. This bigger picture vision, based on new, technology-driven insights and opportunities, brings the potential to redefine the built environment as we know it.
So, what’s stopping this vision becoming reality? I recently attended the All Parliamentary Party Group on AI, where I outlined what needs to be done if we are to fully realise the potential benefits of AI/ML in construction and infrastructure. For example,
- Ensuring the nationwide deployment of the advanced broadband and mobile infrastructure that underpins digital and AI systems
- Improving significantly the availability of data science skills in the UK
- Securing widespread and cheaper access to the advanced computing facilities that are needed for training and implementing AI models
- Improving the understanding of AI at a practical level in government, regulators and the wider public sector
But the biggest barrier is the fact that most of the UK’s built infrastructure is still far from digitised, while data is largely fragmented and of poor quality, so not in a fit state for mining, analysis and insight-generation. The first nation to successfully digitise its built infrastructure, and thereby generate the data suitable for AI/ML, will reap huge benefits in improved infrastructure provision, in better public services, and in generating whole new areas of economic activity and enterprise. So for me this is a top priority for government and industry investment and collaboration.
Meanwhile, at Arup we are taking every opportunity to use ML in our work.
Take the oil and gas industry as an example, where older infrastructure faces increased risk in terms of spillage, accidents and loss of product. Using ML on satellite imagery, we are able to scan areas in real-time and identify not only hazards but thefts, cross-referencing them with the refineries’ locations and activities, to determine the best action and management of the facilities.
We are also using a similar approach to reduce flood risk: applying satellite data to acquire accurate information on small patches of land quickly, before ML, to reveal subtle changes in land use. This means rapid scanning and intervention in these areas, to enable smaller-scale natural interventions to reduce flood risk without the need to invest in significant projects like dam construction.
Elsewhere, we have used ML to train a computer to identify and categorise tunnel cracks. This means we can move away from the traditional approach to detecting tunnel damage, which involves sending staff into tunnels to check and log issues – a method that’s not only time consuming and costly, but also ripe with risk to human safety.
These examples are just the start. In the next 3-5 years, progress will be such that entire new areas of cities and infrastructure could be built with data and digital technology embedded at the centre. For now, focusing on the continued development of tangible examples, while systematically addressing the challenges will build confidence and capability ahead of technology innovations being implemented at scale.
Will Cavendish, Global Digital Services Leader, Arup