Raising the roof with artificial intelligence
Artificial intelligence (AI) is reshaping our lives. From the home to the workplace, technologies are enabling computers to perceive, learn, reason, and assist in decision-making to solve problems in ways that are similar to humans. Already we’re seeing changes to the way we communicate, work and innovate. And the impact is being felt. Whether it’s personalised shopping recommendations on the busiest of high streets or your smartphone telling you the best way to get home before you set off, AI is becoming more visible than ever before.
The emergence of new technologies has often given rise to disruption – such as witnessed with the arrival of steam power, the harnessing of electricity or the arrival of personal computing. But it also gives rise to opportunity. The emergence of AI is creating similar such opportunities. Microsoft, in conjunction with Goldsmiths, University of London, recently conducted a study to benchmark where UK organisations are on their own AI journeys and the impact that intelligent technologies were having on business performance across the UK.
We found those organisations already embracing AI are outperforming those that aren’t by 5 per cent. A significant boost to an organisation’s bottom line few can afford to miss out on. AI is here and it’s creating value for organisations of all shapes and sizes. But this is just the start of a long journey and the value it creates will increase significantly and rapidly.
We are delighted to partner with Ordnance Survey (OS) to explore how AI can transform some of its processes. One project has been to examine how the company classifies UK roof types using geospatial data. By spending just a few short days working together, we found by using AI, the determination of roof types be completed in a shorter timeframe as it could to analyse multiple roofs at the same time.
The key point is not AI can do the job of a surveyor, rather how the AI frees up the surveyor to focus on higher value tasks. After all, few people spend years studying to be a qualified surveyor with the ambition to sit in front of a screen identifying different roof types.
Analysing from the skies
So why does OS spend so much time identifying these roof types anyway? It's a critical and extensive undertaking. Securing information about the structure of UK buildings is crucial for purposes including insurance, disaster preparedness, urban planning and even poverty alleviation. But this work is incredibly time intensive and OS needed to find a way to accelerate this process.
Microsoft and OS joined forces for a week-long ‘hackathon’ to build a proof-of-concept system. This programme first categorised roof types. Once it could do this, we then trained the machine learning algorithm to extract these features from photographic data. At the core of the machine learning model, the teams used a transfer learning approach to "featurise" the roof images. This meant starting with an original model which was trained on a very large number of generic images, and then to use a small set of images from unique datasets to fine-tune this network for the most accurate of results across roof types.
Making work more human, through machines
Upon the workshop’s completion, the positive results were immediately clear. The team found that in the time it took for a human to classify all the roof types in a single image, a machine could process thousands. Indeed, it is now evident it would take the trained algorithm less than a day to classify all 35,700,655 properties in Great Britain and their roof types.
This enhanced level of detail could be used to inform insurance companies, helping them to produce more accurate risk assessments to ensure people have insurance policies that truly meet their needs. Equally as important, this project means OS employees are no longer required to manually input clarifications into a spreadsheet, freeing up their time to focus on more complex work. By using their skillsets in different ways, there is a clear and tangible benefit to OS.
Embracing AI now for the longer term
As this technology evolves, more and more examples continue to emerge of how AI can benefit UK organisations, their employees and broader society. Our work with OS is just one case, creating efficiencies and enabling employees to focus on higher-value tasks that are more beneficial to the firm’s own customers. Yet this is just the start of a bold journey, one which we look forward to embarking on with Ordnance Survey. We have the technology, they have the location data. Let’s go!