AI for good
We founded Tractable as a computer vision company in 2014 with a focus on finding real-world use cases for AI that will have a uniquely positive impact. When you apply AI to real events and go beyond proof to concept to create a robust and trustworthy platform, the results can be transformative for industries, people and planet.
Most people’s direct experience of AI comes through everyday technologies such as virtual assistants on your phone (like Alexa and Siri), automated investing, product recommendations on sites such as Amazon and many more. Other experiences of AI tend to be indirect, such as advertising algorithms that are constantly optimizing the ads you are served.
But one area where AI genuinely excels while also demonstrating tangible value as a force for good in our daily lives is analyzing images, just like a human would. For example, scanning medical images for evidence of disease, or pictures of cars to understand their condition. In fact, assessing car damage has been Tractable’s commercial focus for most of its life so far.
Training our products
Training an AI to carry out a task like this is complex, and the key to creating a solution that adds value is to train your algorithms on very large numbers of historical cases. With cars, you can teach your AI by using hundreds of millions of images, with input from expert humans to correct the algorithm when it needs to improve. Helped by most cars being similar, in time the AI learns to recognize what damage looks like, and when combined with other sources of data, it can estimate how much repairs will cost.
Homes provide a more complex challenge as they are less uniform, but with enough data, it’s still one that is solvable – which is why our position as a partner to insurers is so important. In Japan, our insurance customers represent 90% of the car and property insurance markets. By working with them and by collecting, digitizing and pooling their historical claims, we have been able to produce an enormous training dataset for our AI, helping it get to grips with understanding damage to houses, as well as cars.
AI applied to property and beyond
We’re now launching our Property Estimator solution in North America, and then in Europe. However, insurance is just the beginning. If we can use your phone to scan your car, we can also use AI to help you protect it (by telling you when, for instance, its tires need replacing) or sell it (by valuing it accurately via photos). We can carry out similar tasks for the home too. If we know that your home is in an area that may be affected by wildfires, an AI could tell you, after analyzing photos, that you clear the nearby bushes that could act as tinder and set your fence ablaze. Or in damper areas, it could notice that your drains need some care and attention before you do.
Climate change will only make the preventative role of AI more valuable. According to a 2021 report by the Canadian Institute for Climate Choice, “As sea levels rise and rainfall increases, flood damage to homes and buildings could increase fivefold in the next few decades and by a factor of ten by the end of the century, with costs as high as $13.6 billion annually.” Sadly, this story is global: the 2020 California wildfires are estimated to have destroyed over $10 billion worth of property. Aside from climate-influenced disasters, there are uniquely natural ones. At the time of writing, we just witnessed devastating flooding in Tonga.