Artificial Intelligence has been around for decades, but over the last few years it has become an indispensable part of any company strategy. Why is now the right time for insurers to double down on artificial intelligence? What are the drivers and enablers that can’t be ignored, and how can insurers distinguish reality from hype?
There are many reasons why artificial intelligence is becoming a reality in insurance, but let’s distill them down into the three most important drivers.
The Explosion of New Data: machine learning, an approach to realize AI, in a nutshell is a system that uses examples to make predictions. ‘Examples’ are captured as data and stored in low cost, massively scalable infrastructure (see driver #3). Over the last few years extraordinary amounts of data have been captured via digital devices such as the internet, cell phones, and IoT sensors. All of this data creates examples for machine learning models to learn from and to make better and better predictions from. BlueZoo Wi-Fi and now optical sensors provide critical data for machine learning models to better understand, predict, and prevent risk.
Advancements in Models & Architectures: research from universities, big tech, and even startups has advanced and built upon itself in remarkable ways – much of the most exciting developments are based on mimicking how the brain learns. BlueZoo leverages open source tools, libraries, and standards wherever we can. TensorFlow, OpenCV, ImageNet and a multitude of other community driven initiatives have created a platform for rapid innovation, and we take advantage of these incredible innovations. Simply put, you could not have built a risk measurement system like ours a few years ago. At least it would have been very expensive, and would no longer be differentiated (as many of our peers have realized a little too late).
Cloud Computing and Edge Processing: thanks to the likes of AWS, Azure, and Google Cloud the cost of training, a major inhibitor of delivering compelling machine learning, has dramatically declined. Capabilities such as pay-only-for-what-you-use, store nearly infinite amounts of data at very low cost, access powerful compute resources on demand for training new models – are table stakes in cloud computing. What’s more, advances in semiconductor design to produce low cost custom processors at the edge of the network means sensors can actually perform inference at the edge. As you explore our sensor technology you will realize this adds the extra benefit of providing privacy by doing inference on the edge and never storing or sending information to the cloud. We partner with Google Coral Edge TPUs and benefit from their large R&D investments – so we can focus on measuring, predicting, and preventing risk.