AI will be both revolutionary and incremental
The time to think about how to deploy AI in your enterprise is now. Your competitors aren’t waiting.
No company, large or small, can afford to ignore AI
We’ve stressed that AI is not about the kind of super-human intelligence seen in science fiction movies, but about human-level pattern recognition and prediction skills performed by software that can be deployed almost anywhere. This means that every enterprise needs to be thinking today about how it will deploy AI.
If the business activity occurs in your industry, then you can be certain that someone is already doing it or thinking about doing it. The question then becomes: is that someone you, one of your existing competitors, or a disruptive new entrant to your industry?
However, not every application of AI will involve a revolutionary change-the-world approach. Self-driving cars and trucks undoubtedly have a big future and promise to revolutionize one of the world’s most important industries: road transportation.
Boards, CEOs, and their legal and compliance advisors must recognize that AI is going to be everywhere: if it is possible to apply AI to automate, improve, or reengineer a business activity, then this will happen.
But there are countless other AI applications that, while less dramatic in scope, are nevertheless already transforming entire industries and their business models. We call such applications incremental AI. In contrast to the revolutionary AI exemplified by autonomous vehicles. Both kinds of AI will drive fundamental changes across nearly every industry you can think of.
Incremental AI is quietly making headway in the nooks and crannies of countless business processes throughout the economy. In addition to exploring revolutionary AI scenarios, Boards and CEOs need to make sure that all business units in their organizations are actively exploring incremental AI.
Examples of incremental AI
The key to evidence-based medicine is to treat patients based on the best available research. But at a time when a million or more medical research papers are published every year, determining what the best research says is a daunting task. For 25 years the NGO Cochrane has specialized in conducting systematic reviews of medical research with the goal of promoting evidence-based approaches in clinical practice.
A typical Cochrane review, which may take two or three years to complete, begins by surveying hundreds or even thousands of papers about a given disease, then culls a few dozen of the most relevant to build a unified statistical model of the most effective treatments. Screening all these papers is a superhuman task that cries out for automation with the help of AI. It should come as no surprise then that these days Cochrane is using cloud-based machine learning to automate this task. This use of AI is not quite as revolutionary as a cure for cancer, but it is one brick of many that are progressively transforming the edifice of modern medicine.
Inner Eye Project
Another fascinating case study from the field of Healthcare tells how AI is helping radiation oncologists perform key clinical tasks more efficiently. The job of these highly trained specialists is to treat cancer by bombarding tumors with radiation. But making sure the deadly beam hits only the tumor and not surrounding healthy tissue is critical. Consequently, before they treat a patient, radiation oncologists typically spend hours carefully delineating the 3D contours of a tumor by manually marking up dozens of MRI cross-sectional images.
Now, thanks to Microsoft’s InnerEye project, this work can be done automatically in seconds by a machine learning algorithm. Again, this is an example of AI being used to automate a routine and labor-intensive back-office task. This incremental application won’t save the world by itself, but it will make the deployment of evidence-based medicine faster and less expensive.
Proper maintenance of airliner engines is critical for both safety and efficiency. Over the past two decades technology has utterly transformed the way such maintenance is performed. Today every new engine on an airliner has thousands of sensors that produce terabytes of data on long-haul flights. Engine manufacturer Rolls-Royce has over 13,000 engines in service with airlines around the world. Processing data to reach business decisions has been at the heart of enterprise IT for decades. But gathering and making sense of the gigantic volumes of data flowing from the Rolls-Royce engine fleet would be impossible with conventional IT. In fact, it is only possible by putting the data in the cloud and applying machine learning to it, and that is exactly how Rolls-Royce does it. Data from each engine is streamed to the cloud in-flight or on the ground, then analyzed by machine learning algorithms that tell the airline when maintenance is needed and how to optimize fuel consumption.
This combination of the cloud, big data, and AI not only delivers safer flying and millions in annual savings to the airlines, but it also allows Rolls-Royce to expand its business model from simply making engines to managing their entire operational life-cycle. To be sure, the use of AI to analyze masses of technical data is not a radically new idea. But it’s hard to think of a better example than this case where the smart revamping of basic IT operations can lead to the digital transformation of an entire business model.
As we all know, self-driving cars get a lot of attention from the media, and rightly so. One day in the future, this technology is certain to have a profound impact on how modern cities and road transportation are organized. But building the AI required to make self-driving cars mature is a big challenge. While tech firms and the auto industry work on that challenge, it’s useful to remember that simpler applications of AI to our everyday driving experience are much closer to deployment.
One example is Volvo’s project to use AI and in-car cameras to assess a driver’s mental state in real-time and determine if he or she is in a condition to drive safely. AI can detect if a driver is alert or drowsy, paying attention to the task of driving or distracted, in a suitable emotional state or not. If the AI determines that the safety of the driver or passengers is in danger, it can issue a warning or conceivably even prevent the car from moving. Of course technology like Volvo’s still-experimental driver surveillance AI raises important questions about privacy. Volvo and Microsoft are well aware that AI of this kind must only be used to assist humans with their permission and in compliance with relevant privacy regulations.
An example of revolutionary AI: autonomous vehicles
While the idea of autonomous vehicles has been around since the dawn of the computer industry, today’s global rush to build self-driving cars and trucks was kicked off by the Defense Advanced Research Projects Agency in 2002, when the agency offered a $1 million prize to any team that could build a robot vehicle able to navigate a 142 mile course in the Mojave desert within 10 hours.
The first race in 2004 was a failure. The best performing entry, built by a team from Carnegie Mellon University, traveled less than 8 miles before getting stuck on the edge of a cliff. But the very next year a team from Stanford managed to complete the full distance and won the prize. The leader of that team, German computer scientist Sebastien Thrun, went on to found Waymo in 2009.
Now, less than a decade later, the entire global automobile industry, along with tech giants such as Google and Microsoft, transportation pioneers like Uber and Tesla, and hundreds of startups are investing billions of dollars as they race to make driverless vehicles an everyday reality.