Last month I moderated an AI disruptor panel at the World Summit AI in Amsterdam. Our conversation covered the moral, economic and practical considerations that need to be taken into account when it comes to the future disruptions AI will bring. As MIT President, L. Rafael Reif, was recently quoted as saying, “Technologies embody the values of those who make them, and the policies we build around them can profoundly shape their impact. Whether the outcome is inclusive or exclusive, fair or laissez-faire, is therefore up to all of us.” While this is true for every new technology, we are entering new territory when it comes to AI. How can your business be prepared for the impact of AI? Let’s take a look at three key elements to successfully making this transition.
Prepare for AI by thinking clearly about the data that is important to your business.
As we’ve seen in numerous discussions around Big Data over the last five years, the impact to your business - or AI project - can only be as good as the data that is used as input. The algorithms that determine your recommendations on Netflix or the language translator running on deep neural networks are completely dependent upon the quality of data being put into the system. Now, more than ever, it is critical to understand the data at your disposal before your project begins.
Don’t pick your solution before understanding your problem.
By working to understand your problem through conventional analysis, you learn more about your problem and more importantly, the data that will drive solutions. Once you have perspective and the data you need, then you can choose the technology. If AI is the right way forward, you will be in a much better place than those who start building and applying models before understanding the lay of the land. Every technology over the history of time has been created to solve a problem, and AI is no different.
Test early and often.
The world is always changing. Don’t fall into the trap of using fixed models with a false sense of security. If we don’t test our models regularly, they can become prone to bias. Even worse, if the basis of the training changes, the models can become obsolete and we are left with a false oracle. When autonomous vehicles use neural networks to amalgamate large quantities of data to better process a single action, it’s not easy to imagine a worse scenario than a self-driving car using outdated models. For both the safety of the driver and other passengers on the road, it could truly become a life or death scenario.
Artificial intelligence, as we’ve begun to see, acts to commoditise intelligence and decision making. The first tasks taken up by AI are the most simplistic and monotonous - identifying images, recognising sounds, searching for patterns. Simple and repetitive tasks, and later complex and repetitive tasks, will be ‘solved’ through artificial intelligence.
While at the SC19 supercomputing conference in Denver, I came across many examples of where artificial intelligence is impacting scientific projects, including those which involve some of the fastest computers on the planet. In energy exploration, PhD credentialed scientists are used during the data gathering phase to manually scan through seismic images to identify and weed out the areas of the images where natural salts negatively impact the data. This takes a lot of time, slowing down the exploration process, underutilising valuable scientific resources, and costing energy companies millions of dollars that could be applied in more scientific and less tedious ways. This is a perfect opportunity for machine learning and deep learning techniques that could assist the scientists in helping to identify the salt patches and dramatically improve the efficiency of the energy exploration process.
While your company may not be topping out on the Green500, your business can approach new and innovative AI projects with the same sound practices you have applied across other technology implementations. A good practice is independent of its technology.