AI? Or cognitive computing?
At the other end of the vertical axis are systems that serve to discover issues and insights of which we are unaware—they are the famous unknown unknowns popularized for TV audiences during the early Gulf War by former Secretary of Defense Donald Rumsfeld. The modality of these systems is exploration; the goal is discovery. Fraud discovery in financial services businesses is one example of this kind of application. Network surveillance in the context of cybersecurity is another. Finding new molecules for disease treatments is another. Discovery systems fall squarely in the realm of serendipity, which any innovation executive will confirm has a huge impact on the business.
With this simple model, it becomes clear that one of the basic problems with our current media hype around AI terminology is that it doesn’t distinguish carefully between the augmentation level of function and the truly artificial level. No systems today exhibit artificial general intelligence—many exhibit some level of augmentation for human intelligence. The model also suggests a home for cognitive computing systems. In our view, they fall almost exclusively in the left-hand half of the diagram. They are about augmentation. They can be built with today’s technologies. They help us understand big data and can help us discover threats and opportunities, but they don’t pass the Turing test. Although they are based on learning, they can’t reason and respond the way humans do.
Next time you need to explain AI to the board or to the executive team, try using this simple matrix to characterize where in the array the applications you’re interested in land. By clarifying the conversations around AI, we will all avoid pulling out the snow blowers for the next AI winter.