One of the most frequent questions I get about artificial intelligence (AI) is "What's going to happen next?" Typically, these conversations gravitate toward extremes of the AI spectrum, starting with the realistic technology of today followed by a rapid shift to the science fiction of maybe tomorrow. What is discussed less frequently is the middle ground between those two extremes. That middle ground is the next handful of years. With organizations continuing to adopt narrow AI capabilities at pace, an opportunity will present itself to orchestrate them for broader and more valuable insights via a concept Ovum calls "supersystem AI."
In general terms there are two types of AI: first, the currently available, enterprise-ready capabilities – almost entirely machine learning-based – which are classified as narrow, or weak AI; second, the subject matter of science fiction, classified as general, or strong AI. Narrow AI is a well-defined, task-specific capability, which closely mirrors a human's ability to do the same task but at far greater speed. General AI is usually characterized as a human-like cognitive intelligence and, depending on who you believe, is either decades away or not even possible. (I disagree with both these assertions, but that will have to wait for another opinion piece!) As more and more narrow AI capabilities are rolled out in enterprises, it is my opinion that we will see the emergence of a new concept backed by technology I call supersystem AI.
The first iteration of supersystem AI is a platform for the orchestration of many narrow AI capabilities across an organization, essentially AI managing AI. Why? The reason is twofold: the capability of the technology and the potential value that can be extracted by connecting it. From a technology perspective, the machine learning that is at the heart of narrow AI is very good at ingesting vast quantities of data and identifying patterns; this ability could be applied a level up across multiple narrow AIs as if they were variables of a "larger" algorithm. Applying it in this way leads to the second reason: additional business value delivered by unearthing the relationships and interdependencies between narrow tasks within and across business processes, uncovering opportunities to optimize and make change at a grander, more transformative scale than is possible with narrow AI in isolation. Making individual tasks within a process faster and more efficient is good; gathering these together and learning how to make the whole process faster and more effective is better. I argue that this kind of capability is narrow AI scaled out, bringing together many "simple" machines that can do the work of one very advanced one: or in other words, mimicking a general AI-like system.
Hierarchies of control and management within software are standard fare today, with programs managing programs. Expect it to happen soon in AI – bringing together and compounding many small returns into bigger, more valuable opportunities for transformation.
Straight Talk is a weekly briefing from the desk of the Chief Research Officer. To receive this newsletter by email, please contact us.