Last week, for the first time, an AI system was able to outscore a human on the Stanford University Reading Comprehension Quiz. Actually, two different systems achieved this feat. Systems created by Microsoft and Alibaba were both able to eek out a small victory over the humans. It’s important to put this into perspective that the AI doesn’t truly comprehend what its answers mean, and that there is still a long way to go before the digital assistants of today can really understand our communication.
Still, this does demonstrate significant progress being made. A few years ago we only had Siri, who was often unable to understand what we said. Now we are developing digital assistants that are working on understanding what we meant. That’s a big distinction, and the nuance of human communication makes this next step a tricky one.
At the same time, companies need to be aware that there will eventually be diminishing returns on artificial comprehension. Down the road, the systems will be good enough to understand the context of the conversation well enough to interact with a high enough probability that incremental advancements will be difficult to appreciate.
Where research is starting to focus, and where the most disruption will come from, is strategy, especially related to sales. Of course, we’ve seen the Dota 2, Texas Hold’em, and Go systems of the last 18 months, which can be summed up with the simple expression, “Humans got their butts kicked by computers in a wide variety of strategy games.” We’ve also seen Mind Meld, Gong, and a few others looking to advise humans by giving them feedback and analysis of conversations and sales meetings. As we combine these two areas of growth with increased comprehension, we converge towards an obvious point: Negotiations and Sales.
All of these three areas of research are pointing towards a digital agent being able to not only react to what has been said, but recognizing a desirable outcome for an interaction and strategically moving the conversation in that direction. This might be sales, customer service, therapy, or negotiating terms of a business arrangement. The economic impact of this would be greater than improving data retrieval, shopping suggestions, and calendar management. More than having a digital assistant working for you, this would create a digital partner working with you.