Blended AI: How to Use the Best of Man and Machine

March 30, 2018 — by MediaMath    

Artificial intelligence isn’t artificial in the sense that it’s fake. In this case, artificial means “man made.” Everything about AI is based on human knowledge. Data from searches, for instance, comes from people looking for information.

The image then, of a machine making cold calculations without human assistance is misguided. From the beginning, AI has been about using human knowledge to automate decision making.

It’s logical then that AI should be used to augment human performance rather than act as its replacement. This hybrid of AI and human work has been called “blended AI.”

The reasoning behind blended AI

What we think of as AI today is really narrow AI, a.k.a “weak AI.” That is, it is AI focused on solving a specific task. Systems in autonomous vehicles, for instance, attempt to solve the “problem” of cars that require human drivers. A system that can beat a grand master at chess is another example of narrow AI.

General AI, meanwhile, would mimic the human brain’s ability to be adept at many different things. General AI would also be able to synthesize that knowledge. Some even believe that general AI would mean a machine has consciousness.

Weak AI has gaps that necessitate human help. Personal assistants like Siri, for instance, lack context. As Robert Scoble pointed out, you ask Siri “What’s my favorite gas station?” it can’t tell you even though it knows which one you visit most.

Blended AI in action

Blended AI isn’t theoretical. It is the primary way that AI is employed. For instance, Alphabet uses AI to screen YouTube videos, but it also employs thousands of moderators.

A more common example is in customer service. Sweden’s Swedbank uses Nuance Communications’ chatbot Nina to field customer service queries. Nina solves 78% and sends the rest to human reps.

Looking ahead in 2018, Forrester Research predicted that blended AI would help improve sales and cut customer service costs. Forrester cautioned that customers won’t necessarily like the change. A recent survey found that 71% of U.S. consumers said a bot couldn’t answer their questions or help them.

Even so, Apple, Nike, Uber and Target, among others, are putting more emphasis on chat. Those companies are dialing back on email as a customer service channel in favor of chatbots, according to Forrester.

The success of such programs is likely to hinge on the proper use of blended AI. Though critics say AI is a job killer, AI supporters counter that rather than eliminate jobs, AI will improve them. Automating repetitive aspects of work will let workers spend more time being creative, they say.

Blended AI realizes this vision. Instead of answering rote questions, customer service reps will use emotional intelligence and creativity to solve harder problems. The result in this vision is a virtuous circle: As AI improves, so will human workers. Both improve by exposure. From the consumer’s point of view, the improvement will be seamless.