The Future of Intelligence Relies on Human-Machine Collaboration
Without human ingenuity, technology may fail to fully realize its potential.
By Shazia Manus, Chief Strategy and Business Development Officer, AdvantEdge Analytics
Amazon just announced it will retrain 100,000 of its employees by 2025. The employees, who wear both white and blue collars within the organization, hold positions Amazon anticipates will be impacted by the further integration of robots and automation into their environment.
“When automation comes in, it changes the nature of work, but there are still pieces of work that will be done by people,” Amazon’s vice president of people operations told The New York Times. “You have the opportunity to up-skill that population so they can, for example, work with the robots.”
The recognition that human ingenuity must remain a meaningful part of our work is really poignant. Robots, automation and the artificial intelligence (AI) that powers them are about replicating human intelligence, not replacing it.
Recognizing the Limitations of AI
There are many different facets that make up intelligence, and machines are excellent at some. In other areas, they fall short.
Take image recognition, for example. When humans see a photograph, they instantly understand what they are looking at. They can identify the subject, understand the setting and read the emotional undertones.
Computers, on the other hand, view images as a collection of pixels. For an AI platform to recognize, categorize and make decisions based on an image, it must have ingested mass quantities of similar and dissimilar images and learned to distinguish what’s important from what’s not important. Unless or until it’s been properly trained (with massive data feedings), an AI tool does not have the natural context humans carry with them every day.
Another limitation of AI without a human in the loop is its reliance on massive amounts of data to detect patterns and learn from them. Particularly for organizations like credit unions, which operate within the confines of legacy infrastructure, accessing, categorizing and cleaning that data to feed the machine is an incredibly complex undertaking. To realize the true potential of machines to learn and advise, humans must institute data transformation and data maturity roadmaps to ensure their strategies grow alongside the volume of data AI needs to function.
Recognizing the Limitations of Humans
Of course, humans also bring their own weaknesses to the machines. Today’s AI solutions are learning from algorithms most often developed by humans. When turned into disparate bits and bytes of data, poor judgement or lack of sensitivity can take on a life of its own.
Take, for example, a recent beauty competition that was created for the specific purpose of letting machines do the judging. The idea was to free such a competition from the bias of humans. Thousands of people across the world submitted photos with the hopes of uncovering “true beauty.” The human-developed algorithm had other ideas. It began to associate skin color with beauty and picked winners solely on the basis of race.
Recognizing Our Responsibility
Technology helps humans reach beyond their potential. But, just the like humans whose lives it’s built to enhance, technology is imperfect. When it gets something wrong, the consequences can be devastating. While the vast majority of decisions made in financial services are not life-or-death, even small ones can have massive ramifications on the lives of banked (and even unbanked) consumers.
Because they have been designed by humans, the algorithms on which many of today’s cognitive systems are built are likely to be flawed. People are so much more complex than their data profiles, and the patterns and insights detected by algorithms are only as good as the data fed into them. That is why keeping ethical humans in the loop is essential.
As the credit union industry mines our systems, networks, channels and other sources of data, we will have a responsibility to use that information in a way that is worthy of member trust. Insisting on and training humans to be effective collaborators with machines is an excellent first step in the execution of that vision.
For an in-depth look at much of the hype and myth surrounding AI, download our white paper, How Humans and Machines Will Transform the Credit Union Industry.