Abhishek Gupta, a machine learning engineer at Microsoft and founder of Montreal AI Ethics Institute, highlights some important issues to bear in mind as many companies are accelerating their adoption of more sophisticated AI due to the coronavirus pandemic.
For implementing human in the loop, Gupta suggests that the agents in the loop should have a real understanding of how the AI works to avoid “automation surprise.” Automation surprise happens when things go wrong with automated technology, and the humans-in-the-loop become confused and struggle to regain control.
Gupta explains that a lot of AI engineers spend too much time worrying about how to train their algorithms, and little time thinking about the humans who will use them. He has a lot of good perspective on how to train AI engineers, but incorrectly assumes engineers are the only problem solvers who create and apply AI.
This is where the power of co-creation is incredibly valuable – diverse solutions come from diverse perspectives, and the ability to co-create AI changes the way it is made. Rather than relying different teams (engineering and marketing/customer service/etc.) to come together and attempt to speak the same language, equip your team with a conversational AI platform that allows everyone to come together to co-create the experience. Co-creation will result in better trained algorithms and many more humans who understand how the AI works to avoid automation surprise.