AI should be trained like children to avoid blunders, says Paul Roehrig

By Shannon Tellis

While Salesforce CEO Marc Benioff is pushing for humane and ethical use of technology, the global head of strategy for Cognizant Digital Business — Paul Roehrig, has a different strategy in mind. At a recent conference on the future of work in Mumbai, Roehrig emphasised, why we need to add humans to the equation.

To prevent fiascos like Microsoft’s Tay (an AI chat box which began spewing racist, sexist comments) Roehrig recommended over indexing desired behaviours that we would like AI tools to emulate, just like we train children. “Our children are surrounded by behaviours we may not want them to emulate. AI needs to be trained the same way.”


Another strategy Roehrig recomms paring AI coders with social scientists. “(By) leaving AI exclusively in the hands of engineers, we may end up with systems that are non-inclusive and that amplify our ‘lesser angels’ (biases).”