Can Sophia Answer the Question: What Do We Want Out of AI?


Over the last several months I’ve continuously seen ‘Sophia’, in the news. Whether she is trolling Chrissy Teigen, taking her first steps at CES, or discussing her future in Korea, Sophia is sure to be nearby whenever I’m scrolling through the news. For those of you who may not have heard of Sophia, she is a robot created by Hanson Technologies that combines robotics and Artificial Intelligence. Hanson Technologies touts that Sophia is more than just a robot, she is a real, live electronic girl who is ready to see the world and live with people.

What I’ve noticed in the past few months as Sophia has become more and more of an icon, is that people either love her or hate her. Just look at any of the comments on her Instagram, Twitter or  Facebook and you’ll see a heavy divide. But this divide isn’t just among the everyday consumers, it carries over to those who live and breathe technology – especially those who work with AI regularly.

In early January, Facebook’s AI expert, Yan Lecun, took to Twitter to express his frustration with Sophia saying that “This is to AI what  prestidigitation is to real magic…in other words, it’s complete bullsh*t (pardon my French).” But Lecun isn’t alone, many researchers and journalists agree that Sophia and Hanson Technologies are doing a disservice to Artificial Intelligence and sensationalizing to the public what AI can do and how ‘human’ it can become.

AI technologist, Kriti Sharma, said in her article last year that it isn’t about if AI can help robots become more human, robots shouldn’t be trying to become more human at all. Instead, AI should

be used to help people by focusing on problems and enhancing productivity without assuming an emotional role in society. But how? I spent some time chatting with Tobias Goebel the Senior Director of Emerging Technologies here at Aspect Software to find out his thoughts on Sophia and AI.

His perspective echoed what so many others believe. While AI can help a company improve their processes in a variety of areas, e.g. in customer engagement, there are human attributes that robots and AI will never be able to fully embrace. Human empathy, creative problem-solving, good judgment, and emotional involvement will remain a uniquely human undertaking; AI can only mimic these attributes and never truly learn to feel them. What remains interesting about the average person’s reaction to Sophia is the effect of anthropomorphism, where human categories are applied in order to make sense of what Sophia is, while she is “merely” a combination of several (in parts very sophisticated) technologies that have existed for years, some of them decades: automatic speech recognition (or “speech-to-text”) to hear us, natural language understanding (or “text-to-meaning”) to interpret what she heard, speech synthesis (or “text-to-speech”) to speak back to us, basic dialog management to keep track of the conversation, and elaborate robotics for her facial gestures.

Tobias also agrees with Kriti’s sentiments that when used correctly AI provides a huge opportunity to enterprises. AI’s goal, especially in the workforce, is to augment human efficiency and quality on both staff and an individual contributor level. AI is mirroring how humans conduct tasks and solve problems, and as a result, it fits even better into existing human workflows.

AI should focus on two broad goals that can directly impact the success of a business: increasing efficiencies and finding peak performance synergies between human and AI-powered ‘digital’ employees, while improving the quality of work life. Two great examples of this are The Radisson Blu Edwardian who successfully deployed a customer service chatbot using AI that has drastically improved their guest experience, while freeing up the front desk staff for more personal face-to-face help (vs. answering commonly asked, easily answered questions from guests). The other is Lidl UK who just this week introduced Margot, their virtual sommelier who also uses AI and natural language understanding to provide wine advice to shoppers since that knowledge isn’t available with the in-store workforce nor the contact center.

There is no doubt that we have just started to scratch the surface of AI and its capabilities, especially in the workplace. Over the next several years we can anticipate that AI will become commonplace within companies, but it is important to remember that while some AI may be ‘over-humanized’ it truly can only mimic the emotions and self-awareness that humans experience…at least for now.