When we think of Voice Enabled Digital Assistants (VEDA), one movie automatically comes to mind: Her. In Spike Jonze’s brilliant movie, the Voice Assistant brought to life by Scarlett Johansson might be a sneak preview of what awaits us tomorrow. It is, however, still far from today’s reality.

One might consider the first commercialized VEDA, revealed in October 2011, by Apple CEO Tim Cook, known as Siri.

Despite the pace at which voice recognition technology is improving, the capability to hear and understand ranges from 95% to 98%. Despite the progress made by VEDA in terms of Artificial Intelligence (the capability to answer) in six short years, one should not forget, like any six-year old, it still has a lot to learn.

Publicis Media clinically tested[1], for example, 120 semi-complex questions on two Voice Enabled Speakers (VES), TMALL Genie and Raven (Baidu). Defining semi-complex questions as those with an additional, yet simple, element of complexity, we asked “Play me the best song of all times” rather than “Play me [name your song]” to observe differences. The result of testing showed that while over 95% of the answers provided were accurate, Raven and TMALL Genie only provided 30% of answers to our questions. In fact, in some cases, the feedback we had from the assistants was “I'm still young and naïve, I can’t answer this question”.