NYC-headquarterd start-up x.ai has been working to develop a virtual assistant capable of setting up meetings by working with all required participants so as to optimise arrangements.
Artificial Intelligence (AI) has made such progress that there are now a subsntial number of machines that are capable of holding a conversation in an almost-human manner. So will we soon be seeing people and robots working hand-in-hand, with the robots taking on tedious time-consuming tasks in order to free up time for their human ‘colleagues’ to focus on other activities?
We all know how setting up a meeting can often turn into a sort of jigsaw puzzle exercise, especially when large numbers of people are involved. Meetings are often arranged via email, which frequently means endless to-ing and fro-ing while all the participants juggle their availability until a date can be found that suits everyone…only to be thrown back up into the air at the last minute when somebody is faced with an urgent call on their time.
So what if AI systems were able to completely take over the task of coordinating attendees’ diaries? This is the idea being pursued by New York City-based startup x.ai, which has been working for two years on building a virtual assistant (at present still in beta phase) that specialises in setting up business meetings. ‟I like ideas which emerge not through enthusiasm, but from pain. Almost every time I scheduled a meeting it used to end up being postponed, so I wanted to create a solution – to try and avoid exchanging dozens of emails just to arrange one meeting,” explains x.ai founder and CEO Dennis Mortensen.
A ‘vertical’ assistant
The virtual assistant designed by x.ai is called Amy. She is equipped to undertake all the steps necessary to set up a meeting: ‟Amy checks everyone’s availability by email, writing to them as a human being would, and makes sure that all the participants are ready to meet up at the right place and time,” explains Dennis Mortensen. So Amy is a very different kind of virtual assistant from those available on most smartphones.
Mortensen describes Siri, Google Now, Cortana and other similar virtual assistants as horizontal virtual assistants – meaning that they are set up to respond immediately to a large range of questions and requests, but are not geared to carrying out an entire end-to-end task to perfection. Amy, on the other hand, is what he calls a ‘vertical virtual assistant’ – i.e. she has been trained to carry out a specific task, involving a number of different steps over a longer period of time. Arranging a meeting with several people requires much more than a simple request-response sequence. This job involves a final objective, calling for a series of actions and interactions that will take place over several hours or days before the goal is reached.
At the moment the technology works via email, but the x.ai system has been designed for use on all types of communication channels. So in the near future it could be based on text messages or voice communication. In fact it may well be that in the near future we will be using a range of different AI-based assistants, each specialising in a different ‘vertical’. Time for a change of scenery? All you will have to do is give your desired destination to your virtual assistant, and s/he will take care of booking your air tickets and hotel room. Need to see a doctor? The AI system will make an appointment and send the doctor all your up-to-date health data. In this scenario your horizontal assistant will act as go-between with the various specialised assistants: you will, for instance, be able to ask Siri to plan a meeting, and Siri will depute Amy to get on with the job.
Human, all too human?
Apart from being highly specialised in what she does, Amy has a distinctive personality: neither too formal nor too familiar, not without a sense of humour but focused closely on her work. The x.ai team have made it a point of honour to ensure that, as far as possible, Amy’s emails seem to have been written by a human being, with such success that some users actually believe Amy is a real person. ‟We’ve invested a lot of time and money in humanising Amy. Our goal was that even though users know they’re dealing with a machine, they’ll still treat her just like a human being,” underlines Dennis Mortensen.
So how do you build an AI system that appears to be human? ‟I don’t think the answer is to make her funny, for example, I rather think our humanity lies in the fact that we have a consistent, well-defined personality. So every time you talk to Amy, there’s continuity, you have the feeling that you’re always talking to the same person.” And has it worked? Mortensen points to evidence that it has: users not only tend to use polite formulae when they send a request to Amy, many of them also write emails just to thank her for her work! Some go so far as to apologise to the virtual assistant when they have to postpone a meeting, politely explaining themselves and making excuses which are of course not really necessary.
Researchers at Kyoto University in Japan have already demonstrated that human beings can feel empathy towards robots. The Media Equation, a theory developed by researchers Clyfford Nass and Byron Reeves at Stanford University, also shows that we tend to apply the same rules in our social relations with robots as we do in our relationships with people, resulting in our being polite or showing deference when talking to an AI system. Such findings may raise questions as to the very nature of humanity: is it necessary to have a body, to be made of flesh and blood in order to be human? Is humanity to be found rather in the mind – is it about the capacity to express oneself, to demonstrate a consistent character, to be able to think and reason? It seems quite likely that these ethical questions will take on greater importance in the coming years.