The Edinburgh Fringe is no stranger to abstract and esoteric art and the show ‘Siri’ was no exception. A short theatre piece built around a dialogue between Laurence Dauphinais, the human actor, and Apple’s Siri, the intelligent agent.
Laurence explores the nature of identity, intertwining her own search for self, resulting from being a child born through Artificial Insemination, i.e., an early ‘test tube’ baby, (the first of the AIs), with the interlinking of human identities with our intelligent virtual assistants (the second AI).
Art is an excellent medium to explore the metaphysical, and it seems appropriate that the philosophy and human consequences of AI are being explored this year at the Fringe.
The Art of Artificial Intelligence
Artificial Intelligence has again come to capture the public imagination in the last 18 months, due to a number of significant leaps in capability. In the main spawned from a combination of more processing power, greater access to data and some improvements to core algorithms. However, as is apparent in ‘Siri’, we are still a long way from any form of general intelligence recognisable to the public. One of the key areas to have significantly improved is speech recognition and this was highly evident in the theatre. On numerous occasions, Laurence dictates long monologues to Siri and the transcription accuracy is almost flawless. Siri’s responses, by contrast, are almost always canned, banal and repetitive. This highlights an area of AI that is still really in it’s infancy, natural language understanding (NLU).
Human conversation relies, among other things, upon our outstanding ability to maintain context and resolve ambiguity through time. It is in these areas in particular that the current capabilities of NLU are lacking. However, this is a key area of research for most AI labs, be they academic or Silicon Valley, and one where some progress is likely to be made quickly.
It’s All About Context
Context, put a different way, is the maintenance and access of relevant data with respect to our current interlocutor and the moment of the conversation. All of the current intelligent personal assistants available today (Siri, Alexa, Google Assistant and Cortana, etc…), are manufactured by companies with significant access to our personal data, albeit with slightly different slants dependent on their business model. The agent that, in the future, we are most likely to adopt is that which, is most able to understand us to provide context (a) to the conversations we have with it and (b) to support our transfer of agency to it.
It is this that raises the most pertinent questions, both in the play and for society. Laurence frequently asks Siri, ‘are you like me?’, ‘are you the same as me?’ and ‘what do you know about me?’. Although Siri proves a limited conversationalist in these areas, there are a couple of telling, if not slightly chilling, responses, particularly Siri’s frequent retort ‘like user, like assistant’. For me, this poses us a number of moral and existential questions that society must begin to contemplate, among which are…
The Impact of Artificial Intelligence
Firstly, if, in the future, a personal assistant is able to acquire enough data about an individual to effectively make rational choices in the same manner as them, and to converse in an almost identical way, how intwined do their identities become? May I decide to assign power of attorney to it in the event that I become incapacitated? May my love ones continue to co-exist with it after I die? What does this do to our sense of mortality?
Secondly, the custody and maintenance of these identities is entrusted to a, not necessarily, benevolent third party. To what extent should we be willing to allow a third party to have the ability to modify, restrict access or even exterminate our potentially closest, most knowing companions?
Finally, the language, tonality and level of respect we show in conversations with out intelligent assistants becomes increasingly important. This was highlighted in a recent article in MIT’s Technology Review (Growing Up with Alexa) where the author explores the effect on children of the way they interact with their virtual ‘butler’. The article discusses research relating to whether having a continuous relationship with a subservient and sycophantic intelligent assistant will affect the way we form relationship with our human companions. If we extend the thought experiment to include an intelligent assistant that has in effect become us, what affect will the way we converse with our third self have on us?
A Human Approach To AI
The nature of identity and artificial intelligence was aptly explored at the Fringe, shedding light on our constantly shifting sense of self and highlighting the nature of this new ‘other me’. For me, this clearly underlines the importance of approaching Artificial Intelligence with a human centred approach.
This is important from the technology side, where ensuring designers are involved and educated in AI alongside their developer or research colleagues, is essential to building usable and enjoyable experiences (well explored here). It is also important from the philosophical and legislative side, where adopting a person centred approach to encroachment of automation and AI in society may help spare us from the worst of the possible excesses of individual greed and disregard for our fellow humans.
If you are in Edinburgh this Fringe, its definitely worth checking out the show.
Dan is the Principal Solution Manager for Digital at Sabio where he is responsible for a range of Digital Self-Service solutions including speech technologies, IVR, biometrics, virtual assistants and digital engagement. Previously a Senior UX Consultant, he has over 15 years academic and industry experience in AI, speech and cognitive science. You can follow him on twitter here.