Do you remember that episode of Black Mirror with the artificial Domnhall Gleeson? Well, Microsoft now has a patent for a product that is eerily similar to the service used by Hayley Atwell to ‘talk to’ her dead boyfriend. The patent for creating a conversational chatbot of a specific person was finalized on December 1st, which means the Redwood tech giant now has the legal rights to a product that could allow your loved ones to interact with you even after your death.
The patent claims that “social data” – anything from social media posts, to emails, to blog posts, to audio – can be used to train a chatbot application to ‘speak’ as if it were a specific person. Although the patent largely regards a chatbot product, it also notes that it might be possible to recreate an image of a person. “A 2D or 3D model of a specific person may be generated using images, dpeth information and/or video data associated with a specific person” the document states.
For anyone that’s seen that particular Black Mirror episode, this sort of technology can take you pretty deep into the horrors of the uncanny valley.
Microsoft explains that the technology could be used in a range of different ways. This includes the ability to recreate fictional characters, historical figures, or even yourself. However, while the scope is broad, it’s worth noting that inventors Dustin Abramson and Joseph Johnson Jr. have specifically considered the ‘deceased person’ use case. The patent notes, for example, the fact that there could be a missing period of ‘social data between a person’s death and the present moment. This presents a challenge because it means the chatbot’s contextual won’t be up to date.
“Although the social data may comprise information from the lifetime of the deceased relative, the social data may not comprise information related to a time period after the lifetime of the deceased relative.”
This can be solved using a “set of data acquisition rules[that can] be generated for (or assigned to) the personalized personality index. The set of data acquisition rules may provide instructions for acquiring data related to various time periods of the deceased relative’s lifetime.”
It’s not clear exactly when the patent will be turned into a real-life product that can be used. Indeed, there are also many issues around how well such a chatbot will work. Wide ranging data sources might mean more ‘raw material’ for the application, but it also introduces significant complexity – every different data source necessarily has a different context, in terms of how it was created and the meaning that it can communicate. Chatbots need to be ‘natural’ to work – a collage of a given person might offer up a compelling representation, but its unlikely to be a strictly realistic one.
Microsoft doesn’t have a great track record with chatbots. Think back to Tay – the chatbot that turned racist after 24 hours on Twitter. That was, at least, an interesting experiment; simulations of real people are another matter. As that Black Mirror episode demonstrates, even the subtlest hint of the uncanny can feel monstrous and nightmarish.
This post was published on December 8, 2020 1:04 pm 1:04 pm
If you’ve spent time within a particular type of online space in recent years you…
In The Utopia of Rules the late David Graeber argues that we have moved from…
'Accidentally Wes Anderson' is a social media phenomenon. Starting life as an Instagram account which…
Alice Marwick’s conception of social surveillance (Marwick, 2012) builds heavily on Foucault’s notion of “capillaries…
Sponsored by 10KMedia. Trends like observability have underlined that one of the key challenges facing…
“If I get really excited about licensing schemes, kill me,” software developer Jesse von Doom…
This website uses cookies.