Talking to the departed

In the first episode of the second season of the great British speculative science fiction series, Black Mirror, entitled “Be Right Back” (IMDb, Wikipedia), broadcast in February 2013, a woman devastated by the death of her partner in a road accident, hires a service that offers to “rebuild” the personality of the deceased based their social network footprint.

The episode continues with another set of avatars that I’m not going to discuss so as not to spoil it for those readers who have not yet seen it (if you like this page and still have not seen Black Mirror, I highly recommend you rectify that). The point is that there was a long article yesterday in The Verge, called “Speak, memory” which tells how Eugenia Kuyda, co-founder of artificial intelligence startup Luka, which develops chatbot technology, took the decision to “rebuild” her friend Roman Mazurenko, who died in a traffic accident, using her enormous archive of instant messaging conversations with him.

The article mentions that Eugenia had seen the episode of Black Mirror and that this inspired her to pay a tribute to her friend that would go beyond a simple photograph album or website. It also notes that the reactions — positive and negative — of friends and relatives of Roman, and the extent to which the bot Eugenia created really reflect the personality and way of speaking of her friend, as well as how “talking” with an artificial intelligence version of Roman based on the conversations she had with influenced the grieving process. I can think of one or two people I have lost that I would have loved to talk to in this way, and I think that at least one would also have liked it.

We find ourselves in one of those moments in technology that raise some very serious questions: how does artificial intelligence based on the characteristics of a person develop as it is fed with successive conversations? What happens when artificial intelligence no longer simply repeats answers based on pre-programmed situations, but is able to build and develop on the basis of many different features?

We will all have very different responses to the idea of talking to a dead person whose personality has been reconstructed from conversations, but what the piece in The Verge really highlights is that chatbots based on artificial intelligence are no longer limited to executing commands and replying to questions with set answers, but can now add much more value, or even to transfer values. How smart can a bot get? The answer may very possibly surprise you.

Chatbot technology is progressing very quickly: the creation of languages ​​and platforms and the appearance of bot stores in Telegram, Kik, Facebook Messenger, Microsoft, Skype, Slack and even some wikis and aggregators, or the febrile activity around the topic foreshadow the development of a bot economy, one in which just about all companies will be involved.

When chatbots are not only able to answer predetermined questions from a list of FAQs, but can develop an intelligence based on feedback allowing it to respond to an increasingly wider range of situations, we will be able to move on from using them for customer service or to provide certain direct services, then we will start to come up with much more ambitious possibilities.


(En español, aquí)