Death and the internet

How often do you wish you could speak to someone who has died? How many questions would you love to ask your departed loved ones? Thanks to new technology, the opportunity to do so is not far off.

Huge chunks of our lives are, of course, already recorded on the web—not only in the emails and the documents we write, but in our photographs and videos, which we share on sites like Facebook, Flickr and YouTube. Companies such as Microsoft, Google and Vodafone offer free storage space in return for our life data; accessing our memories, tastes and records. But what happens to all this when we die?

A European Commission-funded project called Companions, which I ran for its first two years and is now coming to an end, created a trial computer programme called “Senior Companion”: a conversational agent designed to interact with a person for a long period, learning their tastes and habits. For the moment, this would be best suited to elderly people, living alone and wanting company, who might need to be reminded when to take pills and so on. It could also, however, help build up a narrative picture of a person’s life for posterity.

At present, only those with talent, resources and leisure tend to write autobiographies. But, if the Companions project or its successors work, anyone could assemble some form of autobiography for their children. Elderly people often possess a lot of old photographs; soon these will be digital images. The Senior Companion starts by asking its owner who is in each picture, where it was taken and what its importance is. In this way, the programme can store a wealth of memories after the owner has departed.

This may seem a futuristic project, but the Japanese have already shown there is a market for surprisingly primitive devices of this sort if they can achieve an acceptable level of naturalness of voice and manner. The technical basis of Companion programmes is a research area called “machine learning”: the ability of a computer to learn, within limits, things it did not know before. A successful example of this is voice recognition technology: the iPad now comes with software that gives close to full transcription of voice to text.

The same underlying process would allow a Companion to imitate its owner’s voice. Stephen Hawking’s insistence on keeping his 20 year-old electronic voice has masked the great advances made in this field. Car satnavs, for example, offer a wide range of plausible artificial voices with a variety of accents. After years of debriefing its owner’s life, a Companion could certainly produce a convincing approximation of their voice. It would also have access to a huge store of images, emails and documents telling its owner’s life story. From this, it is not too hard to imagine a Companion continuing after its owner’s death to answer questions about their life—in their own voice.

Ray Kurzweil, the computer pioneer who built the first dictation typewriter, is devoting his old age to health products so he can stay alive long enough to benefit from what he believes will be the next great technical advance: the reproduction of every human brain cell in a computer, or “in silico,” as he puts it. A Companion that simulated a dead person would be much less radical than that: it might imitate behaviour but would have no tie to any structure in the departed’s body or brain. It would be no more, perhaps, than a computerised and updated form of the goodbye videos from the deceased now shown at funerals, or on some modern gravestones, which have, instead of a stone memorial, a small solar-powered video of the deceased activated by a switch.

There are already four major types of death site available on the internet: memorial and tribute sites created for the already dead; “locked boxes” of assets and secrets for survivors that protect the individual’s interests after death; “legacy” sites containing last wishes and emails to be revealed or sent after an individual’s death; and “life story” sites that manage autobiographical material for an individual creator, to leave some form of self-presentation of their life.

Given what already exists, then, we shall surely begin to see devices like the Companion in the coming years. Some may think that a programme which assumes the voice and screen image of its deceased owner is an unacceptable form of “immortality.” But it seems inevitable that, in the future, the dead will speak—so it’s worth considering now what form that conversation should take.