Will they even care if it's me?

It’s not immediately obvious that JK Rowling’s “Harry Potter” stories have much relevance to AI.?But I recently realised they tell us quite a lot about the future of this rapidly emerging and disruptive technology, and also about the enduring relevance of “genuine presence”.

Our children grew up with Harry Potter, and the whole family adores JKR’s fables and their wonderful performance for audio by Stephen Fry.?So please forgive my use of it to make a point.

At the very end of the saga, in “Harry Potter and the Deathly Hallows”, Harry visits the Headmaster’s study to ask the portrait of Dumbledore for “one last piece of advice”.?At this point in the narrative arc, Dumbledore has been dead since the last volume, and the story makes no compromise with this – the portrait is not a ghost, or a connection to the spirit world or the soul of the protagonist. Dumbledore is quite, quite dead.?So the portrait is a representation of the personality, values, modes of speech and habits of dialogue which once characterised the departed headmaster.?It has no soul, no consciousness and no life.?It is an illustrated chatbot, so convincing that it satisfies Harry’s desperate psychic need at that moment.

Many of us who have lost parents will recognise this yearning to be able to speak to the departed one, to ask them for guidance or reassurance.?Part of the human condition is to learn to cope with such loss.?But what if we didn’t have to?

Recent progress in the development of deep generative models has stunned the technology community.?I don’t need to add to the clamour here, but the more one thinks of the potential applications, the weirder the possibilities become.

One application, presciently foreseen by JKR, will be the training of models on the spoken and written communications of individuals, captured with their consent through listener apps during their lives.?These models will create visual, speaking representations of the person after they have passed away.?I have no doubt there will be a lucrative market for this, both amongst the subjects whose vanity or compassion compels them to create such facsimiles, and amongst their descendants or acolytes who demand such an escape from pain and abandonment.

With well-trained models and enough data, I don’t doubt such models will cross what is sometimes called the “uncanny valley” – the sensation of similarity which is close enough to be familiar and recognisable, but just wrong enough to be creepy.?Like Harry, people will have the very real and believable sensation of conversing with their departed loved one, aided perhaps by some confirmation bias.?What the effect of this on society might be, I struggle to imagine.

But why wait until death??Back in the 1990’s, companies like General Magic started to examine the concept of network agents: autonomous software programmed and sent off to execute the owner’s will in transacting with third parties.?Now let us come forward 30-40 years, to a time when agents that look, sound and transact just like us can engage with service providers, friends and family to conduct relationships and business for us, while we are using our time in other ways.?I always complain I can’t be in two places at once – will this be the solution??And what, exactly, will be the harm in it?

Perhaps some of the harm will arise from the passing off of “spoof me” as “real me” – the deliberate intent to deceive the counter-party.?But what if the counter-party has become accustomed to this kind of interaction, and no longer cares if it’s me?

Harm may be caused if the agent behaves unlawfully, dishonestly or reprehensibly.? Who is to be held accountable??In the world of real people, accountability is vital as a means to maintain social order and cohesion. Accountability permits retribution, which serves as a deterrent for the offender against repeating the act, a deterrent to other members of society who would rather avoid such unpleasant consequences, and satisfaction of an innate sense of justice amongst the wider community.?But how can a machine be held accountable?

These harms can be mitigated if the counter-party knows that they are dealing with a fake.?They may not care, or they may be prepared to take the risk of bad behaviour, but at least they are empowered to make that choice.

When I was in Aachen (Germany), at the Rathaus, I was initially dazzled by the ancient Imperial Regalia of the Habsburgs.?When I read the label and learned that they are but copies of originals in Vienna’s Hofburg Palace, I still chose to marvel at their magnificence but with a greater sense of understanding and of perspective.?I was no longer being unwittingly conned – I understood the nature of my engagement with the objects.

In future, if the water company decides to take instructions from a me they know is an agent, that is a risk they have chosen to run.?A bank may not be willing to do so.

And society may decide that, for the psychic health of a person, they have to listen to their late beloved’s disclosure that they are not real but merely a mechanical representation.

Of course, the absence of such disclosures will be a tool of deception.? Genuineness cannot be reliably inferred by the absence of a disclosure – it will have to be positively proved. ?So in future, it will become essential to have ways to prove genuineness to a counterparty in almost every remote transaction.?That proof must be resilient to the best efforts of AI to produce convincing facsimiles, especially of the face of the subject, which to the human mind is a most powerful signal of identity.

So they must care if it’s really me.?And this confirms the vital importance of what we are doing at iProov.?Proving the genuineness of a remote person will evolve from being, today, a means of defence against money-laundering and a tool in enterprises’ cyber defences.?It will become a fundamental foundation of societal stability and the psychological well-being of its members.?