Unpredictable Patterns #118:Life after death

On the policy and legal aspects of digital necromancy

Dear reader,

A sombre topic this week, but one that I think is increasingly relevant: how do we regulate the digital reanimated dead?

Voices from beyond

In US courts it is not unusual that the victim has an opportunity to comment on the case and the sentencing - they can call for leniency or express their wishes that the perpetrator is sentenced to the greatest extent of the law. In the case of Christopher Pelkey, who was involved in a road rage incident, Pelkey said he believed in a God that forgives, and that forgiveness is key. The judge thanked him and said the statement showed Pelkey to be the kind of man he thought he would be.

Or was. The thing with this particular victim statement was that Pelkey was shot and killed in the incident, and the statement was delivered via an AI-produced video that his sister had produced. In the video Pelkey also showed an image of himself as he would have looked if he had a chance to grow old, aged with AI-technology, of course.

The story got immediate pickup, and has been making the rounds in the US - but it is not an isolated incident. To a greater and greater degree, artificial intelligence is used to simulate minds - living or deceased - in ways that are blurring the boundaries between life and death.

This is largely happening without a policy debate attached to it, and we are just beginning to explore the regulatory questions that this new capability will raise - and that is not surprising, given that death used to be the great absolute, not a negotiable policy equity that we could draw lines around. Now, there are some early attempts, looking at things like the Tupac hologram, but there is more to explore here.

The fungible identity

At MIT there is a project called the Augmented Eternity project. The project is approaching identity as something that no longer needs to be bounded by space nor time — and is exploring things like swappable identities. Do you want to try on your friend’s identity for a while, or get their advice without really asking them? It should be possible, now that we can simulate other minds to some first approximation:

“Can software agents become our digital heirs? Can a head of state, a scientist, or a business owner leverage machine intelligence to complement succession planning? What if you could select the digital identity of a deceased person from a social network and activate it as a pluggable ontology into your iPhone’s Siri and ask a question?”

This question from the project site is then answered, by describing the aims of the work:

“This project uses a distributed machine intelligence network to enable its users to control their growing digital footprint, turn it into their digital representation, and share it as a part of a social network. The project creates an evolving ontological mapping of an individual based on her digital interactions and allows the person to represent her aggregated knowledge-base in form of a software agent. This agent can then be rendered as a chatbot or a voice-based assistant. The project is aiming to open-source a number of "identity render kits" to enable users to quickly share their knowledge base within a trust network.”

The model here is fascinating. It is a future in which identity becomes decoupled from presence. You no longer have to be there to be there. Your identity can be replicated, extended and - presumably - modified in different ways. You can also imagine combinations, of course, although we rarely see such hybrids - but why wouldn’t people start experimenting? Imagine having a chance to speak to a mix of Einstein and Da Vinci? Or Niels Bohr and Lemmy from Motorhead?

One possible reason that we do not yet see a market for remixed identities is that there is no standard here yet - no standard representation of us that can be used to summon us from the digital netherworlds - or otherworlds. That is exactly what the project is aiming at, however.

Note that the people behind the project are not in any way malicious: their vision is one in which we have some control over our digital identities, through this technology - but any such control will also need some kind of regulatory framework.

Regulatory options

So, how should we regulate our right to our identity? It seems natural to think that this is something that data protection law already has covered, and that the regulatory framework is simply one in which we look at the GDPR and figure out how things work. The thing is that the GDPR does not cover us after we are dead. Recital 27 is clear:

This Regulation does not apply to the personal data of deceased persons. Member States may provide for rules regarding the processing of personal data of deceased persons.

So, data protection law - at least European - does not cover the identities of the dead. This makes for a quirky consequence: as identities carry more and more value, you could imagine a company designed to collect - literally - dead souls in the spirit of Gogol. Such Gogol-markets would then be able to collect, resurrect and license these personalities, creating a market for ghosts.

Now, the recital does say that there are opportunities for member states here to craft their own rules, and some have. But there is no concerted thinking about how legal personality should be defined and regulated in a world where it might be reanimated by new technologies - since this is, in many ways, a surprising question.

Another possibility, of course, is to simply state that this is a question about copyright. The materials used are likely to be created by the individual, and this means that they have copyright that extends a number of years after the death of the creator - that should take care of things, right? Well, maybe - it depends on how much of your information I need to recreate you. The mistake we make here is to think that your identity is produced by you. In fact, you are - as Heidegger notes - strewn in the eyes of others.

An intriguing measure of how you have lived is if there are a subset of your network that could describe you well enough to recreate you with AI. Or are you such a solitary individual that no such description could not come close? What is the level of fidelity that you feel your spouse, friends, children and others could achieve? And yes, you should feel a bit uncomfortable answering this question because they may actually be able to do it to a greater fidelity of identity than if we used the materials that you have produced, written and created.

Who you think you are is not as relevant as who you are to others here.

But let’s imagine that we extend copyright to identities - then we should also extend the ability to open source yourself or to license yourself under creative commons licensing. An enterprising individual might create a register right now that allows you to register for “attribution-share alike life after death”.

Or should your identity actually become a public good when you die? There is an argument for instituting a system for identity donation, mirroring existing systems for organ donation - where your experience, perspective and insights might be invaluable even after death. A wonderful teacher, a great musician, a trusted friend — the public utility of having access to these seems greater than the interest of the deceased to not be replicated?

Should their be an archival duty for states to preserve some or all their citizens? A great library of the dead, where you can go to speak with the history of the nation?

But this puts us in a quandry - if this is true after death, are there not people whose value is such that the whole balance of interest should tip towards the public goods model even when they are alive?

Technology and choice

In a 2023 prescient article about digital necromancy Laura Hood argues that summoning the dead through technological means is an extension of our grieving practices - but I think that it goes beyond that. It is not just about grief, it is about our relationship with death. Our defiance of it. We simply do not want to die, or for death to take those close to us. This is why we have put protocols such as “do not resuscitate” in place: we know that we make selfish decisions when someone close to us dies.

But what about our right to really die and pass into memory? Is it crazy to argue that we may want to make an active choice here as to whether we are kept around in digital form? Of course, you could make the argument that it is not really you, and that it is ridiculous to speak of us wanting things after death, but there is something worth considering here.

Since we now have the capability to replicate individuals, we also have choice. When we have choice we should ask who gets to make that choice. That is how technological revolutions work in regulation: new choices are introduced, and those choices becoming increasingly more powerful. New choices require new frameworks.

Technology increase the space of possible choices, and regulation first distributes and then restricts those choices.

So, why should we not have an option to say that we want a DNR-protocol, but one in which the R stands for “replicate”? Do not replicate me after my death - live your own lives, do not stay with me, leave me a fading memory and honor me by making that memory matter in its own way, the way memories do.

Is something lost from society if we lose the memories of the dead, and replace them with the presence of them instead?

The fidelity of replicas

On the other hand, maybe you want to collaborate with the dead - maybe you are not quite ready to leave them yet. But is that a good thing? Laurie Anderson recounts how she has worked with a replica of Lou Reed:

“I’m totally 100%, sadly addicted to this,” she laughs. “I still am, after all this time. I kind of literally just can’t stop doing it, and my friends just can’t stand it – ‘You’re not doing that again are you?’

“I mean, I really do not think I’m talking to my dead husband and writing songs with him – I really don’t. But people have styles, and they can be replicated.”

But she also recognizes that the fidelty is not perfect, far from it:

The results, Anderson says, can be hit and miss. “Three-quarters of it is just completely idiotic and stupid. And then maybe 15% is like, ‘Oh?’. And then the rest is pretty interesting. And that’s a pretty good ratio for writing, I think.”

This is a crucial point. What Anderson speaks about here is something fundamental: the fidelity of the replica. After all, we simulate the dead when we read old letters or diaries, it is just that then we simulate them in our own minds. They do not run on GPUs, but on our wet ware. Should that be prohibited as well?

Ghost photography was an early interaction between technology and death

The difference between a “do not replicate” and a “do not remember” regime is one of fidelity. The mind simulation of your beloved, is a reconstruction not a replica, it is different, it melds with your current situation and who you are right now, and it does not answer questions or perform actions. The replica can be called upon to do things, to perform - it is more independent. It is reminiscent of the tibetan magical Tulpa - the imagined being that finally takes on a life of its own. The dead, writes Axel Gruveaus in “Beyond Peak Death? –The Advent of Digital Necromancy and Functional Ghosts” become a functional category.

The Tulpa is created through intense concentration and spiritual practice, if you believe that it can be created at all. The replica is created through data and transformers.

This gives us a curious observation: artificial intelligence has led to a rapidly declining price for the production of tibetan tulpas. What will this mean for society?

The socio-economic impacts of the dead

The socio-economy of necromantic AI is not simple to assess. On the one hand it seems as if it would be a good thing if we can deepen intellectual capital stores beyond death, on the other hand we should worry if allowing the dead to roam our markets will create conserving effects that slow down social and economic change.

What are the effects of allowing for the dead to remain with us? The reverse question is startling: what is the value of death? What if death is the key to change and progress, and the staying of death’s scythe will mean that society slowly devolves into stasis?

There is a market here, of course. A quick glance gives at least the following possible segments.

But there will be many more. The challenge will be to figure out the macro-economic effects of the dead entering the production function. We can imagine a number of different ways in which this will happen:

All of these require careful analysis, and there is a whole research project here: figuring out the institutional economics of death, after life and the reanimation of the deceased. Will the dead crowd out the living in the labour market? Will the intangible capital stock compound in entirely new ways?

These analyses then need to be connected to our wider policy questions, and explored.

And ethics?

You will note that I have not touched on the ethics of summoning the dead in this way, but there is a reason for that. I think that this will happen, because the allure is too strong for us to resist. We should discuss and debate the ethics, and that discussion will underpin the policy choices we make - but we should also be aware that the ethics decision is not a binary one: it is not a choice of if we engage in this practice, but a question of how we regulate this practice. And yes, that means that I subscribe to some kind of determinism here, one rooted in our species’ soft narcissism in the face of technologies like AI: we want to see ourselves in the mirror of AI, and now when that mirror can transcend the end of life we will want to explore that too.

In one of the earliest stories about this trend, it was reported that Alexa, Amazon’s AI assistant, would be able to read stories in your dead grandmother’s voice. We first recoil in shock when we read that, but today? I think we are seeing a rapid normalization of our new relationship with death - so why not tackle it head on and see what we need to do to ensure that this new practice is regulated well?

Thanks for reading,

Nicklas