Afterlife 2.0 - Posthumous AI

Like two million Australians, I watched an increasingly rare moment of linear, live television – The AFL Grand Final. As a long-time Collingwood fan, I rode every bump and felt more invested than a day trader. I rejoiced in an eventual victory and immediately thought about how much my late mother would have loved seeing her Magpies win another flag. We could've watched it together, followed by hours of post-game discussions, analyzing it play-by-play. While I'll never watch another game with my mother, the prospect of discussing it—or anything else with her—is closer than we think.

How to Create a Digital Twin

It won't exactly be her. She'll have no idea about it, but soon we might be able to 'access' the minds and faces of those who've passed.

Initially, it won't be perfect. The process largely involves constructing a 'persona bot'. End users wishing to create a bot of a lost loved one will need to provide detailed information, effectively creating a databank for the AI bot to draw from. This information encompasses obvious factors such as geographic and chronological details, education, work experiences, travel, interests, and even attitudes. With this data, a fairly accurate replica can be made using off-the-shelf AI tools. The bot will be able to engage in generic conversations. In fact, a version of this was achieved years ago when James Vlahos created a chatbot of his terminally ill father. He conducted deep interviews for months, resulting in a basic DadBot he could interact with posthumously. We've progressed significantly since then; interviews are now optional.

Currently, anyone owning a smartphone — which accounts for 86 per cent of Australians and 63 per cent globally — is inadvertently creating a Digital Twin of their existence. ChatGPT can adeptly replicate the voice of Steve Sammartino, mimicking my writing style. Given that much of my life is public, it possesses an extensive database of not just my work, but my experiences, acquaintances, and hobbies.

The younger an individual, the higher the 'resolution' of their twin. Nearly every action since the inception of the smartphone has been digitized and documented. Both my children, born in 2010 and 2012, epitomize this. Their digital footprints reside in server farms globally. Even though it's not 'public domain', tech giants possess access to all components of their life stories.

Consider what today's youth have shared for their future digital twins:

Photos: Most pictures of my children, taken by my wife, originate from an iPhone. Stored in iCloud and another cloud provider, these images are geotagged, timestamped, and analyzed by AI.

Videos: The data obtained from videos is even more detailed and accurate than pictures, and importantly builds a database of our voices, facial and body movements. All of which are as unique as fingerprints.

Diary: Our family's whereabouts have been meticulously logged. Even if we forget an entry, GPS and Google Maps fill in the gaps.

Phones: My daughter's iPhone tracks her content preferences, friends, and interests. My son, though phone-less, is still monitored by big tech when he uses my devices and they create a shadow profile of his clear habits and preferences. 

Laptops: Both kids use laptops for schoolwork. This offers insight into their intellectual growth and thinking patterns, much like a school principal from a smaller primary school might track and take an interest in students’ development. 

And we can be sure that big tech are tracking all ambient conversations of anyone in earshot of a smartphone, and know exactly who is speaking, despite claiming otherwise. 

This is just a glimpse of the data available. For those online for over a decade, we can create a high-definition AI replica. And for those born post-2007, it'll be nearly perfect. This means we can craft Interactive AI Twins of the living or deceased. When it does arrive, we’ll all be astounded, though we shouldn’t really be surprised. 

The Ingredients

Combine the aforementioned personal data with a Large Language Model trained to mimic human behaviour, and we can produce a convincing deep fake of almost anyone. It's like a Zoom call straight from the afterlife. It will be as simple as logging into and choosing who to chat to. In moments a life-like AI bot of a person you once knew will be there conversing with you in real time, with the same ‘brain’ and life they had before they passed. The kicker is that they’ll continue to learn as you chat with them. Adding what new experiences you’ve had and keeping up to date with everything going on in the family.

Of course, you’ve already guessed what the red flags are. 

Who owns our data and biometric records? Who does the person who died need to sign off on their AI replica being created in their will? Who profits from all this? And is this even a healthy idea for people who struggle to move on in such difficult circumstances?

If we've already relinquished our data to Big Tech, will we need to pay subscription fees to converse with AI renditions of departed loved ones? Perhaps this tech will be reserved for forming parasocial relationships with AI Taylor Swift or our favourite actors?

One certainty is the emergence of a significant industry. The allure of fantasy, love, and loss is undeniable. And it is the usual suspects who are in the best position to be able to do this: Apple, Meta and Google. How could anyone turn down the opportunity to be able to talk again to someone once close to your heart, even if you know it isn’t real? 

One thing that will be real, is the profits. The only question is: who will accumulate them – Big tech, yet-to-be-founded startups, or the people who created the data in the first place?