Digital immortality and Augmented Eternity

Digital immortality or duplication. This is very interesting for me. There have definitely been times I have wanted a virtual version of myself to do all my work for me as if I was actually doing it (especially when it comes to studying ??). A post on MIT Technology Review talks about what Hossein Rahnama is trying to achieve with his application 'Augmented Eternity. The idea is that once a person has passed away, an augmented version of themselves will exist to interact with people.

Now, into the specifics. The blog covers various points and I want to talk about all of them. Firstly, the matter of the training data used to create these versions. For something to accurately predict what a human being would say, billions of parameters are required (please refer to my previous post on GPT-3). Even then, the text generated feels awkward or untrue. The data that is used to train these versions would be our digital footprint. All our tweets, Instagram posts and captions, google searches etc. can be used to create a similar persona. While this may be a lot of data for recent generations, many of the older generations whose lives aren't documented on the internet would not be able to create even a slightly similar persona.

Secondly, when the data being used is our digital footprint there are concerns that need to be addressed. Let's take the example of texting your friend versus talking to them face-to-face. While texting, the emotions with which the text was written are often hidden as compared to face-to-face, when even the tiniest micro-expression on the other's face can clarify their emotions. That takes us to 'context'! People often take actions on the internet that are very mood-dependent but the data doesn't show the emotion, only the numbers. Say, I had a pretty bad day and I wanted to de-stress so I bought many clothes online. The data will register numbers that have no mention of my mood. The Augmented Eternity platform is aiming to overcome this. It 'takes data from multiple sources—Facebook, Twitter, messaging apps, and others—and analyzes it for context, emotional content, and semantics.'

Thirdly, this tool would only be effective as a decision-suggesting mechanism. It will never be able to completely replace the mindset of the original person and thus, we will never be able to fully trust it. While it may have information on all the previous decisions taken by the person, it will simply run numbers to try and generate an accurate enough response. Another point is that time matters. Let's say a person passed away twenty years ago and you ask its augmented version for advice on a career choice. The decision it generates might be accurate for you... twenty years ago. The career choices have changed and so has the market's demand, and this persona will not be able to adjust for it.

Next, this tool need not be only for people who have passed away. With the coronavirus, it can be a way for celebrities to meet many people and give them realistic enough responses through their augmented version. Instead of asking a question to Siri or Alexa, you could ask an eminent personality for their opinion. The AI could convert your professional knowledge from a scattered written record to an augmented persona with your knowledge that people can interact with. 

Lastly, ethics. When a person passes away, to whom does the digital footprint left behind belong to? Who owns that information and who, accordingly, has the right to generate this augmented persona? Rahnama said that he felt that Augmented Eternity was important because it raised a lot of questions about data and ethics in data analytics that need to be clarified.

This also falls into the category of 'fascinating, but I'm going to keep my distance for a bit'. I think the last point on ethics needs serious deliberation and consideration before even opening up this bag of worms. While I have covered the topic of digital immortality, I would like to research many more sub-points regarding this.

Thanks for reading!