Woolly mammoths went extinct thousands of years ago, but they didn’t disappear without a trace. They left behind frozen samples of soft tissue and other bits of DNA — biological artifacts that scientists are now using to try to revive the species.
Like the final prehistoric mammoth, whose death marked the end of the species’ initial incarnation, each of us will someday take our last breath, and our somatic life will draw to a close. We, too, will leave behind fragments of DNA, and perhaps someday our genetic code will be run on new fleshy hardware.
But I’m more interested in a different type of DNA we’ll leave behind when our human bodies die. A variety of DNA that’s already being used to resurrect the dead so that they might teach, entertain, and otherwise interact with us: our digital DNA.
Everything we write, speak, create, publish, post, and comment that is recorded digitally can be used to recreate an agent that is just as much “us” as a genetically cloned hunk of meat is a woolly mammoth. While the mammoth is recreated from its genetic code, the “us” agents are recreated from our memetic code. Imagining our online data as a form of DNA isn’t that big a leap — after all, DNA is just data. Information. And since humans are mental creatures, our ideas — the information we have produced and disseminated — are perhaps better than our genes at encoding who we are.
Why does this matter?
It means that every seemingly inconsequential digital footprint you leave actually has incredible significance. Whenever you post a tweet or a cast, record a podcast, publish a blog post, or make a comment, you’re contributing to a collection of data that will determine your digital legacy. That information, and nothing else, will become the fodder for a generative AI that will learn what it means to be you. Future generations will think of that AI as you. Your physical body will live for a single generation, but this program could live for eternity.
A few weeks ago, a simulated podcast interview between Joe Rogan and Steve Jobs was released. Listen closely, and you’ll notice that Rogan sounds playful and casual, while Jobs sounds more oratorical. Why? Because the data points we have on how Jobs spoke, the components of his digital DNA that codes for his vocal patterns, are sourced primarily from recordings of speeches and presentations he gave. Rogan, on the other hand, has produced mountains of audio with a much more conversational tone and style.
Figures throughout history have recognized that their writings and art would serve as perpetuations of their intellect. But they never expected that these data points would be used to seed entirely new beings that would act as their immortal cognitive proxies. Ideally, we will continue to view AI representations of real people as imperfect impersonators, but as the models become more advanced it will become harder to remind ourselves that they are not entirely faithful representations of real people. Each of us is well aware that our own digital DNA is not fully representative of who we are, but the limits of human psychology mean that it’s difficult for us to apply this same open-mindedness when considering AI likenesses of others.
Is it fair that my opinion of Steve Jobs might be influenced not just by things he said, but also by things a lifelike AI trained on his digital DNA says? When generative models become indistinguishable from real people, will we care whether quotes from GPT-Ghandi are actual statements made from the man himself or something the model decided he might have said?
“They say you die twice. One time when you stop breathing and a second time, a bit later on, when somebody says your name for the last time.”
(— Numerous sources)
This quote reminds us that one day, we will be forgotten. But perhaps a greater concern is that we might be misremembered.
- Loading comments...