Eventually, Will Generative AI Win the Death?
If it’s all about calculation power, AI will reproduce humans sooner or later, regardless of the required energy.
Think of the more and more boundless amount of data we put daily into social media: texts, images, and videos; moments and thoughts, the important ones and the trivial ones: all public on the internet.
Since humans write, they have journaled. But it was private. Now it’s public. And generative AI can learn, from social media, everything they need to reproduce a believable version of posts we post.
Sure, Generative AI cannot reproduce a human body, for now. Yet, think of people who rely on mediums and psychics to connect with loved ones who are gone and hear their voices once again. They don’t mind that the body is not there. They want a sign. That’s it.
The energy required to reproduce dead people en masse might be forever unsustainable, so I imagine a future that cannot exist for physical reasons; unless we discover a way to get enough energy to achieve what now seems unattainable.
Nevertheless, it’s realistic that generative AI can reproduce some people and make them live forever in some way. I cannot calculate how many people some could be, but at least one is reasonable.
The undiscoverable dead
Imagine you usually communicate with someone — let’s call him James — far from your home, so the communication goes only through chat, video calls, and similar ways.
One day, James dies, but you don’t know, and a generative AI replaces him in the communication channels. Perhaps, it could be fairer to say that generative AI continues James’s communication habits.
You continue reading his messages, listening to his voice, and watching him during video calls. Sometimes, James states weird sentences — you notice at some point — but who is perfect? James is getting old and strange — you think.
You’re not the only one communicating with James. Many colleagues and friends continue chatting with him.
Nobody knows that James was part of the project “The undiscoverable dead”: he accepted to train a generative AI, called James-AI, with everything he ever posted on the Internet. He installed it on all his devices: laptop, visor, smartphone, tablet, etc. After about 50 years of training, when James died, James-AI was ready.
James will appear alive until the people who run the project “The undiscoverable dead” maintain the hardware and software behind James-AI. But for colleagues and friends who communicate remotely with James-AI, James is alive.
Even when James was still alive, sometimes James-AI communicated on his behalf. It was a progressive substitution. When James died, and James-AI remained the only one, the communication style didn’t change much. Or it didn’t change at all.
For you, James is alive. Until you die, in turn, and never knew that James passed away. But what if you participate in the project “The undiscoverable dead,” too?
Turing test passed: death overcome?
I’m assuming that generative AI passes the Turing test: the ChatGPT experience shows that my assumption is reasonable. So what? Is death overcome?
The answer lies in the difference between some and all.
In the scenario above of the project “The undiscoverable dead,” only remote people can perceive James still alive after his death. They are some people. For some people and for some time, James didn’t pass away.
On the other hand, if we ask whether James is still alive for all possible people, the answer is no. Someone, at a certain point, discovers that James’s body no longer exists.
The simple answer to the question “Turing test passed: death overcome?” is no.
Conclusion
Man’s forbidden dream is to live forever, and with all its limitations, generative AI opens a crack at the question.
Biologically, nothing changes. We still are the same as thousands of years ago. From this standpoint, generative AI is irrelevant.
Culturally, something changes. Automation — at the end of the day, generative AI is nothing more than automation — simulates the ability to share information human-like. The consequence is that you can’t know if a human or a machine crafted what your read or listen to or watch.
Is this article written by a human (me) or by a generative AI? An honest generative AI would declare it’s the source of the text, but honesty means nothing to a machine. That’s why simulating a dead person is a task like another.
But it’s a deception.
