What Actual AI Builders and Black Mirrors are doing fallacious with the digital resurrection
One day we will all die. Science and technology can put it off for a while, but the passage of time doesn’t stop for anyone. Unfortunately, most of us will be forgotten. It’s a dire forecast, but it always has been. And that is unlikely to change, despite the best efforts of the AI community.
There’s a new technology trend (that’s actually a stupid old trump card) that is making its way through big tech, small tech, and South Korean television networks: the digital resurrection.
The premise is simple. A person living in the modern world leaves tiny marks of who he is in everything he does. If you will, our “digital footprint” has grown so large that we are producing enough data for clever AI to imitate us.
For example, Microsoft was recently granted a patent called “Creating a Chatbot for Conversation with a Specific Person”.
We have seen similar AI systems like this one that, given some articles by a particular author, can mimic their style. When you set this setting to 11 and envision an AI trained on thousands of personal texts, emails, and transcribed voice messages, it becomes easy to see how such a paradigm can be used to create a bot that calls a person imitated … alive or dead.
[Read: How this company leveraged AI to become the Netflix of Finland]
The popular British TV series Black Mirror aired an episode in 2013 called “Be Right Back”. The show took the idea of a chatbot trained on the deceased’s data and basically added a futuristic 3D-printed android that became a real-life embodiment of that person. The big idea is that such a robot could help people find a degree, especially if they unexpectedly lost a loved one.
But as is so often the case, the reality is very different. In the real world, we’ve seen the digital resurrection almost exclusively as a marketing tool or gimmick. And that’s because AI, no matter how much data it contains, cannot “recreate” a person in any meaningful way.
Remember the holograms of Tupac Shakur and Whitney Houston, or the additional blatant addition of Kurt Cobain in Guitar Hero 5 as a playable character who could be forced to play every song in the game.
More recently, Korean television broadcaster SBS has unveiled plans for a game show in which people will sing duets with an AI replica of pop superstar Kim Kwang-seok, who died in 1996. A Spanish beer company called Cruzcampo recently used a deepfake-generated version of the late, legendary singer Lola Flores, who passed away in 1995 as part of an advertising campaign.
There’s a reason you’re more likely to see singers replicated than someone better known for their speeches and ideology, like Winston Churchill or Dr. Martin Luther King Jr., because the AI can’t really predict what a person would say or do in any given situation, no matter how much data it has.
The problem with the digital resurrection is simple: AI can’t do anything a sane, average person can’t do if given enough time. Humans can imitate other people by dressing up as them, copying their mannerisms, and mimicking their voices. But we cannot read the other’s mind. We can only guess what someone else is thinking. And AI is no different. No amount of data in the world can predict what a person will do (or would have done if they were still alive) next.
At best, we can preserve and animate a particular developer’s vision of what a sentiment analysis of your dead loved one or a celebrity might look like.
If your late spouse liked to say, “I love you, cuddle bunny,” at the end of every email, an AI could learn to mimic their unsubscribe. But if your spouse has some gold hidden somewhere and has never written or spoken about it to anyone, no AI can tell you where the money is. Human psychics cannot predict winning numbers for the same reason.
And this is where Black Mirror and the real companies trying to sell the idea of a “digital you” are doing something wrong. It’s easy to polish up a two- to three-minute clip of a famous actor doing what he’s famous for. All in all, making a slightly convincing human replicant yourself is a different matter.
People instinctively look for imperfections in other people because historically this has been essential to our survival. We are easily fooled when we are entertained and have temporarily removed our feelings of disbelief. But as Hollywood and the world of computer-generated imagery have learned over the past few decades, in imitating life, it is easy to convince many people from a distance but almost impossible to convince a person up close.
Whether you believe that humans have a soul that drives and motivates them, or you understand that scientists know very little about how the human brain actually manifests consciousness, it is currently inconceivable that a machine could literally be imbued with that what makes each one of us unique.
Just as a vinyl record cannot convey the gravity of a live performance, a digital record cannot take the place of a living person no matter how well it was received. If you’ve never seen Michael Jackson live, the only way to get a taste of what it was like by watching footage or listening to a recording.
No matter how powerful the AI gets, it can’t tell us what the late singer would have come up with next. I’m assuming that no human or no AI could have predicted that “Thriller” would be the sequel to “Beat It”.
And the same goes for your loved ones when they come over. An AI that mimics it is no more accurate or powerful than just asking someone to impersonate: it is not reality, no matter how skilled the impersonator is.
Like most heroics with predictive artificial intelligence, the digital resurrection is little more than prestidigitation.
Published on January 26, 2021 – 20:17 UTC