A few weeks ago, I posted about the infamous neurolinguistic AI that was trained exclusively on Donald Trump speeches. It currently spends most of its day randomly tweeting angry, boastful, nonsensical rants on Twitter. (I don’t have to finish that joke, do I?) Next up comes reports that the UK, who never really got over the cancellation of Friends it turns out, has proposed their own unholy creation: an AI-powered virtual personality based on Joey Tribbiani.
Now, the more fascinating aspect of this is that the team involved is currently working towards developing software that will scan old television shows, record facial expressions and linguistic quirks, and eventually be able to provide realistic simulations of characters in the show — ensuring that Gilligan’s Island will eventually get the movie reboot that it so desperately deserves only with simulations of the original cast so that we won’t have to argue over whether Adam Sandler or Bob Denver made the better Gilligan. The choice of Joey seemed really odd at first, but makes sense when you get down to it. You can’t use the women. Trust me. I once worked on the Iris program (the Android version of Siri), monitoring EVERY request people made of that very limited AI (which didn’t even have a visual representation beyond a blue dot). Yeah, you don’t need to mix lonely geeks with a simulated female. That leaves you with just the annoying, intellectual one (Ross), the borderline-depressive, sarcastic one (Chandler), and the lovable, funny one (Joey) — which, mercifully, they picked. Now, I’m sure they were thinking he had the most distinctive personality quirks and speech mannerisms, and was the most identifiable of the three characters, so he made for a much more exciting simulation, but really, there are real people who have had just as many screen hours of great speeches, lectures, seminars, and interviews. Is a “made up” character the best basis of reference for a simulated personality?
Obviously, yes. From a simulation standpoint, a television character has a much smaller set of rules governing their behavior. Some are written in a “show bible,” some are intangible traits brought by the actor, but considering the sum of the characters existence is a finite set of scenes and lines of dialogue, it is a much easier data set to capture. Additionally, unlike a real-life example, say William F. Buckley, who also has a very distinct personality, even more screen time and lines of dialogue, and arguably a much larger set of ideas and thoughts that go beyond a love of sandwiches and girls, the large majority of this data source is limited to sitting in a chair and discussing political theory in an even emotional tone. You could make an AI out of this, but it would be an incredibly boring one.
In addition to the quantity of behavioral examples, the idea of using a fictional character makes much more sense when you look at the simplicity of examples you receive. We generally get a wide range of emotions, moods, and circumstances presented in a very “over the top” fashion and without the ambiguities and complexities of an actual person. While real people can simultaneously process and project/conceal multiple feelings at once, if a TV character is experiencing two conflicting emotions, acting happy while secretly being sad for example, this is by necessity conveyed to the audience in an obvious fashion that is simultaneously unnoticed by the other characters. This is true from something as light and simple as Joey Tribbiani, to an Oscar nominated performance like Pachino’s Michael Corleone. (Seriously did NOBODY in his family realize how fearful and unsure of himself the Godfather was as he projected assurance and quiet confidence?!)
We’re still a couple years away from “Tribbian-AI,” to be sure, and we’re probably a decade or so out from actually bringing back Frank Zappa as a software based composer of music and espouser of sardonic wisdom. Beyond that, I look forward to the day when a simulated personality matrix is as common as a DNA sequence, and we can even mix and match which ones we want to power our devices. Personally, I’m saving my money for a Kate Beckinsale simulation with Lewis Black’s sarcastic sense of humor, Jack Kerouac’s command of the English language, and Svetlana Kolmykova’s accent.