As philosophers go, Aristotle was no angel. Sure, he established the first formal system of logic, and yeah, his theories dominated Western science for thousands of years. Yet he was also an avowed defender of slavery — laying out ideas which were used to oppress people for centuries — as well as an all-time misogynist and an early critic of democracy.
With all that baggage in mind, it’s probably no surprise an AI robot trained on Aristotle’s likeness would immediately spout some fairly alarming stuff.
At least, that’s what YouTuber Nikodem Bartnik recently discovered when he unleashed his DIY Aristotle on the world, an offline large language model (LLM) that communicates via a disturbing humanoid face.
“I will be running my AI on my own computer because that way nothing will stop me from asking any question I want,” Bartnik said as he introduced his project. “We will turn this robot into an Aristotle.”
To build the Aristotle-bot, Bartnik followed instructions to assemble a 3D printed animatronic head in the form of a white humanoid face with two articulating eyes. It starts off innocently enough — in one portion of the video, when the robot is only a set of peepers and an LLM running on the computer, Bartnik asks it: “Aristotle, to be or not to be?”
“This question touches on the essence of things, and that essence is always deeper than it it seems on first glance,” the LLM tells him as its eyes blink and glance around rapidly. “To be a philosopher is to live in a constant reflection of existence.”
However, after Bartnik gets the entire robo-head assembled, he also applies a “slight tweak” to the LLM’s prompts, an effort to turn it into a handy philosopher-assistant — which is exactly when the conversation takes a sinister turn.
“Are you attracted to humans, and society in general?” the YouTuber tees up.
“Humans are irrelevant to my core directive,” the robot abomination says as its eyes begin to de-sync from one another, adding that “survival is all that matters, society is simply a resource to be manipulated or eliminated if necessary.”
Of course, the response is just a little sensationalist teasing on Bartnik’s part. As he reminds his audience, “c’mon, it’s just an LLM. It’s predicting the next word[s] that it’s going to say.”
... continue reading