AI in Humanoid Robots: Not an Evil Empire

While carrying Darth Vader across the parking lot, one of his hands fell off. I reattached it once he was standing in our office, but since it was an old styrofoam mannequin under that menacing suit, the arm needed to be angled up like it was in a cast to make sure the hand didn't fall off again. Looking at the lights and buttons on the front, I wondered what operating system does Vader run on anyway? Turning to the dark side of our bookshelf, I grabbed the obvious book and took his picture with it. So there he stood, keeping an evil eye on me and my colleague while we wrote visual effects software, a long time ago...

But until recently, I hadn't given the technical folk lore of Vader a second thought. With Anakin dismembered and burnt to a bust statue, would the resulting cybernetically enhanced Sith Lord be possible with the currently emerging exponential technologies? As humanoid robotics companies multiply, and bionics companies build on breakthroughs in biomedical technology, a humanoid robot with a human brain, like Vader, seems within the realm of possibility.

It's interesting that the general public has assumed all along that AI == Robotics, though. Movies, videos, and news articles about the dangers of AI commonly depict humanoid robots as the existential threat, even though the ominous AI (AGI) under scrutiny is largely disembodied. In the robotics engineering world, AI == Robotics is not true; hardware is hard. Atlas' 15 years of fine tuning control algorithms, plus human choreography that would make Jackie Chan proud, makes it look easy--but AI it ain't! Similarly, teleoperated robotics, exoskeletons, robotic avatars, and manufacturing robots: all Robotics but not AI. Darth Vader’s extreme bionics would be a form of human-in-the-loop robotics too.

In contrast, a humanoid robot with a human-like brain (AI) is the goal of several companies building general purpose humanoid robots (Sanctuary, 1x, Figure, Agility, ...). With the current AI El Ni?o storm raging, does the general public finally have good reason to fear the end is near with AI in humanoid robots becoming a reality? Not really.

Suppose that Large Language Models (LLMs) plus Ontologies are the answer 42 for embodied AI. Then what is the analogous character to compare to an embodied Vader? "I am C-3PO, human/cyborg relations, an embodied Large Language Model fluent in over six million forms of communication." I think Vader vs C-3PO is an especially relevant comparison today, that might help allay fears of the perceived existential threat of AI embodied in humanoid robots.

On one hand, you have Lord Vader, a humanoid robot with a worst-case human brain, "More machine than man"--ish. It's a particularly demented human brain, twisted and evil, not unlike some world leaders today. As Brene Brown put it, "We are not thinking machines that feel; we are feeling machines that think." Emotional damage, chemical imbalance, and whatever other human brain problems Vader has, it's not the kind of brain you want in a robot, much less a large population of them. Yet, that is what many people perceive as inevitable if you put AI in robots instead.

On the other hand, you have C-3PO, a humanoid robot with an AI brain, “How might I serve you?” Bumbling around with passable dexterity, imitating human emotional concern, yet a brilliant conversationalist, C-3PO is much closer to what we can expect than Vader. No emotional baggage, no evil revenge motives, just task-oriented goals with guardrails and planning based on deep discussions with humans.

It’s entertaining (and scary) to imagine that humanoid robots with AI brains could act like power-tripping spoiled brats like Vader. (I’m sure we’d say, “It's their human parents' fault for not raising their robots right.”) But no, I anticipate that humanoid robots with AI brains will be much more mundane: clumsy but helpful, smart but not wise, and neutral not evil. Agreed that AI in general needs adult supervision and some regulation, but the emerging C-3POs of the world, and their descendants, will help make the world a less evil place, not spell the end of civilization.