Why ‘home robots’ are a lot further away than you think

It’s that time of year again. CES is almost here and that means we’re about to receive on onslaught of new robot technology. Most of it’s going to be awful, some of it’s going to be amazing, and all of it’s going to be over-hyped because, hey, it’s Vegas.

Based on the contents of the dumpster fire that is my inbox, home robots are going to make a huge splash at CES. And, following a year in which Elon Musk debuted a spandex-clad human as a prototype for the Tesla Robot, we imagine some of the claims are going to be pretty darn silly.

Unfortunately, the glitz and glamour of CES can make it difficult to discern fantasy from reality. And nobody wants fully-autonomous domestic robot servant companions more than the crew here at Neural.

So, when we throw cold water on the whole situation — as I fully intend to do over the next few hundred words — it’s done from a place of empathy.

We, as a species, are nowhere near having the technology it would take to create a humanoid robot that’s capable of functioning as a general laborer or domestic servant.

The lab

Boston Dynamics has dozens of machines that can perform human-level feats of physical agility. Videos of Atlas models doing backflips and running parkour courses have gone viral for years. So why hasn’t anyone at the company ever thought to put an AI brain inside an Atlas and sell it as a butler? Are all the big tech and seasoned robotics companies just too dumb to see the opportunity?

The reason is simple: no such AI brain exists. It’s one thing to build human-sized machines capable of running amok in a laboratory. It’s another to create a robot that’s human-sized, useful, and safe to operate in the wild.

The reality

Useful machines don’t rely on puppetry to get the job done. If you want to build a robot that changes light bulbs, from an engineering point of view, it doesn’t make sense to give it a face or fingers. Chances are you’d be better served mounting a robot arm with a bulb-gripper on an accordion lift.

For the same reason, we wouldn’t build a robot with arms and legs that was designed to pick people up and carry them from point A to point B like bundles of laundry. We’d just automate a car or build a vehicle on rails. 

If you need a light bulb changed, the optimal solution is to get a human to change it. If you need a million light bulbs changed across myriad configurations and environments, it’s probably optimal to get a bunch of humans to change them. Because there’s no “general” artificial intelligence capable of changing any light bulb in any environment.

But if you need the exact same light bulb changed a million times, then that’s a great task for a modern robot.

Creating a robot that can change a light bulb in any random consumer’s home or workplace would require artificial general intelligence (AGI), which is also known as human-level AI or the holy grail of computer science.

Simply put, any attempt at replacing human general laborers with robots is a novelty act at best. 

As much as we want to believe we’re on the cusp of having robot companions and servants, we’ve got a long ways to go before neural network technology is robust enough to go from “Hey Siri, play Lil Nas X,” to “Hey Alexa, go gather up all the dirty laundry.”

Barring aliens visiting our planet and gifting us with robotics technology from the future, or one of history’s greatest eureka moments happening, 2022 won’t be the year you finally get a real working version of Rosie The Robot.