Atlas replacing human workers? Nope.
It’s interesting that Boston Dynamics is partnering with Google Deepmind to use AI on their Atlas robot, but this is merely hype and side show. The consensus among knowledgeable observers whose jobs do NOT depend on robot and AI hype is still that humanoid robots are fundamentally unsuited to replace humans in most jobs, and certainly not in any meaningful numbers. Specialized non-human-shaped robots are where the practical applications lie.
As a side note, Boston Dynamics is notorious for publishing craftily-edited videos to support unrealistic claims of performance and reliability. Their claims for Atlas should thus be examined with a critical eye, which most press accounts sadly fail to do. (Look at any BD video; each time the camera cuts to a new scene, it’s likely they’ve spliced out Atlas taking a pratfall. Continuous takes of Atlas doing practical things are rare.)
Humanoid robots as competition for human workers are a dead-end path for several reasons, which come down to basic physics and economics and thus are resistant to technological quick fixes, and totally immune to the application of AI fairy-dust.
TL:DR; Humanoid walking robots are prohibitively expensive, severely limited in their real-world capabilities, suffer short battery run time, and dangerous to be around because they may fall on you or accidentally run into you.
1. Any practical humanoid robot will cost hundreds of thousands of dollars to build, and will provide performance results inferior to human workers.
2. A two-legged walking robot requires very precise, complex, and expensive mechanics, motors, and electronic systems to be able to walk at all; these add weight and use a lot of electric power.
3. Humans handle arbitrary objects with trivial ease, but robots cannot. No existing robot “hand” can even approach the dexterity of the human hand; a robotic gripper that can pick up an egg, a bag of rice, and a brick costs tens of thousands of dollars to build and is large, power-hungry, and marginally reliable.
4. Battery life for such a robot is severely limited. Batteries are heavy and limited in storage capacity. This only gets worse as you add bigger batteries and more powerful computers to handle navigating in chaotic environments humans take for granted. Practical battery life for a humanoid factory robot is generally less than an hour with present technology.
5. A walking humanoid robot is a mobile safety hazard for the humans that work around it. Walking robots fall over a LOT (just look at the DARPA Robotics Challenge videos), and a 200-pound metal robot falling on a human coworker will injure them. The more safety features you build into that robot, the slower, heavier, and less capable it will become. In a factory, they would be walking lawsuit magnets.
5. Specialized, single-purpose industrial robots that are not made in a humanoid shape are already in wide use. They are far more useful, much less expensive, and safer to work around than humanoid robots.
For further reading, I recommend Rodney Brooks’ essay on the subject: https://rodneybrooks.com/why-todays-humanoids-wont-learn-dexterity/
Steve Hersey
Waltham, MA
