1. One other element I'd add about our hands is that they're capable of self-repair from accumulated damage and wear-and-tear. Robot hands can not do that - they might struggle to survive with acceptable functionality at a cost-effective rate, especially if they're extremely complex to try and make them match human capabilities.
2. I didn't know about the overheating thing. That's brutal - if it generates a lot of noise, people aren't going to have those things running around in the house anymore than you'd tolerate a vacuum constantly being on at all times within 10 feet of you.
One can't ignore the massive deployment of robots in China. I agree that demos are cherry picked, but many Americans have visited China now and noted the ubiquitous adoption of autonomous robotic systems in many domains.
The amount of capital being deployed on emerging tech has never occurred before, and consistently articles that point out limitations age poorly. If there was a aspect that would actually limit robotics it would be supply chain bottlenecks and government intervention. I'd say that would the primary limit.
Generalization I feel is the most important -- continual learning. The physical world needs continual learning far more than the virtual world of video, images, text and programming language to accomplish equivalent success. But an architecture that updates its weights in real-time in response to environmental successes and failures.. oh my. That's a tough one.
Great post. Also the sheer efficiency of human energetics, metabolism and heat dissapation can not go understated. The fact humans can get up eat a bagel and coffee and have all the energy they need to operate for several hours is nothing short of amazing. Further humans also use that miniscule energy for ongoing self repair. Evolution has been optimizing in this domain for millions of years. Good luck replicating that. That is not to say that humans will gain from a synergistic relationship with mechanical tools/ie robots but I think we're a long way from replicating all things human biology.
Your obstacles decode as system functions. Not all of them must be solved by the robot. Some obstacles vanish if you change the environment: dexterity doesn't matter if you redesign the workspace. Generalization is less critical if you narrow the task. Safety becomes manageable if humans supervise until confidence rises.
Where should we push friction? Is it environment standardization? Regulatory structure? Liability models? Labor market dynamics? The clarity of ROI?
What are we missing about why partitioned-work robotics scales in warehouses but hasn't translated to humanoids or care robotics?
For sure, there will be deployments where tasks and environments are redesigned to accommodate robotic limitations. That's an accurate description of ~all robot deployments today, including of course traditional industrial robot arms.
However, this is just another way of saying that there are jobs that robots aren't ready for. The greater the accommodations the robot needs, the slower adoption will be and the smaller the addressable market.
Generalist ( company name ) is making some inroads on the coordination, planning, generalizing actions and even translating instructions to actions, via the use of "physical prompting". I don't think it makes robotics less hard, but moderate success at one shot and even zero shot actions is encouraging.
Your waymo case is interesting for several reasons.
1. Humans also drive into standing water and destroy their cars and or drown - all.the.time.
2. Even if humans are better with edge cases what they aren't better at is mundane daily driving. Not a week goes by that I don't see someone, probably on their phone, slam hard into the car in front of them.
Very good piece!
1. One other element I'd add about our hands is that they're capable of self-repair from accumulated damage and wear-and-tear. Robot hands can not do that - they might struggle to survive with acceptable functionality at a cost-effective rate, especially if they're extremely complex to try and make them match human capabilities.
2. I didn't know about the overheating thing. That's brutal - if it generates a lot of noise, people aren't going to have those things running around in the house anymore than you'd tolerate a vacuum constantly being on at all times within 10 feet of you.
One can't ignore the massive deployment of robots in China. I agree that demos are cherry picked, but many Americans have visited China now and noted the ubiquitous adoption of autonomous robotic systems in many domains.
The amount of capital being deployed on emerging tech has never occurred before, and consistently articles that point out limitations age poorly. If there was a aspect that would actually limit robotics it would be supply chain bottlenecks and government intervention. I'd say that would the primary limit.
Great article! Very comprehensive, it will make a good checklist to look back at.
https://www.understandingai.org/p/why-humanoid-robots-wont-catch-up
(Kai Williams)
I like your article better though.
Generalization I feel is the most important -- continual learning. The physical world needs continual learning far more than the virtual world of video, images, text and programming language to accomplish equivalent success. But an architecture that updates its weights in real-time in response to environmental successes and failures.. oh my. That's a tough one.
Great post. Also the sheer efficiency of human energetics, metabolism and heat dissapation can not go understated. The fact humans can get up eat a bagel and coffee and have all the energy they need to operate for several hours is nothing short of amazing. Further humans also use that miniscule energy for ongoing self repair. Evolution has been optimizing in this domain for millions of years. Good luck replicating that. That is not to say that humans will gain from a synergistic relationship with mechanical tools/ie robots but I think we're a long way from replicating all things human biology.
Nit: the article headline and URL say 14 reasons and the browser tab says 15.
Oops, thought I'd fixed this – thanks!
Thanks Steve for this fine piece.
Your obstacles decode as system functions. Not all of them must be solved by the robot. Some obstacles vanish if you change the environment: dexterity doesn't matter if you redesign the workspace. Generalization is less critical if you narrow the task. Safety becomes manageable if humans supervise until confidence rises.
Where should we push friction? Is it environment standardization? Regulatory structure? Liability models? Labor market dynamics? The clarity of ROI?
What are we missing about why partitioned-work robotics scales in warehouses but hasn't translated to humanoids or care robotics?
For sure, there will be deployments where tasks and environments are redesigned to accommodate robotic limitations. That's an accurate description of ~all robot deployments today, including of course traditional industrial robot arms.
However, this is just another way of saying that there are jobs that robots aren't ready for. The greater the accommodations the robot needs, the slower adoption will be and the smaller the addressable market.
Generalist ( company name ) is making some inroads on the coordination, planning, generalizing actions and even translating instructions to actions, via the use of "physical prompting". I don't think it makes robotics less hard, but moderate success at one shot and even zero shot actions is encouraging.
https://generalistai.com/blog/gen-1.5
Your waymo case is interesting for several reasons.
1. Humans also drive into standing water and destroy their cars and or drown - all.the.time.
2. Even if humans are better with edge cases what they aren't better at is mundane daily driving. Not a week goes by that I don't see someone, probably on their phone, slam hard into the car in front of them.