2 Comments

nice work. yes, most of the problems do seem like they are operationalization problems. especially in the era of off the shelf hardware and teleop, you can get to a robot thats economically productive much quicker than before.

however, this:

"It means grappling with a seemingly infinite number of edge cases that were not identified in the lab or initial POC. It can mean starting this process from scratch for every new customer."

is not a operationalization problem. it's a problem that ostensibly foundation models can solve. i argue this is more of a software problem.

you should keep writing about robotics

Expand full comment

Love this article, but disagree on a few narratives. Will keep them very surface-level, but happy to dig deeper.

1) Generalist humanoids are just that... generalist. I firmly disagree with the approach of 1x, Figure, etc as the humanoid form is NOT the optimal way to maximize efficiency in warehouse or lab settings. However, the end goal needs to be putting humanoids into everyday households. The only form that can perform the ~30+ household tasks without adjusting the environment is a human form.

2) Just b/c it's hard and hasn't been done yet is no reason for why it'll be a stable curve in progress moving forth. Majority of issues relating to reasoning surround computer vision and the ability to extract data in real time. We're so close to solving this.

3) Hype doesn't correlate with progress. One can argue companies targeting AI verticals like Harvey and Cognition are currently all hype. However, it's merely the correct play to capture market share. Figure with their BMW contract is a great way to ensure distribution channels upon building an actually viable product. I wouldn't read too much into it.

Expand full comment