Interestingly the repeated critiques in the article are about sensor richness: primarily force feedback and tactility, which indicates lacking hardware. Software only robotics has a long and fraught history, but it really feels to me that current industrial hardware could be driven more intelligently without much change. No doubt the "ideal" robot for any given task requires developments in both.

I'm also curious about safety, since generally capable mechanisms need a multilayered safety stack that includes semantics, and cobot certification is likely not enough anymore. Examples: feeding someone the wrong pill, pouring a glass of water into electronics, cutting vegetables vs fingers.

[1] https://substack.com/redirect/82d94852-76b6-4b0d-8595-86e46a...