Sure we could argue semantics, but to me putting together off the shelf software and hardware isn't "building a robot", but maybe that the vernacular these days. None of that is from scratch.
It seems that its being billed as nearly ready to go. > we're building a robot that comes in kiosk form and lives in airports, malls, and offices.
Just let my comment stand for the man-on-the-street view, ignorant of the technical specifics. No way am I letting a robot shave my head.
They're just proving out the idea - it makes perfect sense to buy/rent/borrow a COTS robot arm and use commodity hardware (iPhone X 3D scanner) to demonstrate that there's potential and that the software can be made to work. What follows are progressively higher and higher fidelity prototypes to eventually converge on a marketable solution (maybe with a custom mechanical solution rather than a generic 6DOF manipulator - several iterations are required for sure).
Hardware is a long road, something like this takes years. They are doing exactly the right thing by producing a basic demo - it doesn't have to be pretty, it just has to show that the idea has the potential to work. After this they can go and try to raise some money for the next step.
I think there is real under-appreciation by a lot of software engineers on HN regarding how much hardware development actually costs in dollars and time terms. The robot arm in the video costs north of $20k. The prototype phases between here and a consumer product will probably consume tens of thousands if not hundreds of thousands of dollars in materials and likely a few million in labour. It's not at all like firing up a pretty-looking SaaS MVP with a few months of labour and some negligible AWS costs, and I think the HN crowd often forgets that.
Cutting hair is none of this: your position constantly changes, and the stuff you're operating on is irregularly shaped and constantly moves and tilts and turns. You need to manipulate small tools relying on a mix of visual, touch and force feedback, and you need to interact with hair, which is probably difficult to model.
I don't know much of surgical robots, but I assume they're either executing pre-planned operations with a single tool on a completely still subject and/ or they're being controlled by a human surgeon.