4 ms·

Not a regular commenter, but I had to reply. I think everyone here agrees this is just a prototype/early work. Why would you be so dismissive of a tool that som

Not a regular commenter, but I had to reply. I think everyone here agrees this is just a prototype/early work. Why would you be so dismissive of a tool that someone is building from scratch? They are not claiming it is a finished product. Jesus, tone it down.

kleer001 7y ago
Why? A couple reasons. 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.
Udik 7y ago
That is never gonna work. Seriously. Cutting hair is a highly manual job, requiring precision use of a number of different tools- some pretty dangerous- on an irregular, soft and constantly moving surface. Covered in, well, long, irregularly shaped hair. The idea of a hair cutting robot is fun, but it's decades or more ahead of what we can do.
- perl4ever 7y ago
  It seems a lot easier than driving a car, and your description of the task sounds like surgical robots, which are a thing already, with superhuman capabilities.
  - Udik 7y ago
    I don't think it is easier than driving a car- somehow it feels even harder. Driving is a very well defined task: you are sitting in a precise position from which you get a well defined and purely visual input, and based on that you operate on three or four analog commands- maybe 3 degrees of freedom in all? And while the visual environment can be complex, there are situations- such as driving on a highway- in which the environment has already been reduced to a schematic version of itself: a wide, mostly straight road, with lanes and sides marked, and all other objects on it conveniently vehicle-shaped. 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.
    - kleer001 7y ago
      There are no battle field surgical robots running unassisted operations on humans.
    - perl4ever 7y ago
      I'm sure I've read in the past about robotic surgery allowing compensation for a patient that is definitely not completely still, whereas traditional methods might, say, require stopping the heart. Are they still "controlled" by a human? Maybe, but there are different levels of control.
Sanzig 7y ago
Exactly, it's an early stage proof-of-concept prototype (TRL 3, if you prefer that scale). While I have my reservations about how difficult it will be to actually get this into something you can safely and reliably use on a consumer without lawsuits (that's a long road), it makes no sense to judge the idea solely based on an early bench prototype. 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.

kleer001 7y ago

Why? A couple reasons. 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.

Udik 7y ago

That is never gonna work. Seriously. Cutting hair is a highly manual job, requiring precision use of a number of different tools- some pretty dangerous- on an irregular, soft and constantly moving surface. Covered in, well, long, irregularly shaped hair. The idea of a hair cutting robot is fun, but it's decades or more ahead of what we can do.

perl4ever 7y ago
It seems a lot easier than driving a car, and your description of the task sounds like surgical robots, which are a thing already, with superhuman capabilities.
- Udik 7y ago
  I don't think it is easier than driving a car- somehow it feels even harder. Driving is a very well defined task: you are sitting in a precise position from which you get a well defined and purely visual input, and based on that you operate on three or four analog commands- maybe 3 degrees of freedom in all? And while the visual environment can be complex, there are situations- such as driving on a highway- in which the environment has already been reduced to a schematic version of itself: a wide, mostly straight road, with lanes and sides marked, and all other objects on it conveniently vehicle-shaped. 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.
  - kleer001 7y ago
    There are no battle field surgical robots running unassisted operations on humans.
  - perl4ever 7y ago
    I'm sure I've read in the past about robotic surgery allowing compensation for a patient that is definitely not completely still, whereas traditional methods might, say, require stopping the heart. Are they still "controlled" by a human? Maybe, but there are different levels of control.

Sanzig 7y ago

Exactly, it's an early stage proof-of-concept prototype (TRL 3, if you prefer that scale). While I have my reservations about how difficult it will be to actually get this into something you can safely and reliably use on a consumer without lawsuits (that's a long road), it makes no sense to judge the idea solely based on an early bench prototype. 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.