Why the Robot Industry Is So Focused On Single-Task Robots
They’re Optimally Suited for Their Tasks — and Multi-Purpose Robots are a Long Way Off
Conventional industry wisdom these days is “robots must be purpose-built for the task they’re supposed to perform”. Single-task robots abound, from strawberry pickers to box movers.
There is good logic in this. Many tasks performable by robots can be highly optimized for tight parameters:
- specific grasping and manipulating abilities (a consistently deft and lightweight touch for strawberries, compared with non-discerning grabs for moving boxes)
- the physical dimensions of the task (strawberries being pear-shaped and boxes being mostly uniformly square)
- movement over specialized surfaces (uneven ground or smooth factory floors)
- environment (indoors, outdoors, underwater, in space, etc.)
At the same time, each ability is often associated with a highly specific and quantifiable business benefit:
- Strawberry picking robots can avoid the need to find low-paid, low-skill workers who are in sharp demand right now.
- Box moving robots at Amazon can improve speed and accuracy, avoid workman’s comp payments, and allow for continuous work over multiple shifts.
- Robots shoveling sidewalk snow can work at below zero temps at 3am, while avoiding the risks and labor costs associated with dark and dangerous jobs.
There is often a strong business ROI for these single-purpose robots. Even in the home, there can be strong ROI for some single-purpose robots like vacuum cleaners and lawn trimmers. When the cost of the robot is spread over dozens or hundreds of events, how much would the homeowner pay to skip the drudgery? $5? $10? For many, that’s a fine ROI.
Meanwhile, in the world of multi-purpose robots the world looks very different. Many aspects of multi-purpose robots are either supremely expensive or currently impossible:
- grasping solutions that manage the range of delicate to non-discerning
- handling loads of >40 pounds
- navigating both indoor and outdoor terrains
- recognizing an infinitely variable array of objects
- interacting with humans in an infinitely variable number of settings
- handling all possible voice interactions (I’m sure you’ve heard even today’s most advanced voice systems say: “I’m sorry; I can’t answer that.”)
Even more pernicious for multi-purpose robots (and their makers) is the challenge of meeting consumer expectation.
Many consumers, when presented with a robot that performs a single task (vacuum cleaning, lawn mowing) have no expectations the robot will speak fluently or recognize their cousin.
But ask about their expectations for a multi-purpose robot (e.g. C-3PO), and you’ll likely find expectations that the robot can understand all languages, process any question, identify and manipulate any object, cover any terrain, etc.
No company in the world can come anywhere close to meeting these expectations right now nor any time soon .
But wait, you ask, isn’t Misty Robotics building a multi-purpose robot? Why are you writing this blog post, if your business is utter folly?
Because we get asked these questions a lot: Why aren’t you building a robot for one “killer” use case? Why aren’t you building a single-purpose robot?
And we do have an answer. It is that the path to the multi-purpose robot has to start somewhere. True, we could have outrageously overestimated our market timing. But we don’t think so.
So…. what’s the path?
The path starts with a set of work we did at the outset of our company. We did some brainstorming of the “task space” that humans in the office and home undertake.
We asked ourselves, “What are the roles one human pays money to another to perform?” (Thus proving there is the potential for ROI.) The roles were in areas such as security guard, receptionist, inventory clerk, butler, maid, nanny, nurse.
Then, “What are the tasks those roles perform?” These answers included “patrol, greet and guide people, get the dog off the couch, nag the kids to do their chores, make sure grandma turned off the oven and took her meds.”
Finally, we asked “Which of those tasks requires only indoor mobility, sight, hearing and sound?” Because, if a robot could be built that was capable of independent indoor mobility, sight, hearing, and sound then, we hypothesized, that robot could be multi-purpose within those constraints.
In our estimation there is enough task space to begin. As we’ve outlined in Misty’s 10-Year Plan, we’re building a multi-purpose robot for inventors to invent within that task space. There are literally thousands of tasks we humans perform that require only mobility, sight, hearing, and sound.
The leap of faith we’ve taken in our business is this — that the task space is wide enough, and will exist for long enough, to give us time to build abilities like multi-purpose grasping for an affordable cost. And to build a variety of different form factors (indoor/outdoor, hefty/deft).
We’re also taking the leap of faith that there’s a 12-year-old dreamer inside most programmers, developers, makers, and inventors who has always fantasized of easily programming a robot but has never had a robot that would enable them. Until Misty.
We’re comfortable with those two leaps of faith. Yes, true, we might be early. Then again, tens of thousands of inventors might invent millions of tasks for Misty robots within the task-constrained space, making them highly valuable to consumers in offices and homes everywhere.
And then? We’ll add hands.
And then?
Katie bar the doors.
