Screenshot from Tesla's AI Day livestream
Is a human-shaped robot like Optimus, the Tesla Bot, the right path to travel if we want to achieve useful robots and automated help in all aspects of our working and personal lives? That’s basically the promise of Optimus, which Tesla CEO Elon Musk says will usher in “a fundamental transformation for civilization as we know it.”
And maybe it will.
But very possibly, maybe not.
One thing is undeniable: there’s an abiding appeal to human shaped and human sized robots, as Irena Cronin, CEO of Infinite Retina, recently told me. Part of the value of copying the human form: the world is built for humans. Cars, homes, machines, factories, warehouses ... all of our built environment is designed by and for human beings as inhabitants, operators, drivers, or workers. And Tesla’s not the only one attempting to create humaniform robots. China’s Xiaomi recently hyped a similar prototype bipedal robot: CyberOne, and at least 10 companies have tackled the problem recently, although some have already given up or run out of funding.
The obvious benefit: a humaniform robot is instantly perfectly adapted to serve any desired need in any of those spaces.
If — and this is the critical if — the robot works. And if Tesla, or Xiaomi, or Boston Dynamics, or some other company manages to pull it off the entire business: hardware, AI, production, and ecosystem of service, support, and perhaps apps and attachments.
One skeptic that we’ll get there anytime soon is MIT professor Daniela Rus, who shares the core problem in a succinct statement: the more you generalize, the less you optimize.
In this view, a purpose-built robot or machine like a Roomba, or a Robby delivery bot, or a plain old dishwasher can be optimized to do a good job at the one thing it’s intended to do. But as soon as you generalize — carpet and hardwood for the Roomba, stairs and uneven under-construction sidewalks for the delivery bot, self-loading for the dishwasher — you expand the problem set and reduce the specific effectiveness of the machine.
In her view, a universally useful humaniform robot is less likely in the near-term than specifically-useful robots.
“If we rethink what a robot is, then we turn your chair into a robot,” Rus told me in a recent TechFirst podcast. “We turn your door into a robot. We turn your shirt into a robot.”
In this scenario, the door knows who you are, opens when needed, closes and locks when needed, and perhaps alerts authorities if unauthorized people attempt entry. Your shirt changes color, provides more warmth when needed, reconfigures to a more formal setting for a dinner in a high-end restaurant.
This kind of perspective allows for significant robotic infiltration into our lives, homes, and jobs, but rather than mimicking the human form to be good at everything a human might do, provides specific robots for specific tasks.
Or, offers reconfigurable robots that adapt themselves to meet different needs.
“I have been on a quest to have universal machines,” Rus, who is the director of the Computer Science and Artificial Intelligence Laboratory at MIT and a MacArthur Fellow, says. “My idea is to create universal robot cells that could combine to form different types of machines, each with the same capability. So the shape and the function of the machine would be specialized. But since each machine will be built out of the same building blocks, we could have a kind of a generality towards how we think about robots. And so this area of robotics is called modular self-reconfiguring robot systems.”
And in fact MIT is building such robots.
One example is the programmable matter mini robots that MIT teams are working on for space. Another is the smart sand that Rus envisions: tell a bag of smart sand what form or tool or machine you need, and it configures it out of tiny composable segments. Finished with the tool? Toss it back in the back to be decomposed and ready for the next task. Another are the “roboats” MIT has deployed in Amsterdam as autonomously assembling and reconfiguring components that form bridges, stages, or floating platforms as needed.
That’s probably where the betting money is.
But the dream of golems, or automata, or robots that look somewhat like us and can work for us is a hard one to kill, Rus acknowledges.
“I think this is going to happen,” Cronin says. “People do not like to do housework. They don't have time to do housework.”
Of course, Optimus wouldn’t be limited to housework, although I’d certain appreciate having one around for that, and maybe some yard work with the hand-dandy Week Wacking 2000 attachment, and ... why not for anything else that needs doing around the house?
They’d also be useful for building ... Teslas.
“Elon's alluded to factory work,” Cronin says. “Obviously, I'm sure that he would love to have these in the factory so at some point.”
And deliveries work better with last-mile truck-to-house, so having an Optimus along for the ride in a self-driving delivery truck would help get your packages, if not your new sofa or clothes dryer, into your home.
All of that is compelling, if it can be made to work. MIT’s Rus, after a lifetime of designing and building robots, gets that desire and understands that path. But it’s certainly not possible today — as we saw in Tesla’s introduction of Optimus, which is very early stage — and there are valid questions about whether it will be realistically available any time soon.
“While the robotics community has made huge progress in making humanoid-like robots, we have a very long way to go in order to develop machines that have the agility, the dexterity, the reasoning, and the imagination, the range of intellectual and physical activities that humans are capable of doing,” Rus says.
That’s basically undeniable at this moment.
Cronin and Infinite Retina strategy chief Robert Scoble, whose jobs include essentially predicting the future for clients so they can time entry into the market, are bullish on the possibility in the relatively near future, however. Part of the reason: Tesla’s proficiency at AI, which Scoble says is already unparalleled in its ability to drive him around in his Model 3. Another part: Musk is incentivized to built Optimus not only for use on the Tesla factory floor, but also as an expansion of services in potentially hundreds of other places.
“Why am I excited about the Tesla robot today?” Scoble asks. “It’s the business model that Tesla brings to the robot.”
Business models, of course, have timelines. Return on investment calculations. And business plans. And all of that is dependent on when Optimus could actually hit the market. Scoble and Cronin think they’ve pinned down when Elon Musk’s vision could be reality.
The target year, they say, is 2028.
“I think within the next five years, he'll at least have something definitively ready for his factory,” Cronin says. “So they would've gone through all the motions and testing for all the many types of things that they would need to do in the factory that that would demonstrate save time and would be greatly more efficient than what they're doing now.”
After the factory as a proving ground, of course, the world is Tesla Bot’s oyster. Household tasks, deliveries, retail uses, even farming applications ... the possibilities for Optimus are endless.
Most of those possibilities are right down here on planet earth. But imagine a Starship from SpaceX, one of Musk’s other companies, loaded with a crew of Optimus Tesla bots, setting metaphorical sail for Mars to build a settlement in anticipation of human arrivals? That’s science fiction at the moment, but it might just be science fact in not too many more years.
If Musk can pull this off — and it won’t be easy — the results will indeed be a fundamental transformation for civilization as we know it. And civilization as it might evolve.
Get a transcript for my conversation with Daniela Rus here, and Robert Scoble and Irena Cronin here.
