Let’s start from the beginning, with the announcement on Tesla AI Day, August 19, 2021, where Elon Musk shocked the world that Tesla is working on a humanoid robot, a Tesla Bot later nicknamed Optimus. Specifications shared put Optimus at 5’8″ (173 cm) tall, weighing 125 pounds (57 kg), carrying capacity of 45 pounds (20 kg), & deadlift capacity of 150 pounds (68 kg). And very importantly, speed was restricted at 5 miles per hour, so that we could always run away from it, just in case we found ourselves in the midst of a robot uprising. Jokes aside, Elon did share that Optimus, just like any Tesla car, will have the advantages of the latest Tesla computer hardware 3 — aka FSD (full self-driving) computer, cameras, & battery. Furthermore, it was said that it would be possible to get a working prototype within a year.
By Ganesh Sadare
Where Tesla Optimus Originated
According to Elon, a Tesla car navigating itself autonomously is already a “semi-sentient robot on wheels,” and it makes sense for Tesla to go for it and put all of its expertise & AI in humanoid form. Explaining it further, Elon thinks the Tesla bot directly answers a labor shortage issue. (Note: CleanTechnica‘s Chanan Bos wrote about this idea — because it just seemed like the logical next step — and again leading into the Tesla AI Day event in August 2021.)
“I think Tesla Optimus has the potential to be more significant than the vehicle business over time. If you think about the economy, the foundation of the economy is labor. Capital equipment is distilled labor. So, what happens if you don’t actually have a labor shortage? I’m not sure what an economy even means at that point. That’s what Optimus is about. So, very important.”
Apart from attracting & hiring the best of AI talent for itself — as FSD is still not fully solved problem and is quintessential piece establishing efficacy of Tesla Vision for both Tesla autonomous driving and Optimus, which Elon does think they are close to solving — the potential for population collapse may also be another reason the Optimus project got priority status. More on this later.
A fully functioning AI robot opens up limitless labor supply — constrained only by cost & resources to make it — creating an exponential boost for any economy. There’s no wonder why any AI company would want to create one. But it is a highly complex endeavor and hardly any AI has solved the autonomous navigation problem: Boston Dynamics robots are some of the most highly sophisticated robots out there & they too need an operator with controller and have limited autonomous capabilities.
Tesla’s Approach: Tesla Vision & Vertical Integration
For autonomous driving, Tesla is working on a general solution that can be scaled and tackle all scenarios. Most in the auto industry are still using and working on systems that depend on mapped data through lidar, which only works for that specific data, and radar sensors, which Tesla has removed completely from its cars as it has gained more accuracy with computer vision to detect, identify, & track moving or static objects that outmatch sensors.
Tesla’s autonomous driving system depends on a computer vision based approach called Tesla Vision. Tesla wants AI to see the world as we see it, through cameras in visual spectrum similar to ours, and take decisions accordingly in real time. The AI Day stream went far into technical details on these matters, so definitely check that out.
In short, all Tesla cars see the world in vector space (recreated 3D digital space from camera data), which gets updated in real time and predicts the appropriate response from the neural network model that has gone through data collection, auto labeling, simulation, neural network model training, optimization, and then back to Tesla hardware 3. Meanwhile, all new important cases generated from any Tesla car from the fleet get added in the loop to be part of the model, creating a snowball effect of scaling this solution to be more & more accurate after each iteration.
Every step above is vertically integrated from start to finish. There is no outsourcing, unlike at other companies. The Hardware 3 computer that is inside each new Tesla car itself is designed in-house for the specific purpose of solving autonomous driving. There is no other company that has done this so far. And already going above and beyond that, Tesla has now designed a supercomputer chip & is in the middle of stacking multiple of these chips to create a supercomputer named Dojo specially for AI training & optimization. The amount of effort, cost, & resources Tesla has deployed for its AI solution is unparalleled.
Simply put, in-house designed hardware & software specifically for the AI and ability to crunch, scale, & train vast amounts of data (with the introduction of the Dojo supercomputer) puts Tesla at the forefront of autonomous navigation solutions.
Optimus will navigate itself on the same Tesla Vision based approach. And it might even get efficient quickly, considering 5 miles per hour speed means limited moving objects to keep track of & the extra response time at its disposal compared to the car, both of which will be far less computationally heavy.
The Current Future
An Optimus that can take voice commands and be able to do basic tasks in a Tesla factory is what Elon wants to get started with; a robot that can navigate on its own, assemble or move parts, and basically do any physical repetitive task. It’s certain that the biggest benefit of Optimus as a product will be reaped by Elon’s companies before anyone else can get their hands on it. Especially, Optimus can be a great boon for SpaceX and may even become a key player on Elon’s Mars City ambition. An Optimus doesn’t eat or sleep, doesn’t need any extra radiation protection nor any life support system, and now all of a sudden, all the risk, capital, & technology involved for Mars launches are stripped to the bare minimum. What if all or part of the things that were needed to be done by humans could be done with Optimus? Starship cargo versions with Optimus squad can be the very first Mars launch & landing of Starship. And maybe even on the Moon. The cost & benefit of each Optimus launch will be astronomically low compared with any human spaceflight. Starship solves Earth-to-space cost per cargo problem & Optimus might be the solution for cost of setting up self-sustaining cities with that cargo. Even at present, a Starship that doesn’t have an unloading mechanism may possibly look at Optimus as a whole or partial (system designed with Optimus at its core) solution.
Optimus will do the same basic task there on the Moon or Mars as is meant for the factory floor right now; be autonomous, load/unload cargo, set up and assemble/disassemble different parts. Imagine Starship landing on the Moon or Mars with a squad of Optimus. They unload cargo, set up solar panels, recharge & explore autonomously, maybe even do science experiments. For months, they will clean solar panels, maintain, & use scientific equipments, roam far out & stream HD content back to us. We may even explore caves and lava tubes for the very first time through the eyes of Optimus. Months later, another cargo Starship will arrive with new cargo, the squad will unload base building prefabricated parts, get updated with new construction programs, and start working on building structures with new squad. Now, it’s not possible to just expect a fully environment controlled base ready to welcome us when we will eventually reach there, but it’s fair to say that whatever can be done with Optimus will be done by Optimus, leaving us to do less risky and complex stuff. Does all of this diminish the role of human spaceflight or excitement? No. Buzz that will generate from the first video of humanoid Optimus figures waving at us from the Moon or Mars will only excite us all to look forward to getting there and becoming a multiplanetary species as fast as possible. Optimus will turbocharge the mission of becoming a multiplanetary species.
Even basic load/unload, assemble/disassemble capable Optimuses (Optimi?) with their potential impact on earth and especially in space fully justify the investment & priority in developing them.
In the next article, I’ll talk about different versions of Optimus, use cases, cost, who will buy it, and more. Check it out now!
Note: I don’t own any Tesla shares or have any long position on Tesla.
Don't want to miss a cleantech story? Sign up for daily news updates from CleanTechnica on email. Or follow us on Google News!
Have a tip for CleanTechnica, want to advertise, or want to suggest a guest for our CleanTech Talk podcast? Contact us here.