At the time Elon announced the Tesla Bot, he indicated the objective was to have the robot perform dangerous, repetitive, and/or boring tasks. While this sounds like an admirable scope, it’s difficult to ignore the vast potential use cases we can anticipate over time.
When you replicate the physicality of a human, you open the door to having the bot perform many, and over time, likely almost every physical task that we do. Where the most difficult challenge lays, is intelligence, so the jobs of the future, are likely not going to be in manual handling, but rather creative industries.
In technology companies, there’s a concept called dogfooding.. this is the practice of testing products internally, with your own company first. This establishes that the base product is useful for the use cases you have, before releasing it to a wider audience. This week on the earnings call, we learned that Tesla plans to do exactly this with Tesla Bot.
Knowing this raises the question of which tasks Tesla deploy the Bot on first? When functionality is limited, the most basic tasks seem a likely candidate.
Given Tesla is a company that will make more than a million vehicles this year, thousands of stationary storage products in Powerwall and Megapack, along with Solar Panels and Solar Roof, there’s a lot of tasks associated with each of these.
In the past Elon has spoken about automation in Gigafactories and admitted they went too far down the path of attempting to automate everything. There are tasks that are simply faster and easier for humans to do. This reflection was based on attempts to use mechanical arms to perform these tasks, which were not designed to do things like route wiring and secure it, through the body of a vehicle.
Armed with a new form factor that replicates a human, there’s potential this challenge is re-assessed as the bot grows in ability, but certainly won’t be a likely candidate for the first use case.
First use cases for Tesla Bot
Tesla has two, and soon four Gigafactories that produce cars – Freemont, Shanghai, Texas, and Berlin. These Gigafactories each have regular deliveries, likely multiple times per day, of raw materials like lithium, aluminium, along with thousands of parts and accessories from suppliers.
These will arrive at the factory by the truck-load, which backs up to the loading bay, opens the door and humans use either a pallet jack or forklift to bring the goods inside.
After the products and materials are inside the Gigafactory, they need to be unwrapped and distributed to either temporary storage or in the best case, directly into the bins for installation on the production line, to be used by human operators (at least for now).
These unwrapping, product identification, inventory management, and internal distribution, could all be tasks that would be ripe for the Tesla Bot to take over. These are repetitive, need to occur in a timely fashion, 24 hours a day and likely have some risk associated with manually handling.
There are already automated robots that move about the factory, particularly in the case of battery packs, between the battery division and mating with the vehicle, but these Automated Guided Vehicles are dedicated to the task and will continue to be a more efficient form factor for this task, so the Bot won’t be right for every situation.
Over time, the Tesla Bot could be great at tasks in general assembly, including installing seats, doors, wiring harnesses etc, but that is definitely some time away.
Some may consider the computer vision capabilities of the Tesla Bot as having a place in quality assurance. Imagine the Bot walking around the car and doing a live comparison of the final product versus a perfect model, highlighting any defects. In reality, there are already moves inside Tesla to move to computer vision and an array of fixed, high-resolution cameras to review paint defects, panel gaps etc.
What could be interesting is an internal inspection, the bot has the potential to open the door, get inside the vehicle and check everything is in order. Don’t forget, the car also needs to be moved off the production line into the car park, something currently done by humans. Hopefully, ‘factory mode’ of FSD can reach the point where nobody human or robot, has to enter the vehicle to move it, the car could simply drive itself off the line and go park itself in an available space.
Commercial use cases
Outside Tesla’s own requirements, there’s likely to be a long list of commercial opportunities that the bot would be great for, and businesses should get excited by.
Transport and Logistics aka, the last mile problem
Imagine you order from a food delivery service like Uber Eats or DoorDash, even once we have autonomous cars, the products still need to be loaded into a vehicle and unloaded at the other end. Some companies appear to be planning to pull their autonomous car into the closest park, then send a push notification to the customer to go collect their goods from the car’s location.
That’s possible, but you’ll likely have a short window of time to do this, before the car gives up to go deliver the next job, given the time-sensitive nature of food delivery.
Imagine an alternative future, where a Tesla vehicle with Full-Self Driving is sold to a company with a Tesla Bot to solve the last mile problem. This would see the Bot take the goods from the store, place them in the vehicle in either the frunk or trunk, then get in the car. The car plots the most efficient route to deliver the goods to multiple customers and the Bot gets out of the car, and delivers each order to each customer.
This means they’ll need to understand a lot about the entry and exit of vehicles, a lot about the approach to font doors, including steps, gates, doorbells, knocking on doors etc. The opportunity is massive, potentially removing the human cost of a driver, but the challenge is equally massive with the possible variations involved. At the end of the delivery, the Bot would return back to the store to collect the next order and repeat until either it, or the car needed recharging.
Charging electric cars
There are many fleet vehicle operators right now that are considering how the shift to electric vehicles plays out. While charging your own electric car is a job you need to address, in the commercial setting, charging dozens of fleet vehicles, or in the case of a rental service like Hertz, hundreds is a real logistical challenge.
Whether it’s an overnight charge from a destination charger, or a rapid charge on a road trip at a Supercharger, the Tesla Bot could play a key role in automating the charging of electric vehicles.
In the context of robotaxis, these autonomous vehicles need to run for many hours a day to drive the cost per ride lower than ride-sharing services like Uber and Lyft. To achieve this, the cars will need to be recharged and fast to optimise their returns. Having an automated service, even if paid, would enable the robotaxi future that many are hoping for. The dumb thing to do is to create a new job type for humans, the smart thing is to create a robot that can perform this task all day, every day, and deal with the variances between vehicle heights, parking position in the bay etc.
While having a robot connect an electric charger to your car’s charging port may seem scary at first, what if they mess up and scratch the paint, or break off the charging flap, the visual system used by Tesla Bot, combined with finely controlled motor movements from electronic actuators, should enable the Bot to insert the charger even better than humans, if Tesla can master its hand-eye coordination.
Many countries around the world have an aging population and caring for their needs, is a massive societal challenge. Not only is there the need for medication distribution, but other day-to-day needs like delivering food to each room of a nursing home, mobility challenges like assisting the elderly in and out of bed, to the bathroom, or pushing a wheelchair on a walk outside so they can get some sunshine each day. These are the kinds of challenges that could be answered with a humanoid robot.
Every location has a different layout, different obstacles in the way and each patient in each room has a different set of healthcare needs. What if the residents or patients are not in their designated rooms, but in common areas, the Bot would then need to interact with that person first by identifying who they are, so facial recognition may be important in this specific use case.
Of course, healthcare is much broader than addressing just the physical needs, people are also often after companionship, just someone to talk to. Could the Tesla Bot perform two-way conversations, we haven’t seen that from Tesla yet, but would provide massive additional capabilities.
Don’t forget about that visual display in the head of the robot, this could display a smiling face, to provide comfort to the people it interacts with, perhaps most important if the Bot was deployed in a children’s ward at a hospital.
The thing about healthcare is that it’s relentless, the tasks never stop and as one patient or resident leaves, another arrives and while we hope Covid is soon behind us, right now, health care providers are really struggling to find staff. One of the harsh, but real benefits of a robot in the health setting, is that it’s not susceptible to contracting diseases the way humans are.
If you think about how we secure a building out of business hours right now, we use a series of fixed security cameras, either actively monitored in a control room, or passively monitored through motion alerts. Usually, this is backed up with human patrols on a regular schedule. Clearly, the patrol portion of this job could be done far better by a Tesla Bot.
Assuming the battery life would more than cover the shift of a regular worker, a Tesla Bot would offer a number of advantages over a human. The Bot doesn’t get tired, it doesn’t get lazy, it doesn’t need bathroom breaks, it doesn’t play games on its phone and it can see in multiple directions at once.
Of course, the disadvantage is that the Bot would not be able to give chase or engage in the physical defense of an asset if there was a physical break-in. What the Tesla Bot could do, is notify the authorities and provide the video (either recordings or a live feed), leveraging the bot’s always-on mobile connection, similar to what we have in the cars and live Sentry Mode.
If Tesla really is able to make a humanoid robot that arrives in the form factor they suggested, there are hundreds, of potential use cases possible, if the robot has the right ability to learn new tasks quickly, something we’ll expand on in a future post.
There are obvious applications for humanoid robots in police or military contexts, mostly to replace in the most dangerous scenarios, although I suspect Tesla will shy away from these for PR reasons.
It’s definitely possible that feature filmmakers leverage a Tesla Bot to perform stunts, or even behind the scenes to hold cameras, there really are an enormous amount of possibilities, if the technology works.
Personal use cases
The commercial applications for Tesla Bot will certainly get built first. With the immense benefits and cost savings a humanoid robot could deliver, the price is likely to be significant. While it’s too early for Tesla to name a price, we can expect tens of thousands of dollars, making it out of reach for most ordinary people.
Over time, the cost may come down and then unlock more uses personal use cases for Tesla Bot, so I want to imagine some use cases for a Personal, at-home version of the Tesla Bot.
Many of our homes now include robots that are dedicated to the tasks of vacuuming our floors, mowing our lawns etc, but these all have a very narrow focus, and buying dedicated robots per task, will never stack up against a universal robot that can multi-task.
Sure, a robot lawnmower is great, but what about the task of wiper snipping your edges, then using a blower vac to clean up, or what about weeding the garden, trimming a hedge?
In the past couple of years, our robot vacuums have improved, instead of requiring emptying after every load, they now empty into a larger canister that stores the collected dirt from multiple cleaning cycles, but when that larger bin gets full, you’re ultimately left with taking out the garbage to larger outside bins.
For the Tesla Bot to solve for this, they would need to be able to navigate a home, not an easy task, particularly when you consider they’d need to find an external door, usually locked, unlock it, often a circular handle that needs to be rotated, so this will be a key test for the Tesla Bot, if it can’t do this, it’ll struggle for many applications in the home.
Imagine the bot can learn how to successfully take out the trash, we have to get comfortable with the concept that a robot, could take a secured home and unlock a door to go outside.
When going outdoors, the Bot also needs to be smart about the current weather. While the Bot may be water-resistant, I wouldn’t expect it to go out in the pouring rain, if it did, then it is likely to bring water/mud back into the house. It would only take one ruined carpet incident to see your fancy expensive robot, be turned off and sent back.
Given Tesla’s learnings in the car regarding rain and when they need to fire the windscreen wiper, their ability to detect rain and not venture into hazardous environments, should be possible, but may start with a basic weather lookup for the current location to determine if rain is likely.
For this function to be successfully completed, the Bot would need to return inside and re-lock the door. Imagine the robot does this task for a single bin in your home, the reality is, most of us have multiple bins throughout our home, often hidden away in cupboards, with plastic bags that need to be removed intricately and tied off to avoid their contents spilling.
Each bin is unlikely to need emptying on the same schedule, so this raises the issue of how you’d go about scheduling your robot to perform these house maintenance tasks. At some level managing, the calendar for your robot could be more work than actually performing tasks yourself, so this is a critical interface that needs to be slick.
We’ll focus on how the robot may learn in another post, but if the Bot learns a new skill, that may be the time for it to ask you, how often you need that done. If you say weekly at 7PM on Monday’s, it’ll add it to the calendar and you forget about ever doing that task again.
In the context of mowing your lawn, there’s often a varied schedule, based on the time of year, so that’d also be nice to accommodate automatically, or there could be a network of IoT sensors that the Bot can leverage to make determinations on the time to execute a task.
As humans, we perform many of the sub-tasks necessary for home maintenance without thinking, opening the garage door to retrieve a power tool, grabbing a battery from the charger, attaching it to perform the job, then reversing the process when complete. These all require fine motor movements to execute and align objects to others in the 3D space. As difficult as this is to imagine a robot performing today, all the necessary components are in front of Tesla, so there should be nothing technically stopping them from executing, just time, resources, and budget.
Cleaning your house
There are so many other tasks around the home that could be done by the Tesla Bot, but require a unique skill set for each.
Packing and unpacking the dishwasher.
Breaking these down to their core tasks, you need to retrieve the dirty cups, plates, bowls, cutlery etc. These are likely left in various places around the home, so maybe the Bot could do a walkthrough of each room and retrieve any items it finds.
Packing and unpacking a dishwasher requires millimeter accuracy and a fine touch, to ensure you don’t break anything. Then there’s the challenge of setting the correct washing mode, by pressing a sequence of buttons on the dishwasher, each that varies by manufacturer and model. This is why I imagine the training needs to be done in a way that allows you to show the robot the steps to follow and have it repeat it, while being smart enough to adapt to the difference between having different items each load.
Washing clothes etc
In a similar challenge to the dishwasher story above, the challenge of determining the location of dirty clothes throughout the house is significant. What I hope we can avoid is having to use dedicated laundry baskets with massive QR code stickers on them, to enable the robot to know, I retrieve the contents from here.
What’s interesting about this challenge is not the fact people will have a mix of front and top loader machines, but the fact a load of washing needs to be transported to and from the washing machine in a washing basket. Often we’ll hold something like that against our bodies to balance it, so I’d love to see how the robot handles this, particularly when the washing is wet after being through the machine, the additional weight can’t be more than the robot can carry or the automation is over.
Hanging the clothes on the line may seem easy for you or me, but for a robot, there are another million permutations, it really is hard to imagine how Tesla would tackle this, particularly when you consider the difficulty involved in taking a peg and clipping a sock to the line.
It would not surprise me to see the Tesla Bot take 5 and maybe 10 years to have this level of functionality, but definitely happy to be wrong.
Many of us leverage smart home lighting as part of a security solution, suggesting to would-be criminals that there is someone home. We also use home security cameras to capture any break-ins and remotely monitor a property, pets etc while we’re away on holidays.
It’s easy to imagine the Tesla Bot could be the best security device you have. It could patrol the home while you’re not around, and provide a live video feed, with a Sentry Mode-style monitoring. Given the low-light environment at night could prove challenging to navigate, it would make sense that Tesla uses cameras with a night-vision capability, mostly to ensure the Bot avoids stepping on random toys left on the floor by your kids, and avoid any pets that may wander the home at night.
While external cameras are much more common, some homes do leverage internal cameras and even while the Bot is charging, it should be able to provide this level of security.
If you have done well financially in life, then you may entertain the idea of getting a Tesla Bot not because you need one, but because you want one. In this context, the sky is the limit, as long as the training interface is made quick and easy.
I can imagine a house party of the future could invovle a robot butler, where the robot is able to not only retrieve a beer from the fridge, but actually follow recipes for both food and cocktails. If the Tesla Bot becomes a robot chef in your home, now how much would you pay?
Of course there’s lots of other possibilities, like babysitting your kids between 3PM when they finish school and 5PM when you finish work, help them with their homework
The bot could be great for activities like backyard cricket, totem tennis, and more, but these entertainment activities are likely many years away, with the commercial applications likely to snap up the supply of Tesla Bots in the short to medium-term.
Let us know in the comments what you’d like the Tesla Bot to do at your home or business.