Sorry about that. It is a PITA when long.
FUD increases the volatility which is quite profitable when selling options.
Be happy for me!
ralph
2 Likes
Whatever kind he wants. I’m not saying that Musk shouldn’t have an Optimus program. To the contrary - I think he needs to have something like that going on so that he can maintain the framing of Tesla as more than a car company. Musk has continually positioned FSD as a problem that is almost nearly solved - few machine learning boffins are going to want to work for Tesla’s AI division if they think the division is going to get shut down and they’ll all lose their jobs in a year or so once FSD moves from development into basic maintenance (no matter how awesome Musk is). There has to be a long term machine learning project inside of Tesla in order to make engineer recruitment possible - making cool humanoid robots fits that to a T.
I just don’t think this is something that’s going to be a $10T business (or whatever number’s being thrown out there), for all the reasons we’ve discussed. We’re too far away from when AI can do the sorts of things that AI needs to do for a humanoid form factor to be useful - and we’re way too far away from the point where the “learning” part of the system can be in the robot, rather than in a multi-billion dollar supercomputer in Texas. Remember, if Tesla were to sell me a robot, that robot wouldn’t be able to learn anything - it would (at best) be able to upload data to Tesla’s massive computers so that those computers can try to figure out how to do what I want the robot to do. That’s something that works out fine for Tesla’s FSD program - and maybe it works for Tesla’s internal operations - but it’s going to be really hard to set up a commercial arrangement where that makes sense for outside customers all trying to set up robots to work in their individual (and idiosyncratically unique) work situation, and need Tesla’s AI squad to figure out how to make that happen.
1 Like
It might be worse than that.
I’ve watched a dozen or more videos on Optimus.
Some tubers are saying that the customer owner will NOT be allowed to randomly train their bot for a new task.
The bot will come with the desired skill set, that was paid for.
You, the buyer, will select the skill set/s you want, those will be uploaded to your bot.
If you get the house cleaner with lawn mowing skill set, it will not also do beekeeping on the side. If you want the make balloon animals skill set, you’ll need to pay for it.
In addition, there will be “kill switch” circuits in place to prevent the owner from “teaching” a new skill set.
If you want a new functionality, you gotta pay for that.
Just like lots of apps, and software companies already require.
But, this is a good thing.
The tubers say this helps prevent a maleficent from obtaining a bot, training it to use firearms and then kill a bunch of people.
ralph
1 Like
I doubt even Tesla’s made any decisions about anything like that yet - even the time frame given in the earnings call has Optimus deliveries being a year and a half out in the future. But it does make some amount of sense that this type of product wouldn’t necessarily be shipped with an “everything bagel” software stack, rather than something that might be more tailored to what the customer intends to use it for.
What might be more cumbersome, though, is making sure that the robot has “learned” everything specific to what the customer needs. Right now, Optimus isn’t being trained - and may not be designed to be trained - solely on vision. Per the Optimus lead designer:
“We’ve improved our locomotion stack, frequently walking off-gantry without falls and with a faster, increasingly more human-like walking gait. We’ve built a very low-latency & high-fidelity teleoperation system, used to collect AI training data of the bot imitating humans performing certain tasks… When the time is right, our manufacturing lines will be added to that list,” [Optimus Lead engineer Milan Kovac] noted.
It makes sense that Optimus’ training isn’t just based on vision, but includes data generated through teleoperation. It’s learning by reviewing data generated by humans “puppeting” the robot, not just from watching humans behave, which allows it to collect tons of tactile and kinesthetic data in addition to visual data. That’s something that they can (and apparently intend to) use when training to do jobs internal to Tesla. But that’s a process that won’t let these bots just get shipped out and watch a different factory floor and then “learn” how to use that equipment - Tesla would have to figure out how to cut out the teleoperation part of the training for that to happen.
1 Like
That would make sense. It would also make them vastly more expensive while also making them less useful.
Out the window would be “oh, we’re not doing that today, go do that task instead” without substantial “re”-training, and it would also mean that every task would require individual programming - and approval - before being allowed to proceed. Once trained, of course, you could “park” the training and wait for it to be useful again, but that’s gonna cost.
Indeed, you couldn’t even move it 20 feet down the line from “feather plucking” to “deboning” without retraining, and there would be safety issues involved in every change.
Well that’s would probably go over as well as BMW’s “pay for seat heaters by subscription’ idea. While I have substantial doubts about the use of humanoid form robots on factory lines, I can see them being used as personal attendants in the homes of wealthy elderly, allowing more to age in place rather than being shuffled off to the warehouses of death we call “nursing homes.” But that requires that the robots do a wide variety of chores, from cooking and cleaning to help moving people from bed to chair, general nursing cleanup and other tasks as are currently done by low paid CNAs and personal assistants.
1 Like
Maybe not?
Retraining would be clicking the app on your phone, choosing the desired skill set, downloading it … OTA, of course.
Maybe 2 minutes, and the bot is again making it rain!
Certainly. It’s a subscription model. SaaS with 80% margins.
For the elder age-in-place skill set, the user would choose the desired skills from menu options.
If you want Grandma to have the premium services, then choose the Roadster level.
If not, then choose the Corolla level?
ralph
Where is this statement coming from? Back in Sept 2023, Tesla released a video showing a gen2 Optimus robot sorting blocks by color, a skill learned autonomously (using only its onboard neural nets) and using only visual data. InsideEVs concludes the same thing I do from the video:
Tesla says in the video that the machine’s neural net runs entirely on-board, using only vision, with the description saying that the neural network is trained fully end-to-end with video-in, and controls-out. Watch Tesla Optimus Robot Sort Objects On Its Own, Stretch Out In Update Video
Tesla’s Gen2 Optimus bot does NOT require external computers to learn tasks.
Where is this conclusion comng from? Milas Kunis only said they are collecting AI training data of the bot learning by imitating humans. Presumably this is to assess the efficiency of this type of learning. There is not even a suggestion that this AI training data is being used to train the bot. Do you have other information in support of your assertion?
2 Likes
Thanks for info. Great thing about these debates is one can learn about stuff one ordinarily would never bother about.
If meat processing goes on 24/7 it is certainly a good candidate for a dedicated specialize robot. But one can still imagine a plausible narrative for using general purpose humanoid bots. Which (if any) will be adopted will depend on the capabilities and costs of machines that have still yet to be produced. We shall see.
We do know that retail sales activity is seasonal and so the demand for different jobs in the Amazon warehouse will likely fluctuate. Being able to redirect robots to different tasks as needed would seem to have some productivity benefits. Probably why Amazon is testing humanoid bots.
While it is certainly true as others have mentioned that many warehouses/factories can be redesigned to optimize non-humanoid robots, it is also true that there are many locations that by their function are limited to a human-centric design. These include hotels, restaurants, schools, and such rapidly expanding facilities as elder care and health care centers. For these locations having bots with human dimensions and mobility would be an asset.
If so then what you describe probably won’t be the business plan used. My guess is that humanoid bots will be leased to companies. I doubt that Bot makers will charge for specific functions. Rather they will provide different levels of training and technical support. For simple warehouse tasks such training charges will be minimal, then increasing as the required tasks become more complicated, with world domination and enslavement of humanity being the most expensive.
3 Likes
Actually, not true for the warehouse. Retail sales activity can be seasonal, but the warehouse is designed to handle most of those changes. Why? Because that huge warehouse is not limited to a few items. The system they use is a basic one:
- Items in bins.
- What is in the bins is irrelevant.
- The computer knows which items–and how many–in each bin.
- Empty bins can be refilled with whatever will fit in them.
- So “seasonal” is based on what is expected to sell (by the SELLER) during some specified–usually relatively short–time period.
- Summer stuff next to winter stuff doesn’t matter.
- The warehouse is designed to handle ALL of it (that will fit, anyway).
- REMEMBER: Amazon gets PAID for each and every item stored in its bins (whether sold or not). SELLER PAYS.
- The SELLER determines if an item is added/removed.
- Amazon is a “warehouse space rental company” for sellers–nothing more, nothing less. Equivalent to a large storage locker. Pay to use it.
I’m no expert on retail, but it does appear that Amazon needs a lot of workers for the holiday season. That suggests significant seasonality in the work needed. Seems to happen every year.
That would be a lot of humanoid robots doing sales fulfillment on the holidays that could then have other functions for the other parts of the year.
1 Like
If there were other functions for the other parts of the year, then one might assume Amazon would have hired humans to do them already. Since they don’t, this is probably a work force that appears and disappears according to the need of quite temporary and predictable increases in product flow.
Will it be worthwhile to buy robots (or for a lessor or lessee) to have them available for 1/6th of the year? Of course for the other 10 months I suppose they could go to Jamaica and play golf or something.
I would think there would be a significant limit on supplying robots with pre-training in that some of the training would be specific to the individual work environment.
It will probably cost extra for the “sports” or “golf” apps
Mike
2 Likes
Obviously, your statement (quoted above) documents unfamiliarity with automation and employment. I have looked at using propriety (to the business) of automated assembly machines. Plus, many of the parts were made using CNC machining centers and other similar equipment.
Amazon requires more from an automatic picker/placer than just doing picking and placing of parts (placing = going into stock, while picker = for shipment to a buyer). The current level of AI programming is inadequate for Amazon’s needs. Amazon also does not need to use many mobile robots (i.e. on legs to walk anywhere). Instead, the items could be on fixed shelves in an automated warehouse with robots on tracks that run the length of an aisle. Height NOT a problem for the robots because they could have as many picking/placing arms as needed–and programmed to be used. Nobody said a robot could not be 20’ tall–or taller. Because it is all connected to the central computer, when an order is received and validated as being one or more item(s) in stock at that warehouse, the robot(s) can quickly pick the order and send it down to packing/shipping–likely within an hour, definitely the same day. This is already being done with automated parking (i.e. real cars, etc), so doing it for much smaller things is no big deal.
1 Like
If that is the case then Amazon can lease humanoid bots for the months that they need them. That’s the advantage of general purpose robots. The same machine can be leased to different companies at different times of the year to perform different tasks. Amazon fulfillment in the winter, construction in the summer, agriculture in the spring and fall. Just have to open your mind a bit.
I acknowledge your superior expertise on this issue. Just want to say that we are looking into the future at a time when humanoid bots will have capabilities equivalent to what humans currently do on the warehouse floor AND when the aging workforce of most industrial countries have significantly declined.
During the holiday season of 2023 despite great efforts to automate its warehouses Amazon still felt the need to hire a quarter of a million new human workers at higher wages. That suggests there would be demand for humanoid robots with human capabilities if these were available today at a competitive price.
Let’s see how Amazon’s trial with Agility’s humanoid robots works out.
2 Likes
Sure. But then they’re going to cost more than the rate that’s been posted for having one full time. Just as renting a car costs more for a few days than the ownership cost, or renting a hotel room costs more than typical housing cost on a daily basis. Just trying to keep the playing field level, here.
There comes a point where an open mind is just a hole with air blowing through it. A humanoid robot working in the dirty fields in summer and construction in the spring and warehouse in the winter is going to “age” and require far more maintenance than a purpose built robot which has the required maintenance protocols in place from the beginning (dust shields, traction feet for dirt vs concrete floors, etc.) I’m going to put this prediction up on the same bulletin board with others like “cars are going to be shaped like toasters” and “cars are smartphones, only with wheels.”
Yes. My emphasis added.
Oh yeah, 3D warehouse robot pickers and stockers:
Well, at least they have googly eyes.
1 Like
On the other hand, renting equipment when needed is not uncommon and is often better financially than ownership. Perhaps you are not aware, but it is commonly done in construction, agriculture, and many other industries. Why not the same for general purpose robots that can substitute as a temporary work force?
At what age does the brain become calcified to new ideas?
Rental robots are already here. It is just a matter of time before they become humanoid rental bots. Here’s how one company puts it:
“A robot’s working resource is measured in years, and in this time it can successfully complete tasks at a number of sites belonging to different owners. Intensive use of commercial robots will ensure fast return on invest for their owner and low cost of work for their renter.” https://smprobotics.com/robotics_business/robots-rental-and-leasing/
Forbes put it this way a couple of years ago as it extolled the virtues of renting and leasing robots:
“ That’s why a “rent-a-robot” model—which is gaining steam across the country—made so much sense. “Given today’s increasingly tight labor market, more manufacturers should explore these models,”…As the industrial ecosystem adjusts to the labor crunch, companies placing orders will seek out manufacturers with human-robot integration as the best way to ramp up production.”” https://www.forbes.com/sites/ethankarp/2021/10/25/rent-a-robot-is-a-game-changer-for-manufacturing/?sh=465118327a18
It is called Robotics as a Service (RaaS):
Robotics as a service is a business model where robotics companies rent their robots to clients and customers for short-term or long-term use. RaaS provides consistent revenue for robotics companies, while lowering the cost to entry — and potential repairs and maintenance costs — for businesses who want to adopt expensive automation tools. What Is Robotics-as-a-Service (RaaS)? | Built In
This seems a perfect business model for a humanoid general purpose robot.
I’m afraid the world will pass you by Goofy if you keep desperately resisting new ideas.
1 Like
Only if you have a use for a rent-a-bot. IMO, it is not yet ready. It will be years before the tech to offer robots for rent is even close to be ready.
Note Nissan’s recent concept car, designed for autonomous drive with an emphasis on creating a lounge-like interior and an exterior like a futuristic toaster (or as Autoblog says, an electric box on wheels).
Nissan notes that the Hyper Tourer is “geared towards individuals who appreciate the finer things in life.” In this context, that means a spacious, almost sci-fi-like interior with individual suspended seats linked by an arched center console. LED panels in the floors make you feel like you’re hovering above water while traditional Japanese kumiko and koushi patterns influenced the overhead console and lighting. Nissan Hyper Tourer concept is an autonomous electric box on wheels - Autoblog
When humans no longer do the driving, the emphasis will be on maximizing interior space. It is painfully obvious to those who think.
Meanwhile, Amazon funded Zoox just launched as a robotaxi what it describes as the first “purpose-built, fully autonomous, all-electric passenger vehicle ” Or is it a toaster on wheels? No-Controls Robotaxi 'Zoox' Just Received Its Driverless Testing Permit in California - autoevolution
2 Likes