A Musk hype account posted an AI generated video of multiple Optimus robots shoveling snow, showing yet another use for the as-yet unavailable humanoid robot.
The internet is not having it, noting that what has just been invented is “a worse snowplow.” Heh.
The key difference will have to be flexibility. If the robots can be “cross trained” then they can shovel snow in the winter and mow grass in the summer. A snowplow sits idle most of the year.
DB2
The snowplow will still be sitting idle. We will sit idle as well. Yet we will need feeding regardless. Talk about getting fat.
Only the plow, which is probably 5% of the tool. The rest is a truck, which will be used for hauling gravel, repaving roads, carrying garbage, or whatever. You know, just like a humanoid robot with a shovel, which doesn’t always have a shovel.
And a snowplow will be 1,000 times more efficient. Hard to do a sidewalk with it, I’ll admit.
Actually, I was thinking of a snow blower.
My point was that if you spend the money to lease or buy a robot it needs to be multifunctional.
DB2
Does it? I have a robot which takes me from place to place on its four wheels, but that’s all it does. I have another that keeps food cold, one that vacuums the floor, another that makes ice, and yet another that heats and cools my house. None of them do more than one task, but they do it well.
I’m sure there’s a place for one that can make the bed and wash the clothes (a robot putting the clothes into another robot, then transferring them to yet another robot, and finally folding them), but it’s not like I don’t have machines all over the house doing tasks for me. One task each, pretty much.
Robots, as I understand it, are machines which use sensors and programming to complete a task, or a series of tasks. Roomba, washer dryer, ice maker, refrigerator, dishwasher, TiVo, microwave, air fryer, automobile and more - I’m surrounded by them. Hard to think of any which do more than one thing, tho. Maybe heat pump which runs forward for heat and backwards for cold. Otherwise, one task, one robot.
I guess that’s the promise, and I see a couple applications (nursing homes, where they will scare the bejeebers out of the residents, stores stocking shelves, and maybe assembly lines, although I think single-purpose robots are likely to be more efficient) but this is going to take a very long time to develop and market. I think Tesla is wayyyy out over its skis, but maybe that’s just me.
Anyway, do a task and do it well has been my mantra for a while. Do many of them, but not as well doesn’t really roll off the tongue, does it?
This is how I expect Optimus to perform when trying to shovel snow:
Reading this thread, I’m reminded of why fraudster Trevor Milton decided to create a narrative that his company was working on EVs for consumers as well as for trucks. The retail investors he attracted were happy to see electrication applied to something they might own and use, while paying less attention to the larger market of trucks for businesses.
The same appears to be true here for robots. While I admit it’s fun to let the inner Asminov in us come out, the early applications for robots will be on assembly lines, in manufacturing, in warehousing, and other businesses that today rely on a flexible human workforce to adapt to the needs of the business, which may change on a daily, if not hourly, basis.
Sure, there are manufacturing applications for which dedicated robots make sense - anything where the volumes are consistently high enough that they can be kept busy for however many shifts are needed. But even in something like automotive, no assembly line is human-free. Many more businesses employ workers that are given mulitple roles, and should a robot be capable of handling enough of those roles to be kept busy enough to be worth the investment in money and training, those robots will be commercially successful.
To bring this back to where this board’s collective heads are at, I assume you have a cell phone. Is it a really good phone, easy to hold for long discussions, and are you able to write something down while you hold it? Do you will have a landline? Does it have video? Does your business still have dedicated video meeting hardware? Is your cell phone a really good camera? Do you use it as a calculator? As a remote control for your car? How about as a device to represent your airplane ticket and boarding pass? How about as a way to order products, track shipments? Do you still buy Walkmans? Has the cell phone replaced the broker you used to talk to decades ago to buy and sell stocks? Has it replaced ticker tape machines? Do you use it as a clock, alarm clock? Did it replace your Rolladex? Do you configure your Wi-FI or other electronic devices with it? What about trip planning - do you still go to AAA to get a paper TripTik?
Now, go back to the 1980s and 1990s, and remember all the people saying they didn’t need nor want cell phones. Now tell me this time it’s different.
I think that’s completely wrong. The early applications for these kinds of robots won’t be in any of those fields. Because those activities take place away from ordinary people, in environments that are completely controlled by the business and are already designed to be accommodating to non-humanoid robots. They’re also jobs that don’t require a flexible human workforce, since even the human jobs are mostly repetitive and within a narrow field of activity. The humans that are breaking down poultry in the processing plant or straightening candle wicks are doing that all day long, not switching between dozens of tasks. So nearly job that can’t today be done by a robot, but could possibly be done by a robot with better AI, in these environments is probably going to be done by a non-humanoid robot.
Most likely early application of a humanoid robot? Not any of that. I suspect it will be package delivery.
The only thing that we have actually seen humanoid robots be any good at - at all - is moving themselves around. They’re learning to walk, climb, and even dance really well. The only other thing we see them do on occasion is take an object and carry it to another place. Not manipulate the object in any way that requires knowledge of how physical objects respond to forces - but to carry it.
So package delivery actually fits that narrow intersection between what humanoid robots might be able to do in the world without a ton of data (that they don’t have), and a need to be able to navigate a human-oriented space (filled with stairs and doors and elevators) different from the inside of a warehouse or factory. Packages for delivery are light, the main obstacle to complete automation is the “last mile” of getting from the curb to the doorway, there’s an actual advantage to a humanoid form, and you can adapt the packages so that they’re easy to be carried by robots.
OK: It’s different. All of those things are done digitally. Every single one. (Phone, calculator, video, camera, etc.) With a humanoid robot, not a single thing is done digitally - so it requires form, weight, sensitivity to pressure, sight, sound, and all the other things that shape the physical world. They have about as much in common as apple and auto tires.
I’m not saying there might not be any applications for the Optimus & similar, although it seems wildly egotistical for us to think the human form is the penultimate for doing so many things. I mean, why two arms? Why not four? Or six? Why not travel on four legs, or on track mounts? Why not use lidar and radar as well as vision?
PS: There are already completely automated factories using hundreds of robots, not a one of them humanoid.
Yes, [dark factories] (or "lights-out factories") are operational in China, heavily utilizing AI, robots, and automation to function without human intervention, lights, or heating, particularly for producing EVs, smartphones, and electronics. Examples include Xiaomi's Changping Smart Factory, Jetour's Fuzhou factory, and various facilities under the "Made in China 2025" initiative.Key Aspects of Dark Factories in China:
- Definition & Function: These facilities are entirely automated, relying on robots and AI for assembly, packaging, and quality control, which allows for 24/7, lights-out operation.
- Examples:
- Xiaomi’s Changping Smart Factory: A $330 million facility in Beijing that can produce 10 million smartphones annually.
- Jetour’s Fuzhou Dark Factory: Produces SUVs with near-total automation.
- Changying Precision Technology: A factory that replaced 90% of its human workforce with robots
Note to albaby: not package delivery. Sex toys. Big driver of early adoption of VCRs. Of HBO & adult channels. Of streaming. Of phone pay-per-minute services. I’m guessing there will be a market, and I won’t detail how or what, but leave it to the readers’ imaginations.
That’s not the thought.
The thought is that a robot that is human form can more easily fit into existing workflows that were built around the human form. Robots replacing humans without changing anything else.
Yes, as I said:
But, there are many jobs in many industries that aren’t high volume enough or profitable enough to justify dedicated special-purpose robotics. A human form robot could, as I also said, be kept busy doing a variety of jobs that single-purpose robotics cannot.
The way I see it, the main obstacle for Optimus is going to be the training. Elon says FSD needs 10 billion training miles to be unsupervised. Waymo trained on 15 billion virtual miles.
Elon said on the conference call that you would be able to upload a video or just demonstrate to Optimus how to do something. That’s not how today’s neural networks work. You would need to upload 10 billion training videos or demonstrate 10 billion times to Optimus for it to have any chance of working. That’s going to make the cost to use Optimus exponential.
Absolutely correct but you haven’t said why that is, so I’ll reveal the secret.
All the above are now done by humans and business knows their cost. If the humanoid undercuts the human then buying the robot is a done deal.
The Captain
There are many tasks robots could do that are simpler than driving.
What is simple for a human can be very complex for a robot. I suspect this is why we haven’t seen any demo videos of Optimus doing useful work in TSLA factories. Elon hasn’t gotten Optimus to do anything useful yet.
Define “useful.”
The Captain
A task in a factory that is currently done by a human.
That won’t sell a robot unless it is cheaper and more reliable than the human.
The Captain
Yep, that’s why I think TSLA will sell very few robots.
I keep forgetting that the mantra of this board is that whatever Elon decides to tackle, no one will want:
• EVs: Well, Elon looked to be right for a while, but now no-one really wants EV, amiright?
• Robots: See above
• Self-driving: Yeah, Google is wasting money on Waymo, and its current capital raise at a $125B valuation is stupid. /s
• Boring Company: Tunnels are too limited.
• Nerolink: Bad Sci-Fi
• Space-X: Elon’s not really that involved, which is why it’s successful. /s
• xAI: Why does Elon think he can do AI?
Um, well, never mind that Space-X and xAI just merged!