OK, then. Let’s not bother to reduce CO2 emission in the US because we can’t fully offset the increases coming from other countries. Let’s just lay down and give up instead of doing what we can.
Makes perfect sense.
3 Likes
But once you replace the humans with robots, there is no reason to continue to design the environment for humans. You can design it for robots with form factors better suited to the task at hand.
If robots are going to work side by side with humans, sharing space designed for humans, then it makes sense to make them humanoid.
Why aren’t the robots moving goods around the Amazon warehouses humanoid? They share that space with humans. Make a humanoid robot and give them a dolly or a cart to move the goods. No, they made a robot that does the task that needs to be done, using a form factor completely unrelated to the human form.
3 Likes
And yet the robots we send to wander around the Moon or Mars are not humanoid. They’ve had wheels. I’m pretty sure those environments were not build for humans. 
One even had wings (of the rotor variety). Well done, all you nerdy types who put together Ingenuity. Flew 72 of it’s scheduled 5 missions. After Three Years on Mars, NASA’s Ingenuity Helicopter Mission Ends - NASA
4 Likes
Fossil fuel companies are sowing that sort of disbelief in every channel. The idea is to leave the fossil fuel companies with trillions in profits. Just BS the entire world that no progress can be made.
No, that’s not the point. We should of course reduce CO2 emissions as much as possible. But the point is, where do we spend our effort jawboning about it? Do we rant and rave against ICE vehicles, and make all sorts of unrealistic pronouncements about them mainly to make ourselves feel better. Or do we at least attempt to discuss where the best efforts should be directed to get the biggest bang for the buck when it comes to CO2 emissions reductions?
Do we actually try to do something useful, or do we simply ridicule those who want to talk about the 2 million extra tons of CO2 and instead focus on the 100,000 extra tons of it?
3 Likes
How unrealistic? The goal might be close to met. Why screw around with the verbiage just because the oil companies are giving you ideas that suit them?
The oil companies have over $1 trillion in the ground which makes them want to fool the heck out of the entire world. Stop reading how nothing will work. It is BS to fool you purposely. It does not matter what else has happened in the past.
Jawboning costs nothing. So pretty much anywhere and everywhere.
Actually doing things that cost money, that’s where the rub lies.
Or do we at least attempt to discuss where the best efforts should be directed to get the biggest bang for the buck when it comes to CO2 emissions reductions?
Ideally, that’s what we should do. Unfortunately, steve is not entirely wrong about the outsized influence of JCs on public policy. Trying to limit CO2 emissions by any particular type of business is going to be met with significant pushback from significant campaign donors. But limiting CO2 emissions by consumer devices (like cars) where the costs are mainly borne by the consumer is a much easier thing to accomplish politically. So that’s what actually happens, at least in the US, and likely to some extent in most of the industrialized world.
Getting done what can get done is better than not accomplishing a better solution. Keep working for that better solution, but take what you can get when you can get it.
–Peter
We have started to put in newer electric grids. That will save us on fossil fuels in a huge way.
India and China are going EV. Yet in this conversation, some are worried India and China are going entirely the wrong way. That is misinformation.
I wonder if that includes the costs to the remaining staff of keeping them running. Not just “maintenance”, but actually giving them a kick in the pants every time they stop. Like, um, this:
https://www.wsj.com/business/robots-taking-jobs-still-need-humans-warehouses-63bc0306
Companies Brought in Robots. Now They Need Human ‘Robot Wranglers.’Wandering and confused cyborgs create a new job. ‘We’ve found them on a receiving dock, just lost like a child in the park.’
Caroline Rutenberg got an alert one day that one of her charges needed help at the [Amazon.com] warehouse where she works in Windsor, Conn. walked across the warehouse floor and found the crew member covered in white paint and refusing to move after an accidental spill.
Rutenberg took a cloth and gently wiped off the worker’s eyes, getting a glassy, expressionless stare in return. But it’s not like she expected a “thank you”—her underling is a robot, a squat machine on wheels that looks like a Roomba vacuum with a conveyor belt on top.
Androids of various shapes, sizes and functions are now employed doing everything from autonomously ferrying goods around warehouses to [delivering food on college] campuses. About 21% of [warehouses used some form of robotics]in 2023, up from 15% in 2018, according to research firm Interact Analysis. But far from [replacing humans] entirely, companies are finding cyborgs need a little hand-holding to learn how to function in the real world.
So while I have no doubt “robots” will continue to proliferate in settings where it makes sense (warehouses, agriculture, etc) I am equally sure that most will not be “android” but will be purpose built to the task required. As I say, there will likely be some in human form, although it’s difficult to imagine why we would spend $250,000 for the form in great numbers when cheaper, task designed and more reliable would suffice.
The 25 robots at Roper—which come in a variety of shapes, including one that Samples says looks like “putting wheels on a pumpkin”—work on tasks such as delivering parts and materials to the assembly line and moving finished products. They are programmed to follow a digital map of the facility and to use camera vision and light detection and ranging technology to stay on course. But occasionally they stray out of bounds, where they can no longer locate themselves on the digital map, shut down and wait for help.“We’ve found them on a receiving dock, just lost like a child in the park,” Samples said. Other times the cyborgs are “over by somebody’s desk or trying to get under their desk,” he said.
When Samples gets a call about a robot wandering off, he pulls up that particular bot on his computer and looks out of the cameras and sensors attached to the device to try to figure out where the android is within the 1.4 million-square-foot facility. He then goes out, manually finds the robot and guides it back inside its boundaries.
The era of “set and forget” isn’t going to include robots that mimic the human form, at least for a pretty long while.
3 Likes
Well Agility disagrees with you. 10,000 humanoid robots a year to start out with is nothing to sneeze at, and that is just from Agility. This is just the start of their production also.
Andy
I don’t think that’s accurate. Looking at the article, it doesn’t say they’re building 10,000 robots any time soon. It just says they’ve got a facility that has the capacity when it is fully built out to build 10,000 robots in a year. That’s not where they’re planning to be in the immediate future:
At first, production will be in the hundreds, but eventually RoboFab is “going to have a significantly larger capacity of 10,000 robots per year, peak,” Shelton says.
Still in early days, yet.
https://www.axios.com/2023/12/05/humanoid-robot-factory-agility-bipedal-amazon
2 Likes
That is what I was talking about tightening. Investing isn’t about the past, because if you wait till it is past, you will have missed it. Investing is about the future. We have already missed this because Agility has the production facility almost ready to go for 10,000 robots. Now could something happen where they still fail? Yes but the ability to make them is already here. Now the question becomes will they make more factories? Nobody can see the future so you make bets on the probabilities of future outcomes. When the future gets here you have missed it.
Andy
2 Likes
Again, I think you’re reading something into the article that isn’t there. They have a property and building big enough that one day, it could produce 10,000 robots per year. That doesn’t mean that they have the equipment and production lines installed in that building to produce 10,000 robots a year. From the article, it appears that they’re concentrating right now on just finishing the first production line, with a capacity of a few hundred robots.
Investing is indeed about the future - but if you’re making assessments about how close or far that future is based on what’s happening in the present, it’s important to be accurate in assessing where the present is. Agility hasn’t yet built out production capacity for 10,000 robots. Like many other robotics firms (of which Tesla is only one), they’re making an initial limited stab at these products. They’re not currently proceeding on the basis that their current iteration of the product is far enough along to generate a demand of 10K per year.
2 Likes
I think you are understating everything. The article states: The new factory, which Agility has dubbed the RoboFab, will produce up to 10,000 units a year and employ 500 people, according to COO Aindrea Campbell, formerly Apple’s senior director of iPad operations.
So the company does plan on producing 10,000 robots a year. Of course you will (future tense) argue about if it will happen. But that is the companies plan and I take them at their word, while you will wait to see if it happens. Much different outlooks on the world and one we could argue about all day. Just like when you said we didn’t have humanoid robots in factories. They are there but maybe now they just are not humanoid enough?
Andy
1 Like
“Up to” 10,000 robots a year. They’ve designed a facility that they can grow into if their product generates adequate demand. That’s very different from saying that they are at “10,000 humanoid robots a year to start out with.” That’s why I looked for a more recent article than the one you linked to (which was from September of last year) to see where they plan on being on opening day. They’re starting out with a few hundred, not 10K. We’re not yet at the point where humanoid robots have proven themselves to be a viable product.
https://www.axios.com/2023/12/05/humanoid-robot-factory-agility-bipedal-amazon
It’s not that they’re not human enough - it’s that they’re there as part of pilot and test programs. They’re being “deployed” in the sense that Amazon is experimenting with them. In contrast to the hundreds of thousands of non-humanoid mobile robots that fill Amazon’s warehouses already, that are already deployed as part of Amazon’s operations, not a pilot program.
I find this discussion about what form robots will take quite humorous.
AI will make the final decision for us. I assume they will take many forms, depending on the need. Some best as humanoid, maybe some will look like spiders or an octopus.
I’m hoping if a bot needs to have large legs, a monstrous jaw, and itsy bitsy hands, it will look like a T-Rex! They say I’m a dreamer ……
3 Likes
And then came AI.
Musk jumped into humanoid bots because he believes the major problem facing humanity is depopulation, an aging and declining global population will reduce the human workforce and thereby reduce prosperity, hence the need for better automation. He also found that “dumb” robotics could not produce the kind of automated gigafactory he envisioned. He needed smart general purpose bots that could do multiple tasks and also quickly learn to do different tasks as car production evolves, which at Tesla it constantly does.
He feels that the humanoid form factor is best suited for that type of higher order task flexibility (why argue with natural selection and evolution?).
The goal isn’t to completely replace humans. The goal is to find the efficient balance of robot to humans. Suppose it turns out that the most efficient way to automate a fast food restaurant is to have Bots do 75% of the work and humans present to supervise and fill in as needed during the rushes. Now you could have a specialized robot make burgers, and another one make fries, and another one make milk shakes, and another one clean the bathroom, etc. Or you could get a couple of general purpose humanoid bots that can do multiple tasks without having to redesign restaurants for non-human form factors and that can also learn how to make ten varieties of chicken sandwiches when you decide to expand the menu.
I think the general purpose smart bot is a better investment and advancements in AI make these possible.
Don’t be so sure they are not. This is more in the weeds than I want to go, mainly because I don’t understand it, but here’s a modest attempt. The oldest and presumed leader in the field of humanoid robots is Boston Dynamics that I believe is using a basic form of machine learning. For their bot to learn how to pickup an egg without breaking requires a “model” or algorithm that defines the decisions that have to be made. In this case the model would say that the bot would have to adjust the pressure used to pickup the egg. The bot would then learn either by being fed structured data (e.g., strength of egg shell) or by trial and error, also called reinforcement learning.
Tesla is using a form of machine learning called neural networking. In this case there is a hierarchy of algorithms that by some magic allows the bot to make decisions and learn using unstructured data such as visual images. The bot can learn by simply watching something similar do the task correctly. That’s why a humanoid bot can learn from watching humans. Now, keep in mind that this is a highly simplified description made by someone who admittedly doesn’t understand the field. Hopefully someone will correct the errors. But if this is broadly true, then the efforts Tesla has made to develop FSD via neural networks and info primarily from optical cameras could directly benefit developing similar neural network bots that can learn to do much simpler tasks.
Tesla has also already vertically integrated much of Optimus manufacturing. Tesla designs and makes its own actuators, batteries, and software.
I think this statement is based on ignorance. Musk for example understands first hand what type of robots would improve his gigafactories because of his earlier mistakes with automation. He sees the limitations of current forms of automation with respect to the building EVs the Tesla way and is designing Optimus to overcome these.
2 Likes
Well then if we are going to go by your parameters you are incorrect. They haven’t built a few hundred. I have only seen one so that is all they have built. I think you are inflating the number by saying a few hundred. Can you prove that they have built that many?
That is how everything starts. You can’t start from the end, that is why they call it the beginning. You have to be able to project into the future or you will never get there. Who would have ever thought that Tesla would have the number one selling car in the world?
Andy
2 Likes
There will always be a market for specialized robots. That doesn’t mean there won’t be one for humanoid robots.
Suppose you build your automated factory with a bunch of robots that do a single thing well. That’s how the current ICE OEMs build their cars. It is very efficient, but because they have invested so much capital into this infrastructure, much of the future development of their product is constrained by what these factories can build. This is why the initial OEM BEVs generally suck rocks. They just tacked on a battery to the cars their factories could produce rather than design a car that optimizes the use of the battery.
In a reality where technology is changing at an increasingly faster pace, future manufacturing will increasingly require flexibility. That is what Tesla is demonstrating to the auto industry. Nothing we know of so far in the universe is more flexible with respect to doing different tasks and learning different tasks than humans. That’s why it makes sense to design robots to be like humans if the goal is to automate a factory capable of adapting to rapidly changing technology.
3 Likes