I am reminded of two things:
First, this is a humanoid doing modern cleaning:
This is the robot version of that humanoid:
I personally have three of these things (one for each level). They will not make it 30 years.
3 Likes
Would a $250,000 one last 30 years? Or even a $25,000 one? Maybe. 
Heh, I felt like I paid $25,000 for the last one.
I have my doubts that the economics are ever going to be good enough to replace low/no skill manual labor (that requires a degree of finesse) with a robot.
Will it ever be cost effective to have a specialized robot that can clean a toilet, make the bed, remove laundry from the dryer and then fold it and put it away? I serious doubt it. I think manual labor will forever be cheaper for some jobs.
5 Likes
Some industries (mfrs) could offer a variety of features to “add on” to some product(s) that might make things better/easier. Consider AC in today’s cars. Common/required NOW, but not before (about) the 1960s.
Yup, I am speculating about the future (which I thought obvious). I believe in the near future (within the next decade) we will have humanoid robots capable of mimicking most of the simpler human physical behaviors and that can be mass produced for the current cost of the Model Y, if not cheaper given the smaller amount of raw materials and batteries needed. I don’t see any technical barriers preventing that from happening.
Just two examples. Here is the bipedal capabilities of Atlas (Boston Dynamics). https://www.youtube.com/watch?v=-e1_QhJ1EhQ
Here is the hand dexterity of the Tesla Optimus 2. https://www.youtube.com/watch?v=fPeGPf7jvEQ
Combine these with neural network machine learning and you have a humanoid robot capable of simpler human work tasks at the purchase price approximating one year at minimum wage + fringe. More likely, these will be leased to companies as needed to maximize work force flexibility.
This same data shows that this has not led to a reduced demand for low-skilled labor. From the MIT Sloan School:
Since 1917, American jobs have been classified by their skill level. “Unskilled work” was said to, “require no special training, judgment, or manual dexterity, but supply mainly muscular strength for the performance of coarse, heavy workdexterity.” It was the largest category of work and included mostly non-white and foreign-born workers who were employed in farm labor, factory labor, servant occupations, and “other labor” occupations. In 2020, the largest category remained “low-skilled work,” in farms, factories, in low-wage service and care occupations, and in on-demand gig and warehousing jobs. A Background Note on "Unskilled" Jobs in the United States - Past, Present, and Future | MIT Sloan
In short, nonhuman automation has not demonstrated that it can reduce the need for unskilled human labor in the general economy. It hasn’t done so for the last century. On the other hand, there is no reason I can see why a humanoid robot capable of mimicking the unskilled laborer would not be able to.
And since it is projected that due to aging demographics, most industrial countries will have a significant shortage of such workers over the next 50 years, I think it is very likely that the market for such humanoid robots will be very large.
3 Likes
The trends we are seeing today are that economies are getting larger, creating more jobs for unskilled labor. Populations are aging such that the human workforce is not projected to keep up with the need for unskilled labor. Technological advances in actuators and AI are leading to humanoid robots that can mimic multiple forms of unskilled labor.
Seems like there is a market there for such robots.
Here I believe is the gist of your position, which I think is based on a limited understanding of the capabilities of neural networks, deep learning, and machine learning as well as limited knowledge of recent advances in actuators and tactile sensors.
We have an economy where many jobs done by humans require multiple functions but at relatively low skill levels. I believe that the technology to build a multipurpose humanoid robot that can do these jobs at a competitive cost will happen in the next decade. You believe it is “beyond imagining”.
Time will tell.
1 Like
Not for nothing, but the market for unskilled labor has expanded dramatically as the economy rises. Using my own experience, when I was growing up in a middle class neighborhood, there was no such thing as “landscaping service.” You mowed your own lawn, or in rare cases (the elderly) you hired a neighborhood kid to do it. Now you can hardly drive down the street without seeing a truck and a bunch of guys happily driving a zero-turn (*a robot, of sorts) and leaf blower (*another) and tidying up the properties of half the neighborhood.
Back when, a restaurant was comprised of a cook in the back who knew how to assemble variety of ingredients and servers in the front who brought them to you. Now fast food restaurants have all but eliminated “the cook” (a skilled position) and replaced it with teen-age “assemblers”, and in some cases “robots” in the form of “warmer trays”, “fry machines”, and even (experimental) “Flippy”.
What happens is that the universe of low skill jobs expands as 1) other people move up in skill and 2) the economy allows people to buy more “service”, whether that is “home cleaning service aka “Merry Maids”, or nail salons (women used to mostly DIY), or “oil change” or “car wash” or “hair color” or “dog walking” or a hundred other things.
If the economy were to crash you can be sure nail salons and home cleaning services would be impacted. And, I note, we have “automated with robots” some of these tasks already. A car wash is just a big bunch of robots doing a single job: spraying, soaping, brushes rolling, rinsing, drying, etc. No need for a “humanoid” factor. Along with most of the others.
It will not be economically viable to have a humanoid robot mow the lawn, or frankly in my view, flip a burger. I’ll submit that there may be some uses where a humanoid robot will have advantages, but I’m pretty sure they will be fewer than you are predicting, and they are further away than even the pessimistic projections envision (much like FSD and other dreamy ideas.)
2 Likes
I’m not sure that you can conclude that ‘data has shown it has not led to a reduced demand for low-skilled labor’ from the statement that ‘In 2020, the largest category remained low-skilled labor.’
The largest category, low-skilled labor, might have shrunk and still be the largest category.
1 Like
If there aren’t enough human workers to do those jobs, what will happen to the economy? When scarcity creates a need, the market tries to fill that need. That’s why there is so much investment in humanoid robots. If the scarcity is humans…
Let’s not complicate the topic by broadening the definition of “robot” to be synonymous with “tools” or “machines”. Robots requires some element of independent or autonomous behavior.
It becomes more economically viable if a single humanoid robot can use human tools to mow the lawn, edge, trim trees, spread mulch, blow leaves, and move stuff. You know, multi-task. In this and many other instances, having one robot that can do multiple things adequately may be more cost effective than requiring several robots that do single things exceptionally.
Sure, but that’s a bit of a quibble. Largest category is the main point. If the number of human workers is projected to decline while a rising economy maintains the same need for human workers in spite of all our non-humanoid automation, what happens?
I’m guessing we build humanoid robots that can do many of these human jobs.
1 Like
Prices will go up for those jobs, and people will decide whether it’s worth it. Or technology in the form of highly specialized robots will replace. See: lawn mowing robots, not android robots pushing lawnmowers around.
It seems unlikely that nail salons are going to have android robots doing women’s nails, or that women are going to accept that as a good idea. Likewise for almost any of the examples I mentioned upthread.
OK. The definition, which I just looked up, of a dishwasher is that it is a “dishwashing robot”. Likewise a “clothes washer” and “clothes dryer”. They are “given instructions”, they follow a program, they accomplish a task. That might not fit “zero turn” or “leaf blower”, but it encompasses everything else from toaster to microwave to Roombas.
It becomes more economically viable if a single humanoid robot can use human tools to mow the lawn, edge, trim trees, spread mulch, blow leaves, and move stuff.
Well, you can continue to dream up this stuff, but cheap human labor is likely to be available in some quantity for-almost-ever. As the costs rise, automation will replace some of it (see: kiosks in fast food outlets) and those workers will downshift into other areas (landscaping, fence hole digging, etc.)
Doesn’t mean there won’t be dislocation of some and automation of others, indeed, that’s exactly what I expect. Humanoids? Very specialized, very expensive, very far away. Meanwhile, for your entertainment:
My general guy’s understanding of that task is that while nails get done, the main purpose is to have a conversation with a sympathetic human being. Nail salons are cheaper than a therapist AND you get nice nails to boot.
Of course, film makers have weighed in on robotic beauty salon attendants. See the film Wall-e, for example.
–Peter
2 Likes
I agree, since mobility isn’t needed and there is only a single task. In comparison, consider a grocery store. A humanoid bot could be used when the store is closed to refill the shelves, take inventory, identify misplaced items, and do security surveillance. It can use various sensors to identify and remove spoiled produce. When the humans are around it can move crates and make coffee. Or walk around and look for shoplifters and yell “Danger, Will Robinson” when it sees something suspicious. A 24/7 robot like that seems pretty useful to me. You seem to believe it is not technologically possible. I think it is technologically inevitable.
Yeah, but it won’t be nearly enough for most places not called India, Africa, or Central America.
If your theory is right, this should be a failure.
1 Like
On fingernails some people want chit chat and socializing, some to be slaved over by a menial, and some want their nails done well and as quickly as is possible consistent with “well”.
None of those are a slam dunk for robots, although the latter has possibilities.
Different strokes for different folks and the same applies to a lot else.
From the article:
… the bot can’t even remove existing nail polish, let alone shape nail tips, trim cuticles, or apply acrylics.
Hardly a manicure, then. Just fingernail painting. And if it can’t paint designs, it likely will fail. If you’ve got to remove the old polish and keep your nails shaped yourself, the work of actually painting a plain finish is pretty trivial.
But if it can do various designs, I would not be surprised if it survives as a niche product,
–Peter
And further, were it not for all of the automation, perhaps demand for such labor would be even higher.
No one can know these things, we can only guess at cause and effect and guess at the future:
We do know this. In the last year,
many nonhumanoid robots have been newly implemented
and
comparatively, basically 0 humanoid robots have been implemented.
Presumably many of those nonhumanoid versions will retain utility in the years ahead (like many such prior automations).
But I really have no strong opinion on something so unknowable as the future of robotics (other than it is moving forward).
Today, I’ll take one R2D2 (no offense to C-3PO).
I think you are missing the point. Automation does reduce the need for human labor in the task being automated. Automating the car assembly line reduced the number of assembly line humans needed. But it also made cars more affordable, which increased travel by the masses and greatly expanded, as for one example, the hotel industry. Need lots of workers to maintain all those hotel rooms.
The problem for economies with declining workforce numbers is that hotel housekeeping is not the most desirable profession and is more difficult to automate than say spot welding on an assembly line. What to do if there is a labor shortage?
Here is what we know. Humans currently do the job in a cost-effective way. So we know that the humanoid form factor works. I believe current technology has reached a point where it will soon be possible to produce a humanoid robot that can copy the behavior of human hotel housekeepers and so automate this job.
Goofy believes this will not be technologically possible for the foreseeable future, leaving a continuing need for humans.
You seem to be arguing that this automation will occur through the use of specialized non-humanoid robots: one to vacuum, one to strip sheets from the bed, one to clean the toilet, one to refill the soap containers, one to collect towels, one to empty the waste baskets, etc. Many machines that each do a single task very efficiently.
My disagreement with the two of you is that I believe that general purpose humanoid robots capable of basic human behavior will be developed in the next decade, and that there are many jobs in this economy where a single humanoid robot capable of performing many tasks adequately will be far more cost-effective than requiring multiple machines that do specific tasks even if very efficiently.
That’s why I think Tesla’s Optimus project could become a significant revenue generator rivaling its automotive business.
3 Likes
Entirely misstates my case. I think it will not be economically possible for the foreseeable future. Big difference. I have little doubt that a robot could be “trained” to make a bed (to use your example) but I doubt it is or will be worth it in the near to medium term.
To start with, the hotels reaction to “labor shortage” to this point has been to reward customers for not using this service, by discounting the room rate by around $5 per day. So there are other workarounds that don’t involve buying a $250,000 robot to do the most menial of labor.
More to the point, I have doubts that this super-robot will have the “onboard” capacity to make a bed of one kind in one suite and do the same for twin beds, or folding cots if the parents have brought kids into the room - in the same way Tesla is failing at having FSD execute flawlessly in the field. (That doesn’t mean it’s not possible, and someday it will be, the question is “at what cost?”)
Pre-programming a humanoid robot for every eventuality is fraught, will take a lot of (probably centralized) processing power for some years, and seems a gross misapplication of technology: using expensive doodads to do the most menial and cheapest of labor.
Indeed, I think the whole meme about replacing “cheap labor” is probably the wrong way to go about this: if there are to be humanoid robots in our future they will likely be used to replace expensive labor (although that’s already being done in law and medicine without the need for humanoid form) perhaps on automobile assembly lines (I have my doubts) or other highly specialized roles which are “overpaid.” (I offer the “pharmacy tech” who is essentially a dispenser of pills, making $100,000 per year and using the human brain to remember about “drug interactions”, although I doubt customers will accept this role, either.)
We have non-humanoid robots doing menial agricultural work (digging up tubers, shaking trees, planting seeds, pulling weeds) and while there are still a few which are devilishly hard, I doubt an expensive humanoid form would be much threat to migrant workers or other cheap labor either. You see? I don’t see much case at the bottom of the economic ladder, and at the top, well, human brains are used more than the brawn.
Where is the economic case ?
3 Likes
Training is probably a better word here than pre-programming.
Training means an existing program (not very purpose built or task-specific, except for maybe “humanoid” functions) consumes data for learning/training and updates some of its parameters/data. After so updated, that program can execute tasks in which input of new data leads to output of tasks executed (eg, see bed in certain way, then remove linens for subsequent laundering).
The program does need to be trained on all tasks in some way, so it is pre-programmed in that sense, but not the more traditional sense.
2 Likes
Yes, I agree. That said, needed to do that is a crap-ton of electricity and somewhat immense processing power - not the biggest issue when it’s in a $50,000, 3 ton automobile with a thousand pounds of battery on the chassis, quite another when it’s “humanoid” weighing, what, a few hundred pounds total ?
Yes, processing power is still coming down in price, and batteries too, but even extending that out I think it’s a long way off, practically speaking, at least for this application.
So. Does each robot “learn”, or is there a centralized location to run the experiences and then retransmit out to the field? If so, how much can each one “carry”, and would it be enough to handle all the varied and simple-yet-intricate tasks including edge cases (spilled water on the bathroom floor, towel tied in a knot, food mess in the microwave, bed sheets soiled inappropriately, missing TV remote, etc. etc.).
Eventually, assuming continuing technological progress, I would say yes. Anytime soon enough to make an economic case for deployment? I would say no.
My best case for humanoid robots involves dangerous work: cleanup at Chernobyl, walking around Mars, that sort of thing. Menial labor? I don’t see it.
1 Like
Jumping in to raise two points, again: the fact that the humanoid form factor works doesn’t mean it’s the best or most efficient form factor for the job, and for a humanoid form factor to be worth a durn it requires a “brain” running the robot that’s far more advanced than anything we can build today or likely for a decade or more.
Yes, it would be awesome to have a super-advanced robot brain. But we don’t have one. And if we did have one, for many (most?) situations we probably would want to put it into bodies that don’t exactly track human ones - bodies that are better suited for various jobs.
Again, what leaps to mind are legs. We use legs because we evolved that way - and half our body (by length) is made up of legs. But legs are pretty much useless for most jobs we would ask our robots to do. Our future robots would have much more stability with a quadrapedal or even a flat base with wheels - or build them around a cylinder that you could store tools or materials in rather than half the body being made up of useless legs.
For many jobs, a rolling cabinet with arms is your best bet. You don’t that cabinet to be shaped like a human torso with legs.
2 Likes