[GuB-42]

The title sounds to me like: I am going to spend $1000 in groceries and dance lessons. That is, two very different things lumped together.

Memory chips are like groceries, essential commodity parts, a no-nonsense investment. Humanoid robots are like dance lessons, it is cool, it is sexy, and it may pay off in the future, but the value is much less certain.

[timerol]

$585B on new fabs, $357B on AI data centers, and $5.8B on humanoid robots. One of those numbers is not like the others

[hobofan]

I don't have the historic numbers at hand, but I would assume that for each of those categories this is a similar proportional increase, so it's similarly notable to mention the increase in spending on humanoid robots.

[b112]

Androids (humanoid robots) will require loads of ram, and loads of model training under the current paradigm. So it sort of makes sense. At least, I see robots as the top of the pyramid.

[gavinsyancey]

Autonomous (non-teleoperated) humanoid robots that can do useful work in an unfamiliar environment do not exist. And nobody's close enough to making them to understand if they're possible with our current level of technology, let alone how.

[mkl]

Most initial work for them would be in familiar, well-controlled environments - replacing humans in existing factories. I think whether they'd be cost effective for that will remain unknown even after a few years in service though.

[gavinsyancey]

Factory work that can be straightforwardly automated by robots generally already is, by special-purpose factory robots. The remaining tasks are generally:

* Tasks where poorly-paid humans are cheaper than expensive factory robots. Humanoid robots are more complex / fiddly so will be more expensive than existing factory robots. No help here.

* Tasks where human dexterity has an advantage over state-of-the-art robot actuators (e.g. sewing fabric panels into garments). Better robotics could help here, but the advancement needed is better actuators, not AI and a humanoid form-factor. And if you solve this you'd be better off putting your new end-effector on an existing 6dof platform.

* Supervising robotic equipment and handling exceptions. But then you get to "handling poorly-specified unfamiliar tasks in the physical world" which is not currently a solved problem and there's no guarantee just throwing more compute at it will be sufficient to solve this. So far all humanoid robot demos have either been either known tasks in a tightly-controlled environment, teleportation, or so poorly-functioning as to be obviously not fit for purpose.

[cogman10]

Totally agree.

I'd also point out that what makes this whole thing smell of being a grift is the fact that what's being chased is humanoid.

Humans are not the pinnacle of dexterity or stability. And freed from biological constraints, it makes no sense why we'd use the human form as a reference.

Making a humanoid robot is a hard thing to do, for sure, but it's also not particularly useful. The routines needed to balance a robot or correct for a slip are interesting to solve, but don't really make for a better robot which is more capable of doing the dishes or folding the laundry.

If I'm amazon, for example, then the most useful form factor for a general purpose robot is 2 arms on a 4 wheel omnidirectional rolling platform.

[p1esk]

We’re experiencing gpt-2 moment in robotics now. This means in about 2-3 years they will do useful work (cooking, repairs, cleaning, etc).

[ben_w]

The extrapolation cannot be justified. It may be much longer or tomorrow.

[gavinsyancey]

So far all humanoid robot demos have either been either known tasks in a tightly-controlled environment, teleportation, or so poorly-functioning as to be obviously not fit for purpose. It's possible throwing more compute at the problem will work and they'll be useful in a few years. Or maybe we're actually experiencing a Markov-chain / ELIZA moment and are multiple decades away from anything actually useful.

[cogman10]

I'll do you one further.

Self driving cars is far from a solved problem. It's something we've been working on for the last 20 years at least. We are getting closer, some solutions are impressive (like waymo). But even those ultimately need operators in the area of the cars to solve the problem of the car getting stuck.

Self driving cars are an infinitely simpler problem to solve vs a general purpose humanoid robot. You have basically 2 outputs, acceleration and steering. You have rules simple enough that I was driving at 14. With enough input, you'd think self driving could be completed in a snap. But it's not there yet.

Humanoid robots which are useful will come after widespread deployment of L5 self driving cars. Since we don't have that, I have no faith that we are close to useful humanoids.

[mathgeek]

They're already driving, as an example of why your claim of "gpt-2 moment" isn't as far out there as it may first seem.

[dyauspitr]

We said the same thing about Waymo, that it was perpetually in the future. It took them less than a decade. The robots today are functionally capable, they don’t have the right fuzzy intelligence yet. It’s purely a data problem (lack of) and a lot of people are working on it.

[darkwater]

Are you saying autonomous driving is a solved problem, even at scale? I haven't seen any Waymo in my small town in Southern Europe yet.

[OkayPhysicist]

From what I've seen with Waymo, autonomous driving in relatively good conditions (light to medium rain, fog, or sunny) is a solved technical problem, even in small, winding streets. Snow is the next big hurdle, and they're actively working on that in Denver and Detroit.

They're being conservative with their rollout mostly because of civic issues. Most places do not have a legal framework for "what if your autonomous vehicle hits someone?" yet. Even if Waymos never were at fault for a collision with a person, you can always have cases like the one in Georgia back in October where a bicyclist wasn't looking where they were going and rammed into one. The shaky legal ground is a pretty big impediment right now, and that's in the US where we have much laxer laws about corporations killing people.

[dyauspitr]

If they can pull off year round driving in Manhattan, and they have, it’s a solved problem. Rollout it slow due to bureaucracy, appetite for progressiveness and catering to entrenched business interests but it will show up everywhere it’s profitable eventually.

[sdfsdfs34dfsdf]

> "small town" in "Southern Europe"

I've highlighted the two main issues you are currently experiencing.

[logicchains]

It's not just a data problem, it's a hardware problem. Transformer-based robots require even more processing power than plain LLMs, as they also need to process visual and spatial/touch input. We don't have GPUs capable of fast inference on a SOTA LLM that would fit in a robot brain form factor, let alone also run fast enough spatial and visual processing. And there's currently nothing even approaching a feasible solution for cooling such a device.

[b112]

https://www.1x.tech/

[RealityVoid]

I have been following them since they were Halodi robotics. They're cool, but nowhere near the level of autonomy you need. My theory is that before household robots become a thing we should have self driving cars be a common occurrence, since that is a much much simpler problem.

[red75prime]

If there's no unknown unknowns in the brain, it's most likely possible. As the universal approximation theorem and empirical results of scaling SGD+RL suggest. Whether it will be economically viable remains to be seen. The human cerebellum has a peculiar structure and 80% of the brain's neurons after all.

[ben_w]

The parameter count equivalent of a human brain is not yet known, but if it was one per synapse then a full human brain replica would need about 1.5e14.

We also don't yet know how to be as efficient with training examples as any living creatures' brain, and we only partially make up for this by training on so many examples it would take you a million or so years to do the same, so we'd still stuggle with something proportionally smaller-brained such as a cat.

That said, remote controlled androids are going to be economically disruptive, as they make every (unlicensed) job open to outsourcing from an office in a low wage country.

[xvilka]

Real neurons are orders of magnitude more complex than their artificial pseudo-approximation (it is all based on the century-old understanding of how neurons work). You can think of _individual_ biological neuron as an analog of the small artificial neural network. You can see this simple visual explanation on YouTube[1]. So we aren't even close. It doesn't mean the AI is impossible, it just means people underestimate the "computing power" of real brains, as well as that AI, even the future one might be totally different in how it works from the natural intelligence.

[1] https://www.youtube.com/watch?v=hmtQPrH-gC4

[jampekka]

Brain's information flow also isn't just through direct neuron-to-neuron connections. Firing also releases neuromodulators into the extracellular space which affects how other neurons operate. Furhermore neuron connection architecture is very different between brain and feedforward ANNs, with the former exhibiting a lot of recurrent connections.

[red75prime]

But for biological neurons to do something that can't be efficiently approximated on a digital computer (but conductive to useful information processing) they need to have unknown unknowns (well, partial unknowns like an unknown quantum algorithm will do too).

We don't know the violations of the physical Church-Turing thesis that are conductive for machine learning. We don't have evidence for their existence in the brain (although, the brain would be the prime candidate for finding them as evolution works directly with the true physical laws).

BTW, large ANNs don't try to model how the brain does things. They are trying to mimic what the brain does. So, using "how many transistors/artificial neurons it takes to model a biological neuron" is not a good approach.

We have no evidence. We even have no solid theories how this can work (Penrose's OrchOR is "OrchOR somehow taps into mathematical knowledge somehow encoded into the structure of spacetime"). But people, for some reason, insist that there should be something there. I can't attribute it to anything else but to deeply entrenched feeling of human exceptionalism.

[wqaatwt]

We certainly don’t need to replicate humans or overly anthropomorphise robots. Just like cars didn’t need to imitate mechanical horses.

[Dylan16807]

If you're running a massive model for logic you're probably better off not putting it in the robot. And it'll be a long time before there's enough robots to make up a significant share of usage.

More basic movement control doesn't need loads of ram as far as I know.

[b112]

The same logic for why self-driving cars can't be cloud based, applies for robots. Something cannot be in the middle of a delicate operation and then "oops!", no network, it just stops.

The larger the context window, the better with models. Having a few TB of RAM would be exceptionally helpful.

All this just made me realise something however. Having your robot dormant and charging, is a bit of a waste. You could have robots dormant, but its compute in use to act as a compute node. If the distribution of robots is similar world-wide, we'd need a fraction of the datacenters we have now.

Using such nodes for training purposes would be beyond advantageous. And the company which can slice up the work and having training done in batches would get the big bucks. And actually, with consumer facing products soon all laden with extra ram and gpu for local compute, that applies there too.

Imagine leasing out idle time on your desktop or even laptop for cash. There may be a market here, especially with the cost of new datacenters. Any company able to securely package compute without risking data safety is going to make a mint.

Anyone have any ideas?

[Dylan16807]

> The same logic for why self-driving cars can't be cloud based, applies for robots. Something cannot be in the middle of a delicate operation and then "oops!", no network, it just stops.

I don't think you understood my post. The equivalent of self-driving is the movement control I was talking about.

Self-driving cars don't have high level logic, except for route planning. Which often is offloaded to the cloud. An extra 30 milliseconds on understanding your speech is nothing.

> Imagine leasing out idle time on your desktop or even laptop for cash.

https://vast.ai/

[b112]

You're missing the point. The issue is "how is the thing used" not trying to make identical break points in differing tech. A self driving car cannot have object detection, collision avoidance, human detection offloaded to the cloud. Ever. At all. A network drop can't mean it smashes into things. Or stops unsafely.

The same is true with an android. Imagine it turning on an frying pan, cooking dinner, and then going offline part way through. Or turning on a tap to wash something, and going offline while the sink overflows and destroys the house.

There are myriad of such scenarios, but local compute is absolutely, 100% necessary. Anyone betting the farm on putting network controlled devices into homes for any serious task is going to lose their shirt. Local compute is an absolute requirement, and a few TB of RAM and local compute will be nothing over the scope of this discussion (a few years minimum, just to build and kick off all these new fabs).

By the time these fabs are online, expect most smart phones to have 1TB of RAM and significant llm capable compute (gpu or other custom silicon). I would be astonished if flagship model phones in 2030 weren't sold with 1TB RAM. Note I'm saying flagship, there will be of course economy models as always. Certainly laptops will be sold in multi-TB RAM configs.

[ForOldHack]

So... like vertical markets? Ram for the data centers, data centers for the robots, robots for the Humanoid? or... the other way around? Robots to build data centers, and robots to build Ram.

I see... tech giants and R & D at the top of the pyramid, and I have seen Gumee, the fab that started all this Korean ram prodution.

[aenis]

How can you spend more on a still early stage tech, well in the depths of R&D?

[Schiendelman]

Humanoid robots that can do manual labor are going to be make or break for wealthy economies in the next two decades. Aging populations need help, and most successful nations do not have enough young people to do half the work they need done.

[Barrin92]

>Aging populations need help,

They're pretty good at helping themselves. Close to where I used to live in Bavaria we had a pilot project of communal living for the elderly in a community of about a hundred people that included people with quite severe conditions such as dementia.

Medical and care personell routinely checked in but they were largely self sufficient and did a remarkably job of taking care of themselves, maybe most importantly the were happy and quite dignified, something I cannot imagine is the case when your only contact is a humanoid robot.

Of course in an age where every solution is yet another technology rethinking social life isn't very high up the list.

[red75prime]

The "help" goes beyond taking care of themselves. What about food, clothing, infrastructure maintenance, and so on? An inverted population pyramid requires massive increase in the productivity of the economically active part of the population.

[mikem170]

Or a regression in the standard of living. Ideally to a comfortable sweet spot for people.

[Schiendelman]

The elderly already have a significant regression in standard of living, often catastrophically. That's a pretty big problem I would like to improve. There are so many people who don't leave their house because they struggle to get clean and they're embarrassed, and then their social connections collapse. It's going to be a long time before we have the technology to really help with elder care, but boy is it worth it.

It's also something we could help by getting rid of zoning in cities, of course, because we constantly push low income elderly out to the edges of urban areas where they can afford to live, rather than allowing supply to keep up with demand in dense, walkable, accessible places where people who can't drive can actually have quality of life.

[mikem170]

I agree, we could do so much better. Zoning can be a tremendous quality of life issue, and needing a car to participate in society is an expense other people in other places aren't burdened with.

[HerbManic]

This is the path that Japan tried to go down and it hasn't worked out yet, but we have also solved a lot more of the technical issues since they began. going to be interesting to see if we pull it off this time.

[missedthecue]

Humanoid robots barely progressed between 2000-2020. There have obviously been incremental improvements in things like dexterity, vision, self-balancing, and locomotion, but in terms of having a useful humanoid robot, Honda's ASIMO released in the year 2000 is not crazily behind what we had in 2020. So it's not surprise we haven't seen economic dividends yet in the real world.

I think AI is what could make humanoids turn from parlor tricks to huge amounts of utility, but we're really going to have to see how it plays out in the next 5-10 years.

[Schiendelman]

I think they mostly tried to go down this path before we had the transformer. With VLA models, or really now "Large Behavior Models", what's possible has changed dramatically. I've seen robot arms fold laundry now. Textile work is insanely hard, now it's just putting a lot of learned behavior together.

[numpad0]

The current humanoid hype don't have much substances or key technologies in it, and incumbent industrial robotics companies like FANUC are already in the process of rolling the techniques created for humanoids into their robots. I personally think this is going to be just series of incremental gains for big welding bots, and nursery equipment becoming mildly robotic, like Aperture Science wall panels, than humanoids walking into retirement homes and doing dishes in the future.

[imtringued]

Those were never techniques created for humanoid robots. Google researched using transformers as vision language action models to drive robots back in 2023 on a mobile manipulator and probably did non VLA work even earlier.

This is something people refuse to understand. The shape of the robot changes absolutely nothing about robot intelligence unless it abandons the basic concept of joints and links. Continuum robots are very difficult to control but they are also incredibly niche.

[numpad0]

I really can't shrug off my suspicion since one of Chinese humanoid startups had added few female type body bumps on a robot and the online humanoid hype cluster("but does it do laundry or dishes" types) responded in panic with shared sentiments of defeat, that, those humanoid hype types might be just yearning for HBO Westworld bots, to state it with modesty, not even Data or Marvin who we can share drinks with. The absolute humanoid obsession doesn't make sense to me otherwise. The cookie cutter "humanoids are best poised for leveraging pre-existing human infra" response make close to zero sense otherwise.

[bitwize]

Sounds like "That's a bold strategy, Cotton, let's see if it pays off for 'em."

[Teynah]

I agree they may be strategically necessary. I’m just less convinced that “commercial humanoids by 2028” means anything close to replacing a meaningful amount of labor by then

[Schiendelman]

Certainly. I would imagine that by the end of 2028, there are less than 1000 humanoid robots replacing a human worker. By the end of 29, maybe 5000.

[stouset]

I honestly think the opposite.

Successful nations haven’t had enough young people to care for the elderly because they’ve all been employed as knowledge workers. Knowledge work is going away rapidly. The only work that will be left are things that require actual human bodies (for now).

In short, jobs like elder care are suddenly going to have an enormous number of new job seekers, since it requires human presence and isn’t taxing physical labor.

[Schiendelman]

Are you familiar with Jevon's paradox? I think knowledge work is exploding.

[stouset]

Food has gotten progressively cheaper. At the same time, both the compensation demand for farm labor have plummeted.

Demand for the output of knowledge work can increase while demand for those who produce it falls. “Knowledge work” is not a product with inherent demand, just like you and I don’t go to the store to buy farm labor.

[Schiendelman]

I'm not sure I agree with the premise of that analogy. Demand for farm labor has increased, but more of it is done by machines. The number of people employed in agriculture has increased if you include agriculture technology, over the same period. Not that I'm not saying the proportion of all humans working in agriculture has increased, I'm saying the raw number has, which is what matters to people along their lifetime.

Also, both the pay rates for agricultural labor and the pay for developing agricultural machinery are much higher than the pay for agricultural labor was in the past.

Knowledge work is an ever expanding umbrella, just as agricultural work is.

[summerlight]

The title is kinda misleading. The actual wording is not specific about humanoid robots but "physical AI" which encompasses every machinery that can be potentially integrated with AI, especially focused on mass manufacturing for the Korean case. Basically this project is about all physical infrastructures to automate high tech manufacturing industry.

[chaostheory]

> Humanoid robots are like dance lessons, it is cool, it is sexy, and it may pay off in the future, but the value is much less certain

They need a solution to their plummeting birthrates which are officially worse than either China’s or Japan’s

[Teynah]

I think there's a reason they're being lumped together. Memory is the reliable industrial bet. Humanoids are the speculative "what will all this AI hardware actually enable?" bet

[dist-epoch]

https://en.wikipedia.org/wiki/Commodity:

> The wide availability of commodities typically leads to smaller profit margins and diminishes the importance of factors (such as brand name) other than price.

I'm not aware of many commodities which have only 3 world-wide sellers.

[taneq]

I’d say it’s more like “on groceries and a fancy dinner”. Humanoid robots sure do need RAM, both in data centres for training and in the robots themselves. :)

[kijin]

Vertical integration. Produce the chips, build data centers to run LLMs on the chips, and the robots to deliver the result to end users.

[fragmede]

The value is pretty clear. The problem is the pay off is uncertain.

[ekianjo]

we will need humanoid robots since nobody makes kids anymore