▲ 259 points
back
192 comments
One quote that really stood out to me:
> Rodney: The robots—they’re not embodied. I always say about a physical robot, the physical appearance makes a promise about what it can do. The Roomba was this little disc on the floor. It didn’t promise much—you saw it and thought, that’s not going to clean the windows. But you can imagine it cleaning the floor. But the human form sort of promises it can do anything a human can. And that’s why it’s so attractive to people—it’s selling a promise that is amazing.
> ... I always say about [a language model], the [linguistic] appearance makes a promise about what it can do. [Clippy] was this little [cartoon paper clip]. It didn’t promise much—you saw it and thought, that’s not going to [write the next great novel]. But you can imagine it [offering limited help]. But [human language] sort of promises it can [write] anything a human can. And that’s why it’s so attractive to people—it’s selling a promise that is amazing.
All these companies are pursuing humanoids precisely because we've built the world around that "form factor" (and we evolved ourselves to fit to the world). It's a general purpose design. It's the same reason why OpenAI pursued LLMs, as they are general purpose.
Yes, like LLMs, they will over-promise in the beginning. But it still make sense to pursue that form factor from an investment perspective if we think it's feasible.
~"Brooks believes the sheer size of Nigeria is going to make it an economic and technological epicenter".
I laughed when i read this. The sheer ignorance and false assumption that large population = epic economic powerhouse is ridiculous. I critize CCP a lot for various reasons but their effective and efficient governance over the last 40 years can not be easily replicated. Good governance is an incredibly rare asset. India has big pupulation but has no good governance. The chance of Nigeria having as good goverance as China is very close to zero.
> India has big population but has no good governance.
As an Indian, I would dispute that. When I look back at what we were as a country, when we became independent, to what we are today, our achievements are really astounding.
It's worth noting that China is investing heavily in Nigeria and it is also its biggest lender, it might turn out Nigeria's gov is not that far from CCP.
To be fair China has a top down control that India and the like don't
I agree about Nigeria but not about India. My understanding is that they have some very able people in government.
There are large countries in Africa that are undergoing fast development though, Kenya for example.
Why is it very close to zero? Nigerians in the United States are the most educated immigrant group in the country.
This reads as rather naive. You think India’s governance is “no good”? It is literally the most culturally complex place on earth, so you might need to recalibrate your metrics and criteria.
Additionally, if you believe Nigeria has a near zero chance at good governance, then I suspect you have not read very much history or anthropology. Every instance of civilizational development looks pretty much impossible, and yet “good governance” emerges over and over again. The story of humanity is one of surprising social innovation, and I would never count Nigeria out.
Well since you, of all people, find it hilarious, and honored us with your iron clad, deeply reasoned opinion, it must be immutable truism
Time is not a linear function; creation and decay are not a normalized process. A large number of people can change inductive truism.
This is roughly the same content Brooks posted himself a few years ago, and was covered on HN last week.[1]
There are already many companies selling automated carts that run around warehouses. Search for "automated guided cart".
I thought the humanoid thing was silly until I saw the pricing. Here's the Unitree G1, starting out at US$22,000, less than a Toyota Corolla. I though these things were going to cost like Boston Dynamics products. No, the hardware is already much cheaper.
This is still a low-volume product and prices are headed downward. Humanoid robots are going to be cheaper than cars. Having more degrees of freedom than you really need for any single job will be outweighed by the cost advantages of mass production and the advantages of interchangeability.
The manipulation problem remains tough, but with moderately priced and standardized hardware available, more people can work on it.
> Here's the Unitree G1, starting out at US$22,000, less than a Toyota Corolla.
The sticker price might look cheap, but what I've heard from people actually looking to buy (even in large quantities) is that by the time you assemble the hardware and the tooling and the dev kit, the cost is $80-100k per robot.
> This is roughly the same content Brooks posted himself a few years ago, and was covered on HN last week.
That Brooks' post is his latest one, from a few days ago.
nobody has solved manipulation at any price point, you can't scale your way out of that
It also tells a story about any US robotics company like Tesla now claims to be. How can you possibly hope to compete against manufacturing costs that low.
I really liked this guy. Maybe it's Boston but here's a guy with two wildly successful startups in robotics and he's having trouble raising money because his idea isn't sexy enough!
I thought investors were smart and knew how to calculate odds. There is never any absolutely sure thing. But if a guy has been wildly successful twice aren't the odds pretty good he can do it a third time?
Having been through the VC system, I don't believe his face value statement, and I 100% believe you're right.
They want to invest, but not at his terms that he's demanding.
The thought experiment is - he wants to raise 1m @ 2m val. 50%. There's not a VC/angel/seed who wouldn't take that deal given who he is. So he's pricing himself at a level he "is willing to do this startup thing again" and it's simply too expensive.
I know a very low profile entrepreneur who has essentially built the exact same business (with slight variations) 3 times over the past 20+ years, selling the first two, so yes I believe there is some corelation between his 2 previous successes and being able to do it again. Even if it's imaginary the funding markets swear by previous success being a strong marker for future success.
iRobot got crushed by Chinese competitors. Rethink failed early, they had a poor quality product and Universal made a much better cobot. And this new venture is pointless, there are 10 other companies in the Boston area alone doing warehouse automation...
I mean, VCs want stories (or at least hype) - not accolades, i.e., Altman, Neumann. Heck, accolades might hurt you. They want the unfound gem/they're so smart/cerebral...
Define success. Without a doubt iRobot was a category leader for a long time. They created the market but were they ever profitable? At this point it feels like a company on life support that is getting their butts kicked by the Chinese who offer double the product for half the cost. Without a doubt though he created the category. Was the second company ever a success, all I could find is it being trade for parts over and over.
This third company looks interesting but it’s also a flooded market at this point.
So for me the guy never had a real success and is currently building in a market that has been for years flooded with products like his.
AGI doesn’t need to be “solved” for humanoid robots to be valuable at scale. The role of teleoperation is often underestimated; in the near term, many humanoids will likely be operated remotely by people halfway across the world, performing deliveries and other tasks cheaply.
Is that supposed to be something we want?
How does that increase efficiency (beyond not having to pay normal wages)
> It’s simple intelligence, which is what we can do today and make reliable. It’s not sexy. It’s technology in the service of making things easier for workers and more efficient.
Perfection.
>At MIT, I taught big classes with lots of students, so maybe that helped. I came here in an Uber this morning and asked the guy what street we were on. He had no clue. He said, “I just follow it.” (‘It’ being the GPS—Ed.) And that’s the issue—there’s human intervention, but people can’t figure out how to help when things go wrong.
I live on a culdesac, and theres no instructions I could give uber drivers to help them find my house, when uber added a random roadblock on our street.
I tried guiding them by voice, but none of them read street signs. Its crazy. They just rock up to the dot and complain. Issue being, the dot was the closest accessible street, so one street in either direction.
One guy made the same wrong turn twice before cancelling.
I hate taxis, but at least taxi drivers can be expected to have some basic local knowledge.
You need to work out what the nearest routable adress is. This works wonderfully with taxis here.
FWIW cul de sacs are an evil invention, and what you describe is a common routing problem. The rescue services here solved it by mandating addresses with geolocation and the nearest streets for every house around here. That failed me because everyone uses that official data that is convenient for fire truck rescue.
They're just a tired company not trying to innovate at all. My eufy is better at a fraction of the cost. How many versions of a robot vacuum can you make...
Going back to the tired quote about AI doing the dishes and laundry, fundamentally what IS so hard about making a robot that does that?
Is it the range of motion of my arm and hands? Is it computer vision not knowing the clearances of objects ? Even if the thing couldn't drive itself to the kitchen or laundry room, I would be fine wheeling it there and standing it in front of dirty clothes or dishes opening the appliance doors. But yet we don't have this.
Each individual problem is solved: we have water proof electronics, we have precise multi axis stepper motors, we have computer vision that can map 3D spaces. We could even sell custom dishwasher inserts that are designed to make it easier for the robot. Why the hell do I still have to do the dishes?
Reliable, general-purpose gripping is still an unsolved problem. Especially in the consumer space. Your body has the near-miraculous ability of repairing itself continuously. Most gripper technologies require regular maintenance or they start to fail: suction cups wear out, sandbags start to lose elasticity, rubber gripping points get worn smooth. In an industrial setting, this is something that's manageable with maintenance schedules and the like, but will consumers be happy to replace suction cups every month so their robot doesn't start dropping dishes on the floor?
The dexterity, ultra precise motor control, and senses of the human hands are the bottleneck. That’s my guess at least.
Seems like a fundamentally hard engineering problem. Evolution is mighty impressive!
Things like simple robot arms could do wonders in crucial sectors like construction and agriculture where most the world does not have the industrial solutions of the US. Depopulation and aging is causing massive abandonment of agriculture in europe for example. Indeed we don't need a humanoid to lay bricks, but a not-too-expensive device could help with the housing crisis.
> You can do so much more computation, sensing, some actuation, but people underestimate the long tail of the natural environment.
Indeed, there are so many different application requirements that a single generic platform will never be possible. And humans are inefficient, so why do humanoids? We need a modular robotics platform backed by a big player/operating system
Having interviewed at a few industrial robotics startups, I am concerned a bit for the short-term future of industrial robotics. Lots of companies by founders who have no real robotics experience, who seem to have lots of funding with no real customer base.
Not to long ago I interviewed with an embodied AI company who's founders were from the Google DNS team and had no real robotics experience. They were obsessed with graph theory which, while cool, is solved problem in the context of robotics and their startup. Another pitched itself as advanced AI but was just a camera and YOLO object detection. I worry there is going to be a harsh market readjustment.
> We have affordances on the cart that lower the cognitive load. [...]
In comparison, the state of the art is that people have scan guns, and on their wrists are tiny screens with character-based software—it’s ’80s or ’90s technology emulated on an Android device. They have to read that to know what bin number, what thing to do.
Which means the literate person will get fired and replaced with an illiterate, because all the robot needs is a compliant gripping tool.
"It’s much easier to fund the promise than a real business, because real businesses have limitations on how fast they can grow. Whereas if you don’t know, you can live (and fund) the dream."
Says the guy whose robot vacuums only now don't drag dog shit across your house in a robotic square dance purely because they now have a camera that uses AI to detect such things.
1. When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong.
"Rodney: That's why I'm skeptical of the Tesla taxi system. At the last earnings call, Elon said they're going to have safety drivers in the Teslas and they're hiring remote drivers. It's sort of a charade.
Rodney: I model myself as a realist. I've lived through so many hype cycles in AI. They weren't as big in public as this one, but they were brutal amongst AI practitioners. The arguments were strong and deeply held-screaming matches would happen. I've seen that happen again and again. Neural is ascendant at the moment, but neural was ascendant four or five times before and then got crushed. Something else took over, came back.
You can see that in agentic AI. [OpenAI promotion of "agentic AI" is currently on HN front page, directly adjacent to this submission.^1] Now suddenly everyone's got agent-based AI. They didn't have it six months ago. I suspect it's a little more marketing than reality. But when was the first paper on agentic AI published? It was in 1959 by Oliver Selfridge. There's been agent-based systems-SOAR, there's been lots. They come and go, all these ideas, and they get improved every time they come back. I'm not saying it's stupid, I'm just saying as someone who's been involved, it is not just the shiny new thing. This thing that looks shiny now may not be so shiny in a few years."
I really enjoyed this interview.
I didn't know that much about Dr. Brooks. Seems like a really practical, capable chap.
He also looks a lot like Asimov (at least, in that photo).
How many years were they on the market before they added a dust buster to the dock? 3-4?
Aren't iRobots really not that clever and maybe can be improved with LLMs?