Irobot Founder: Don't Believe the AI and Robotics Hype
Key topics
iRobot founder Rodney Brooks cautions against the hype surrounding AI and robotics, emphasizing the importance of practical, simple intelligence in robotics, sparking a discussion on the limitations and potential of current robotics technology.
Snapshot generated from the HN discussion
Discussion Activity
Very active discussionFirst comment
3h
Peak period
90
12-24h
Avg / period
22.9
Based on 160 loaded comments
Key moments
- 01Story posted
Sep 29, 2025 at 4:19 PM EDT
3 months ago
Step 01 - 02First comment
Sep 29, 2025 at 7:39 PM EDT
3h after posting
Step 02 - 03Peak activity
90 comments in 12-24h
Hottest window of the conversation
Step 03 - 04Latest activity
Oct 4, 2025 at 5:55 AM EDT
3 months ago
Step 04
Generating AI Summary...
Analyzing up to 500 comments to identify key contributors and discussion patterns
Want the full context?
Jump to the original sources
Read the primary article or dive into the live Hacker News thread when you're ready.
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?
The Rethink hype video, with their cute little face on a screen.[1]
What the Rethink robot could actually do: take PC boards off a conveyor and put them in a fixture.[2] That's a routine robotic load/unload task. Like this one with Fanuc robots.[3]
[1] https://www.youtube.com/watch?v=S4mULTknb2I
[2] https://www.youtube.com/watch?v=nnzmxJS4Rp4
[3] https://www.youtube.com/watch?v=Za5z1lb0hdU
The conclusion to draw from "tried and failed" certainly isn't "it's impossible and everyone should give up on the notion".
China competitors are always going to be tough to beat in terms of margins.
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.
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.
His other business was a failure and his third current is in a crowded marketplace. Humanoids are the minority in warehouse automation.
Who is even talking about his argument on humanoids? What does that have to do with my comment. My response was on a comment praising his triple success in business and I am questioning that definition of success.
Are you aware that from an investor perspective, creating a household brand and high IRR though IPO is considered a wild success? All companies eventually decline. By your metrics GE wasn't a wildly successful company either. Or AOL. Or Yahoo. Or Myspace. Every one of those companies eventually declined and your return would be poor if you held on to your investment over their entire arc.
I am not attacking the man only a perspective that I am not sure investors would see him as a wildly successful entrepreneur.
iRobot has sold over fifty million Roombas to date. Packbots are also highly profitable and have saved many lives.
Again as I keep saying. My reply was to someone who said he had three wildly successful companies. Yes iRobot created a segment but I still struggle to classify it as a wild success.
https://macrotrends.net/stocks/charts/IRBT/irobot/gross-prof...
From that, they had 13 straight years of profitable net income starting in 2009. Rough recently because they got eaten up by cheaper upstarts; happens to almost everyone eventually. But clearing several hundred million cumulative in net income over a decade+ (peaking 2018-2020, so they were consistently growing from 2009 to 2020 even) is far from "were they ever profitable" territory. They don't have pre-2009 there, but from some other googling it looks like they were profitable by when he left in 2008 after their 2005 IPO. So he was there from inception to IPO and millions-a-year-profitability, and then you're discounting that because 12 years after he left some companies in China had better copies. Building something for 15 years to IPO isn't some VC pump and dump scam, it's harder.
If that's not success you have a ridiculous bar. "So for me the guy never had a real success" - get out of here, do your googling before making claims like that.
Did shareholders and investors collectively make money from the whole ordeal? Looking briefly at the numbers, it looks like they didn't. During the profitable years IRobot made $639 million (sum total) but they lost -$737 in the collapse that followed. No dividends were paid either during the good years. Shareholders were left holding the bag.
Building a business from nothing to IPO is a real accomplishment and I won't diminish that. However, if a business collapses and incinerates more money than it has ever made during the sum of all profitable years calling it a "real success" is a bit of a stretch.
> 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.
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.
Space is practically inaccessible to most people.
It's crazy how many people purposefully become ignorant of the things they see every day with their own eyes.
We've had buildings with flat surfaces for centuries, so I'm not sure of your point there. And Sears was shipping things in cardboard boxes long before warehouses had robots with suction cups.
Back on the ground, I guess it's possible that some general purpose non-humanoid robot will be able to use our tools and built environment better than people, but I have a hard time imagining what that would be.
Just look at the animal kingdom, there are a bunch of stuff that would be extremely useful, tails, extra limbs, extra joints, no separated head, longer limbs &c.
Some of your listed features could still work though. A humanoid robot doesn't have to have exactly the same form as humans, as long as it's close enough to use our stuff, ride around in our vehicles, etc. The alternative is to build all new infrastructure and tools that don't work for humans; I'm not sure why we would want that.
The point is I don't need a general purpose robot of any form. I have various specific tasks I need done, I don't care how many robots are involved. I do care that the total is affordable. I do care that the total doesn't take up "too much" space. I do care that the job is done well. I don't want to think about any of it unless I want to think about it (sometimes it is fun to watch my robot vacuum work. Sometimes I want a specific mess cleaned up first and then go away before guests arrive). Maybe a general purpose robot is the best answer, maybe not - since I'm not [currently] designing such robots I expect those who are experts to figure out the best form - which is probably a compromise.
If you have a factory, idle equipment costs you money. Right now our factories have only specialized robots, but they still have lots of people because it's too expensive to do the remaining tasks with specialized robots. We need general-purpose workers for the tasks we're not doing 24 hours every day.
This makes general-purpose robots look like the cheapest option for many tasks, but that could be offset if they cost a lot more. But the more of anything we build, the cheaper the thing ends up being. A very large number of identical robots will probably end up costing less than smaller numbers of lots of special purpose robots.
True, but that isn't the whole story. Where I live most factories have their own snow removal equipment even though it is only used a few days per year and sits idle the rest of the time. While it costs money to have it sit idle, it costs even more to have the whole factory idle until the snow melts. (or the people you hire get around to your factory).
> Right now our factories have only specialized robots, but they still have lots of people because it's too expensive to do the remaining tasks with specialized robots.
And when they decide people are too expensive they replace them not with general purpose robots but more special purpose robots.
> A very large number of identical robots will probably end up costing less than smaller numbers of lots of special purpose robots.
Maybe. It isn't clear. A general purpose robot could be more expensive - I only need one while I "need" one vacuum robot per floor meaning that special purpose robot scales better. And that vacuum robot is also a lot simpler meaning it will be cheaper to own 2 than to have 1 general purpose robot.
The question then is the general purpose robot cheaper because you only have 1 instead of an army of special purpose robots. We do not know (today).
An urbanized, general purpose robot may take an entirely different form factor than bipedal homosapien, ie. tachikomas from Ghost in the Shell: SAC.
A true "general purpose" robot should be able to walk into a space built by humans for humans and perform useful work there. That's the reason why humanoid frames are desirable.
Chicken and egg problem there. Nobody will adapt the environment for your robots unless your robots have proven to work really well. So your robots have to first work really well in unadapted environments.
There are always first adopters who will if it isn't too expensive. Rich people often live in a house with big enough doors (well most doors big enough - enough for this discussion to say they do), and likely have a need (and the money) for something even if it isn't very good. If your robot starts to prove useful people will ask for it - we build thousands of new houses/apartments every year just in the US, if builders see a demand they will make changes to their new models - it doesn't cost that much more so they don't even need a real demand, just a marketing feature that you could even if nobody does will work (for a few years: you better prove it useful to those who buy new houses fast or the fad will pass!)
It isn't easy because as you say, chicken and egg. If these prove useful people will make changes of the next few decades so they can have one. (though of course your competition will be looking to see if they can make something that doesn't need a remodel to use)
Humans evolved to be bipedal to see over the bushes and tall grass in the savannah. Maybe the optimal form for a robot is an elephant with prehensile trunk, or giraffe, or a frog.
When a humanoid robot fumbles, its mistakes are obvious because the physical world offers immediate feedback.
It's the difference between lying on your résumé that you're a world-class gymnast, and having to actually perform.
So the robot might be equally convincing that it is capable to clean your windows as it is capable to repair your car brakes.
You saw it clean your windows and are satisfied, and both its form and words are promising that it can repair your brakes equally well...
I’m kinda torn between “genAI powered robots will have a ground truth reality as a reference, so they will ultimately be more grounded and effective that LLMs” and “LLMs are like drunk uncle Steve with his PHD swimming in vodka, and using genAI in robots will end up as well as having drunk uncle Steve drive home”.
Guardrails on tasks it will attempt are inevitable, but I can also see that becoming a paywalled enshitification farm.
With the gymnast example, as a non-gymnast, I don’t know the difference between a high and low scoring routine on the floor or beam. If a humanoid robot did a routine and didn’t fall, I would assume all is well. I don’t know the technical details of what is required for a gymnastics competition.
This seems like the same idea as an LLM writing a paper that looks correct to someone who doesn’t already know the answer.
In a home context, this could look like the robot not practicing proper food safety or storage around someone who doesn’t know the details about that kind of thing, which is a good number of people. What it’s doing might look correct enough, and it produces food you can eat… all is well, until you get sick and don’t know why.
I’m not an expert, but I know there are specific moves with various degrees of difficulty. I believe there is a max score based on that difficulty level, and any imperfection will lower that score, such as a foot pointed or flexed the wrong way at the wrong time, taking an extra step on a landing, etc.
I know all these rules exist, but I’m not an expert where I can say someone had their foot flexed when it should have been pointed. These details would go over my head, where a humanoid robot might get a pass from me, while an actual gymnast or judge would be able to see faults.
Welcome to the world of hard tech not easy machine learning models. Capital is in short supply, it doesn't go nearly as far and you don't get wild multiples in return if you even get any.
If you just ask an LLM to write something off the cuff, it'll be bad. But doing a lot of prep with a human author guiding it? Not Dostoyevsky level, but not pure slop.
While I have seen LLMs produce some ham-fisted attempts at manipulating the state of mind of the reader, I think that the human process is so obfuscated that it only shows up in occasional echoes and shadows in the training set.
It might be possible to develop a training set that reflected perception and internal mental state vs input using (magic brain scan technology) that could change this, but right now the emotional state of the reader is just missing from the training data.
"It's writes like us, it must think like us, and will be able to think anything we can think!"
"It's embodied like us, it must be be like us, and will be able to do anything we can do!"
Flawed thinking layered upon flawed perceptions, but get enough decision makers to buy into it and heaven and earth are moved to further it.
The "API" of trainable algorithms is essentially "arbitrary bunch of data in -> thing I want out" and the magic algorithm in the middle will figure out the rest.
Because "thing I want" is given as a list of examples, you're not even required to come up with a clear definition of what it exactly is that you want. In fact, it's a major "selling point" of the field that you don't have to.
But all of that creates the illusion that machine learning / "AI" would be able to generate a robust algorithm for any correspondence, as long as you can package up a trainset with enough examples and shore up enough "compute" to do the number crunching. Predict intelligence from passport photos? Or chances of campaign success from political speeches? No problem! Economic outlook from tea leaves? Sure thing!
The setup will not even tell you if your idea just needs more tweaks or fundamentally can't work. In both cases, all you get is a less-than-ideal number in your chosen evaluation metric.
I think it is possible to avoid, though, by asking if humans can be generally good at the task in question, if working through the implied interface restrictions, and then evaluating whether the required skills can be reflected in an available training data set.
If either of those cannot be definitively answered, it’s probably not going to work.
An interesting example here is the failure of self driving vehicles based on image sensors.
My take is that most of the problems are because a significant fraction of the actual required training data is poorly represented in data that can be collected from driving experiences alone.
As in: If you want a car to be able to drive safely around humans, you need to understand a lot about what humans do and think about. - then apply that same requirement to everything else that occasionally appears in the operational environment.
To understand some traffic management strategies expressed in infrastructure, you’ll need to understand, to some degree, the goals of the traffic management strategy, aka “what were they thinking when they made this intersection?”.
It’s not all stuff you can magically gather from dashcams.
A model might potentially be able to understand those situations, but it would need a lot of highly task specific training data and it would never be clear if the training really covered all possible situations.
The other problem I see is that a lot of situations in traffic are really two-way communication, even if it is nonverbal and sometimes so implicit we don't realize it. But usually pedestrians will also try to infer what the driver is thinking whether he saw them, etc. In those situations, a self-driving car is simply a fundamentally different kind of traffic participant and pedestrians will interact with it differently than they would with a normal car. That problem is independent of machine learning and seems much harder to solve to me.
The sort of questions you're talking about are primarily popular in academia. Run some MLRs against some random dataset you found, publish a paper, maybe do a press release and sell a story to some gullible journalist. It doesn't have huge value. But generative AI isn't like that.
It plays off of the "if it looks like a duck, quacks like a duck, and walks like a duck" idiom, which of course isn't foolproof and gives avenue to the kind of spectacular advertising that is fueling this hype.
My mom was lamenting car insurance quotes, so I told her to ask AI. She did, then had it do a Monte Carlo simulation for all the insurances she the AI felt she was qualified for.
It happily replied that it did 1 million monte carlo simulations and here was the result.
To this day I don't think she fundamentally groks that LLMs cannot calculate.
https://rotalabutterfly.com/rex-grigg/dosing.htm
Can't most LLMs trivially use Python and other languages and libs and calculate?
They're getting quite good at that now.
At first I've liked it but the bruise rotor is already making rattling noises. It advertised itself as Matter compatible but I had to buy a hub thing to integrate it in HomeKit.
Worst thing it relies on Chinese servers for its metrics. I got China blocked on my router due to reasons so I can't check the status of the parts without leaving my wifi. I already bought a box of 1 year replacement parts for this but I'm getting rid of it as soon as I can.
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.
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.
The work around in the interim was a combination of:
1. Writing a long description of how to access our property. 2. Trying to limit the distance an uber driver had to travel. <1km made it more likely for them to review the map and comment before driving. Dont ask me why. 3. Oscillating the drop off marker on the map to make it more likely that the driver was sent to our street at all, where we could wave at them to get them to drive beyond the psychological roadblock. 4. Attrition lead some drivers to remember my name, and just come straight to the property ignoring the map.
Any other mapping service delivered people to our property fine. Its just that Uber randomly blocked the end of our street. Meaning our street had no entrance.
Problem was that, uber has an exclusive relationship for delivery with one of our local supermarkets, and my wife doesnt drive and was stuck at home with an angry baby for the duration.
Due to this issue we did push the maximum amount of our orders through other delivery apps, probably taking significant income away from uber. But they have deliberately structured their support channels to be impossible to deal with.
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.
[1] https://news.ycombinator.com/item?id=45392922
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.
Thus my question - what scale are they targeting. At some level of sales their expected costs are right. However if demand is a lot more than they expect it could probably scale down to $10k, maybe less, while if it is less costs go up...
- Ford F-150 base price: $39,645.
- Ford F-150 Raptor R: $115,015.
It's extremely common for dev tooling to be extremely expensive, even for widgets that sell for a few bucks at the retail level.
I know. You don't get the GPU for AI in the base model, for example. Do you get the articulated hands?
A robot package with all those custom parts, motors, and motor controllers, along with cameras and a short range LIDAR. for only $22K is pretty good.[1] As this technology starts to work, it's apparently not going to be insanely expensive.
[1] https://robostore.com/products/unitree-g1-robotic-humanoid
It can lift 3 kg / arm. That's like barely enough for a super-sized coffee! What human role are you replacing with this? Some desk job? Why bother with the robot.
His point was that mass production will make it more likely that someone will solve it
(Wait this is still the rodney brooks discussion!?)
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.
Don't. It's good for us that they overestimate China and underestimate us. For the next couple decades, we need breathing room to develop; why would you destroy that space by drawing attention to yourself?
Everybody and everything is growing, even North Korea in some aspects is. Thats called a progress and nobody interacting with rest of the world can escape that.
What is unfortunately true about India, and many other similar places is the potential of the country and the people and reality where they got and how far they can reach, given their own limitations.
From what I've seen, the gap (potential vs reality) is huge compared to western standards, and I don't see an easy way to push through that ceiling. Society isn't fair, its everybody for themselves. Castes are very much a thing still, absolutely horrible and deplorable thing (while this will be read mostly by brahmins or rich merchant castes who rarely see an issue there, so I don't expect much sympathy here). Corruption is abysmal. Women and minority rights are questionable at best.
No doubt great and important things will improve in upcoming decades. But in 21st century India will not resemble ie western Europe in any meaningful way, I just don't see a way. Real India is not some Bangalore or Pune bubble, its very opposite of it.
And China... its doing its own superpower things. I get why US is scared of prospect of having an actually worthy peer for a change with very different values, but thats not a main concern for Europe. putin and his FUBAR russia is and will be, all focus and energy goes there as it should. Unless US is going to properly help with that instead of licking various orifices of your own sworn enemy while it spits on you, we leave China to you guys (which just wants to sell their cheap EVs and other stuff here, hardly an evil plan we should lose sleep about right now).
In Europe you very rarely see consumer goods labelled "Made in USA" (although we will frequently see products that are "Made in India"), yet the US is still an economic powerhouse because its strengths are in services, tech, finance, aerospace, and defence rather than mass-market manufacturing.
India is similar with most of its global impact through IT, software, generic pharmaceuticals, and space tech. Its the 4th largest economy in the world, and among the fastest growing - potentially surpassing Germany for 3rd place by 2030.
India is also the 2nd largest exporter of food in the world, and among the largest textile exporters (clothing, etc) - but for some reason you wanted to exclude commodities.
I'm hopeful for India's future and they're on the right track, but currently, they're punching well below their weight. Germany is punching well above theirs. But investment from companies like Apple shows that India has what it needs to become the next China (or at the very least, next Vietnam). Nigeria seems to lack similar investment.
Let's see India's GDP per capita.
> India is the world's largest provider of generic medicines by volume, with a 20% share of total global pharmaceutical exports. It is also the largest vaccine supplier in the world by volume, accounting for more than 60% of all vaccines manufactured in the world.
https://en.m.wikipedia.org/wiki/Pharmaceutical_industry_in_I...
This I am less certain on, but I believe there is a fairly robust space program (not many countries have that). I’m sure there is more.
It's not comparable to the broad based productivity gains which happened in China or the UK across every industry as they developed, from agriculture to manufacturing to services. Instead it's much more like the Norwegian or Saudi hydrocarbon economies but with human capital exports.
That's simply not true.
India is making steel, pharmaceuticals, textiles, cars, movies, you name it.
https://en.m.wikipedia.org/wiki/Economy_of_Saudi_Arabia
BTW, while I would call the US or EU better, both have very significant governance issues that they need to work on. This shouldn't be seen as a game of who is better it should be seen as a game of how can I improve, looking at others only to see what areas they are better you can copy, and what areas they are worse you can avoid.
Since I live in the US that is where I focus most of my efforts, and I'm constantly stopped by people who have their heart in the right place but are doing something that because of complex factors they refuse to acknowledge is worse than what we have now. I'm also constantly being informed of something I didn't think of and having to readjust what I advocate for. If you are not in the same situation you are making things worse!
I grew up in the USA and the changes in India even in the last 10 years has been completely astounding. For example I’m from Hyderabad and some of things changes I see: A national freeway that didn’t exist 10 years ago, a metro that covers most of the core of the city, skyscrapers, trains that now ship containers instead of boxcars, massive power plants that have recently been built.
Is India perfect? No. It still has a hierarchical society, sectarianism is still pretty acute, like everywhere else in the world India’s fertility rate has dropped off a cliff, and traffic in Hyderabad proper is a nightmare.
But culturally the submissive Indian mindset of my youth is largely gone. The millennial India is entrepreneurial and everyone is looking to get ahead. The next 20 years of India will be amazing at which point a lot of the issues of low fertility rates will start affecting India as well. Lastly, I think India really only has capacity to send a sizable amount of its citizens overseas only for the next 5 years or so at which point migration out of India will largely subside.
It's amazing when people flag this as a bad thing when it's undoubtedly a key component of getting places to prosperity in the first place. Got to get people away from being starvation-limited.
> I think India really only has capacity to send a sizable amount of its citizens overseas only for the next 5 years or so at which point migration out of India will largely subside.
That's how you can tell a country is ""winning"" in the international rankings, when more people want to move in than move out.
Exponentially falling fertility rates can create dynamics which can be destructive in its own right. As with other complex phenomena it would be for example foolish to rapidly cool the earth's climate. Stability is the key, here. Right now India is just below replacement which short to mid-term looks very promising but will it stabilize? Looking at worldwide trends I very much doubt that. A growing economy needs some demographical stability so coming from a long-term view fertility dropping off a cliff, now, could be bad news later (in one, two generations).
Turning some knobs one way or the other does not produce linear results, quite the opposite, there are thresholds, there is criticality. To draw on another more time compressed analogy here: I guess some operators thought back then: What could go possibly wrong by running a nuclear reactor (RBMK) at safer lower powers?
The baby boom and a couple previous generations in the US is also associated with getting to high prosperity and high fertility. This part of your thesis is debunked and thus your whole argument falls apart.
There are large countries in Africa that are undergoing fast development though, Kenya for example.
Time is not a linear function; creation and decay are not a normalized process. A large number of people can change inductive truism.
32 more comments available on Hacker News