Congratulations on the launch! This is really cool.
I'm not super active in the humanoid robot space anymore, however I did my PhD about 9 years ago in HRI.
That was the time of Boston Dynamics, DARPA robotics challenge, and Aldebaran's Pepper and Nao robots.
You mentioned you are building everything open source. What happened with ROS and related projects? Do you build on top of that, or is that all super outdated that a reboot was needed?
Another question I have is: why are you choosing a two-legged human over a four-legged one?
My experiments with two legged robots were mostly bad. Not only did they fall basically all the time but they also had a big drift. So far, I have not seen any large improvements. But again, I might be very outdated.
I always said to my colleagues. The main point stopping robots from picking up is a stable platform. And with the platform I mean walking.
Eh. I think we got a bit nerd-sniped on some things and we ended up trying to build most of our stack ourselves. The control loop is just a pretty simple Rust loop for running ML models. ROS is kind of annoying to work with and the control loop is pretty simple so we just wrote it ourselves.
For two legged - I think it will eventually be quite a bit cheaper, it's mostly a software problem to get them to be stable. RL based control has gotten much, much better. For example, I was talking to Trevor Blackwell a few weeks ago, and he was saying the IMU on the original Anybots robot was over $2k, whereas if you throw a noisy IMU into sim, you can get away with something basically from a cellphone. So yea, personally I'm a big believer in needing to solve the robotics intelligence problem, and once you solve that, humanoids will make the most sense as a form factor.
Human feet have metatarsophalangeal joints connecting the toes to the rest of the foot. But humanoid robots don't have these (at least, the vast majority don't). Why? These joints are very useful.
Also, the bottom of the human foot is soft and has thousands of nerve endings. Can we really expect robots to get anywhere near human mobility performance without this level of compliance and sensory sophistication?
Feet/ankles on humanoids are really hard mechanically. In many ways the kinematic requirements for the ankle are similar to a wrist, but while the wrist of a robot arm is the least-heavily-loaded, the ankle area can be the most heavily loaded. Humans get away with it by having most of the highly-stressed actuators in the lower leg, not the ankle itself, whereas many humanoid robots end up putting more of the actuators in the ankle assembly itself.
In general, I think the best way to think about the differences between human muscles and robot actuators is that human muscles are simultaneously incredible in terms of strength and power density, and also incredibly fragile. Robot actuators are fairly robust, but comparatively poor. Humans are essentially falling apart at all times, but the repair mechanisms in the body do a good enough job that it doesn't matter (although you probably know someone with a "career-disruptive/ending" sports-related injury that shows these mechanisms have limits). Robotics is a long way away from making actuators that can be fixed online in such a process. Even cable stretching in cable-driven mechanisms remains hard to handle effectively, and that's one of the simplest types of mechanism wear.
So, most humanoids you see that are relatively cost-effective are just "humanoid" in that they look like humans, but there are significant mechanical differences between them and real humans. It's almost always driven by the cost of manufacturing different components. A good example is the lead screw you see in the knee and ankle on Optimus - while it is more human-centric to have tendon-like actuation, it drives the price up quite a bit. Put differently, humans have a lot of specialization which is not great if you want to mass manufacture something.
For walking, the most important thing is that the robot can be simulated well, so in our case, we tried to model the foot contact with the ground in simulation quite accurately. In fact, we found that force sensors in the foot probably help but they're not necessary in simulation, and we wanted to shave off anything that wasn't necessary. I am not sure how close we will get to human mobility - it's definitely a learning process - but you can get much further in simulation than you'd expect.
The stuff of nightmares is this being adapted by the DoD. I can almost imagine your website as a scene in the prologue of a terminator like movie.
Nightmare 2 is this becomes a companion of some sort. Detroit Become Human goes into this. You have a theme of the robots basically wanting freedom. Which throws out a moral conundrum, if someone buys an AGI enabled bot just to be mean to it, have they done anything wrong.
I like technology , but this feels like step one to a whole lot of weird stuff.
I suspect that these kinds of things will happen with or without my involvement. Assuming that they do, I would rather that a good open source option exists
What ML algorithms do you intend for full autonomy? Multi Modal LLMs for planning that control the robot by generating s.th. like code? Or s.th. that requires more learning from the environment?
When I click "get in touch" on your github I just land on the website where I can buy the robot. Building the hardware for an autonomous robot is orders of magnitudes easier than the control. Do you think anyone with the capability do develop an autonomus robot will buy this and then just give you the code because its open source?
My overall plan is basic joystick control -> VLA with RL -> self-supervised embodied representation -> end-to-end RL -> end-to-end control. I suspect there will be some very good multi modal models coming out in the next few years which we might use as base models, although more likely, we will adapt their techniques to work on data from our own robot.
I agree that the hardware is easier than the software - I am a software guy, personally, but I felt that it was important to do the hardware first so at least we can have a baseline product which we can offer to people. I would personally like to work on this software problem (or rather, build a company to work on this problem), and this seems like the right way to go about funding working on this problem.
I like the K-Sim Gym. Im looking forward to fiddle with it a bit when I have more time. I could see that you get something usefull out of people competing on your leaderboard xD
It's my hot take that the next big ML breakthrough needs s.th. that learns from its own actions in an environment, so this goes in the right direction imo.
On the other hand a lot of big companies struggle with self driving cars even though they predicted to build this years ago. Also probably all big AI companies work on AI for autonomous robots. Where do you intend to do s.th. different to get a shot at competing with them (when they have so much more capital)?
I really do think that building through the open source community is the best way to compete with the big players, even without having a lot of capital. Of course, it doesn't mean we can't execute well, but I do think it's a good way to make a lot of progress without spending a lot of money.
Pretty sweet! Don’t have the time (or budget) to directly invest in the hardware but do you have a list of open source “open” software problems you are looking to solve?
a) some of these definitely look they could be done without hardware or with light hardware support from a staff member?
b) if you do a) and are open source completely I bet you don’t even need to do bounties. The increased participation could mean some great community generated solutions quick.
I think it's tricky to manage an open source community effectively while still being opinionated on the product. I don't want to get too sucked into coordinating adhoc contributors - we do have a strong core team of people, and we all live together, which is pretty important. But yea, still figuring out what the right dynamic here will look like. Thanks for the suggestions!
That's fair, especially around firmware and things like OTA updates and lower level controls/safety you'd want to keep that closer to team.
But on the long-tail of autonomy options? Implementing the latest papers, trying cutting edge solutions, I bet a thriving open-source community could be very helpful ala PR2, given that the hardware is already open-source.
Nothing stopping you from picking and choosing from the various implementations to build into a streamlined product offering on the front end.
Yep I definitely agree. Our team is pretty small and bandwidth limited
I basically think our goal is to solve all the boring stuff and make it work reliably, so that other can people can try out the cool ML stuff more easily
I love the idea of humaniod robot and commercially available. I like to think of such expensive things as an investment rather than a toy if I have to buy. Question is what are some good use cases that can be solved with such a humanoid robot ?
To be honest, I don't know that there are many applications today which humanoids are the best form factor to solve. I would view it more as a form factor that is likely to get much more useful over the next few years, and having the hardware in your home lets you try out new techniques as they come online.
Personally, I think the first real use cases will mostly be entertainment. Humanoids have a high "coolness" factor. Also, I think there's a long tail of random problems which you don't want to buy a new robot to solve, but which, if you have a robot lying around that isn't perfect but is "good enough", might be possible to solve imperfectly. For example, I just had a newborn baby, and I was thinking it would be nice if I had a static arm that could hold his bottle for me. There's a lot of tail end problems like that in your day to day life. But I think the really interesting capabilities will come once there's very good end-to-end models running on-device.
Thanks for an honest response. I did some google search and I see that even a simple Robotic arm costs over $15, so K-Bot is at a good price point . If I have to invest purely for learning and trying to integrate with something like Gemini Robotics SDK, do you think it will work ?
Yeah, this is exactly the kind of use case we intend to support. Basically, we want our robot to be the best mass-produced robot for this kind of experimentation.
Yes, we are aware of this customer segment. I don't think they have thoroughly considered the implications of what high torque actuators can do to the human body
Every technological leap has it's Chuck Yeagers and Yuri Gagarins that will put it all on the line with early tech for humanity to take that next step - we have to accept the inevitable and hope that luck is on these brave soul's side
> the demand side, the basic problem with humanoid robots is that they're mostly useless right now ... ... to square this circle, ... we will provide free hardware and software upgrades until we are able to make the robot fully autonomous...This way, we can have some extra cash upfront to kickstart development
Congratulations guys! The technical stuff is above my paygrade, but you have a cracked team and with open source you will have a great chance to be close or at SOTA level at your price point.
However, it looks to me that your core thesis is yes, when the autonomous robots get good enough, even at a medium family car price range they will sell like candies. Sure. But since you also want to have the cash now, to who exactly are you selling? Yes you promise that you will support the full autonomy option, but this sounds weirdly similar to Tesla selling cars promising the FSD, which we all know how that story went.
I'm not saying you won't deliver, I'm just saying you might need to a bit more careful in your story selling/narrative for this. For example, i would be super interested to get one for like 2k if it's not useful now, but paying 10k for essentially promises and possible upgrades is a bit iffy. Hence i would like to at least see some plug in and play current usecases? Even if they are just for fun.
I spent two years on Tesla's FSD team, and I think from a cash flow perspective for funding R&D this model did make a lot of sense - basically, it takes cash upfront for training models, but there's zero marginal cost for distributing the models once you've developed them.
I think this kind of "promise the future, pay now" model does alienates some people, especially when the tech is not ready today. That's why we're open sourcing everything, to avoid the feeling of overpromising on what is ready today. The core idea is that the people who bought FSD early on were very invested in it's success, and that feedback loop is very important for improving machine learning models at scale. The problem happens when actually delivering on the tech takes a long time, but I think we have a fairly clear technical roadmap to make our robot useful. At least, I think there are a lot more intermediate benchmarks for driving value for a humanoid robot than there are for self-driving cars, so I think people who buy it will have a stronger feeling that it is constantly improving.
From a cash flow perspective of course it makes sense to sell the future before you have it as working product. It just needs a great salesman or narrative to keep it going, im not arguing that.
> that feedback loop is very important for improving machine learning models at scale
Oh will you have your own feedback loop with let's say user's data? Or you meant as an example?
> * That's why we're open sourcing everything, to avoid the feeling of overpromising on what is ready today*
I agree here, it helps the today, but I dont think it helps the feeling of overpromising on what is ready today, its more like, even if it's open source , it does not increase the chances of it being ready/autonomous in the future. (im just playing devils advocate here)
I also agree with the intermediate benchmarks for sure, this is more to what i was referring to, it would be nice to see some more short term usecases/fun applications that are realistic to hit today or in the nearer future, that would drive a lot of sales value, at least for me, rather than go from now to full autonomy. Good luck!
> Oh will you have your own feedback loop with let's say user's data? Or you meant as an example?
That's more or less the idea - obviously since it's open source we wouldn't scrape peoples' data without their consent, but I would hope that people would contribute to the project in some form. Like, the core idea of the open source ethos is that building something like this collaboratively is a better / cheaper way to scale data collection / experience than us trying to collect all the data ourselves.
> it does not increase the chances of it being ready/autonomous in the future.
Yea that's true. At the end of the day it's just technical execution, so it's pretty risky. I just prefer that if people sign up for something risky, it's pretty transparent what exactly it is they're signing up for :)
Yeah of course. That's kind of the whole point - I don't think you can really trust a humanoid robot in your house around your family if it is not clear what it's running. Basically, for myself, I would not want to buy a closed sourced humanoid, and I view myself as relatively representative of the early adopter mentality. So personally I think this is the right way to build a great product.
I basically believe that in a world where humanoid robots are actually useful, we will not have any trouble monetizing. Probably we will verticalize manufacturing at some point in the future. I think the bigger risks for our business model are not from people copying us or something, but from not making progress fast enough.
Your current market seems to be "niche toys for rich tech people" and the future market seems very uncertain. I am impressed that you were able to get funding for this idea. How do you get around the "solution in search of a problem (SISP)" objection from VCs? In fact, your founding story indicates that you just liked the technology meaning you had to work backwards to find the business case.
I'm asking because I think many of us would like to get funding for ventures in areas of technology that we are passionate about, but for which the future market potential remains extremely speculative. How do you do it?
I have a pretty bad mental model of how most VCs think, but I think good VCs will fund smart people who demonstrate extreme conviction, regardless of how they initially size the market. The opportunity cost for me doing K-Scale is making quite a bit of money at Tesla or Meta, so assuming I am not acting irrationally, either I have extreme conviction or I am a masochist. In my experience, VCs are pretty bad at telling the difference.
Congratulations! This looks really great. What've you found to be the best hands / end effectors these days? When do you think we'll have good, reliable 5 finger hands that are ~reasonably priced?
I'm not convinced that 5 finger hands are necessary right now, but there is a really long tail of hand suppliers that we've been exploring to help get the price down.
I think at volume the price for a good set of hands should settle somewhere around $300-500. Most of it comes down to meeting suppliers where they're at and negotiating mutually beneficial deals. It's not magic but it does require having a good understanding of the hardware in order to negotiate well.
Actually yea, the benefit of our parallel gripper is that we get some proprioceptive feedback which we can't get from the current 5 finger hand. I'm not sure how important this will be long term - I think vision can eventually mostly compensate if the ML models are good enough
I give you a lot of credit for your communication and openness. I am afraid full-autonomy is a fairy tale to rationalize the ungodly expense of LLMs, I personally would find taking money on spec for something like fully autonomous robotics to be a little shaky, ethically speaking. But then, I don't believe this is an unsolved problem, I believe this is an unsolvable problem, so more of a philosophical difference than an empirical position at this point.
This is awesome! How much of your team's time goes into working on the physical hardware, versus RL simulation environments, versus managing all the training data from the real robot and the simulations?
I'm super interested in learning more about the training process of world and robotics model and the data challenges there.
We're all pretty cross-stack - there are some hardware people and some software people, but the product is quite integrated. Personally, my time has been mostly focused on the RL stack recently, and after there are more robots in the wild, I suspect my time will transition to working on building this data feedback loop.
I try to answer questions pretty actively on our Discord so happy to chat there about whatever you like
My friend and I are so excited about this bot that we're actively looking for AI grants to apply to for funding the purchase! The price is incredible for what you get, but we both work in the public sector :/
Consider the ROS simulators first, and look at the platforms used in RoboCup events. There are better platforms than Talos or HECTOR around.
It can take years to get a basic bipedal platform operational, and in general it takes 3 times as long to tune the software/firmware. Unless you see actual proof of platform operation for more than a minute, than take any claims as marketing hype.
Could also try a cheap servo hexapod or turtle-bot kit first, as stable well-studied platforms are easier to code on. =3
How do you plan on competing with existing Chinese manufacturers. Unitree for instance sells there robots at a reasonable price and already has walking working.
Well, I believe the competition up until now has mostly been on hardware, but moving forward it will mostly be in software. I don't think we will be outcompeted by Unitree on software. And I hope to capitalize on our engagement with the open source community in a way that Unitree has not.
Yeah, really like that project. The main difference is that it's much shorter than ours, and shorter robots are cheaper. The downside is that it can't really reach normal human spaces. Actually, we made another robot called Zeroth Bot which you can build for $350 if you want: https://docs.kscale.dev/docs/zeroth-01
We are planning to release a similar-size robot later this year (calling it M-Bot) that will be closer in height and price, but our current focus has been on launching the full-size humanoid.
Mechanically, I think Berkeley Humanoid Lite is pretty similar to the 3D printed one we made last year. Our main focus with the K-Bot redesign was to make it not break so much. 3D printed components break a lot and repair time can be quite long. Also, having the wiring routed internally makes a huge, huge difference. So there are some benefits to doing QA on a manufacturing line in terms of quality and consistency.
That is a good question. I think mechanically we're probably there today, but on the intelligence level, who knows... If I had to guess, probably 1-3 years
We try to sell it for a fair price while still making money. Actually I think there will be comparably priced humanoids coming out from other companies soon
Sorry maybe I should have been more specific. I had Unitree in mind, last time I checked the humanoid with no hands was $20k and each hand was an extra $20k. Yours seems to ship with 2 hands for extra $1k (surprisingly cheap, which is great of course!). I was curious to read more about what the hands are capable of doing.
Oh I see. Yes, we're currently exploring a few different five finger hand options - we will choose whichever option provides the best value. I actually quite like the Inspire hands and we might be able to get a volume discount
Yeah, the basic robot is just a robot, albeit with open-source software and hardware
For compute - we're exploring a few boards right now, but the base model will be something from Amlogic and the higher end model will be something from Nvidia
nice! For manipulation taskes, is kscale planning to train in sim mostly and transferring to real, or is some imitation learning used, like with gloves and such? maybe both?
I think we will just focus on making really great, low-cost hardware and a nice SDK, and let other people experiment with different approaches on the intelligence layer.
Hey, just chiming in to say that I think this project is really cool even though it's outside the price range of what I can spend on a cool hobby.
I'm disabled, and one thing I'm really interested in long-term for humanoid robots is disability support work. Disability support work involves a huge variety of individual tasks, as many as a typical person will do in their life, so it's a good fit for an extremely general platform like a humanoid robot. Motorised wheelchairs and dishwashers exist, but a support worker might need to push a wheelchair, do sensitive dishes, do laundry, accurately open and place medications without destroying them, weigh & dose powders, help someone with going to the toilet, cook meals, drive a car, control pets, manage the level of noise/light/smells in the environment to stop someone from being overwhelmed, sanitise surfaces including themselves, navigate confusing interfaces on a phone or computer, help someone drink from a bottle, remember what sort of activities helped a disabled person in the past to be able to do them in the future, help someone with physical fitness activities like punching or kicking a pad, talk to people for someone, carry someone safely in the event of an emergency, make coffee in the morning, monitor intake of various drugs/nutrients/macronutrients, be able to reach and catch someone before they hit the floor if they pass out, help someone walk if they're unsteady on their feet, etc etc. It makes sense to me that it would be cost effective to have one platform which can do all of that with similar performance to a human, rather than automating many of those tasks individually in ways that might not be accessible to some disabled people.
In terms of TAM, absolutely huge amounts of money are spent on disability care (keeping in mind that elder care is also disability care), by both governments and private citizens, and this number is forecasted to continue growing as more people become disabled by COVID-19 and demographic changes increase the elderly population relative to working age adults. As well, there are constantly scandals about how bad conditions are in some area of disability care, almost always due to underpaid, untrained, or unmonitored staff, so there's a lot of demand for both more reliable quality & lower prices; that demand is only going to grow with time. Various government bodies are very large sources of funding that are very concerned with value for money and would pursue any option that could do the job without costing as much - in my country (Australia), there's the NDIS, National Disability Insurance Scheme. They are always looking for ways to consolidate care for less money.
I strongly suspect that any humanoid robot which was good enough to do disability support work would be in extremely high demand in the general population for obvious reasons, as well as being useful as a platform for labour automation, but those are much more speculative. Disability support work is a lot of money for incredibly varied tasks being spent right now. Something to think about.
This is really interesting to read about. To be honest, I know very little about this space, but it's something a few people have approached me about tackling.
I do think that this is a great application of a general purpose robot. I'm not sure what the technical timeline will be, but it would certainly be cool for my parents to have such a robot when they are elderly.
I spend a lot of time thinking about it day-to-day because of my disability and reliance on multiple disability support workers, as well as living with my husband who is also disabled, so if you ever wanted to talk to someone with disability support workers feel free to ask.
This, to me, is the most compelling and humane application of humanoids. Often I think people jump to humanoids taking jobs, but wow, it would be so incredible for elderly folks to have a humanoid that can help them.
Tao Motors reached out to us about a month ago after our soft launch because they just bought a big factory in Texas to manufacture golf karts, and the supply chain for golf karts and humanoids is actually pretty similar. Their subsidiary invested some money in our company recently and we are partnering with them to scale manufacturing. I discussed this in our launch video.
Nothing against the company but I'm waiting for the tech to get backlash. I have a feeling people are going to want to end the techno autocracy and so-calledb advancements that go on to become weapons while people continue to go back to old ways and learn old skills that prove to more useful.
Yeah I think humanoids are a pretty fraught area. There's definitely been some backlash but overall I have really appreciated the responses we've got from people. Like, I just want people to care about what we're doing - if positive that's great, if negative then we can learn how to do better. As long as people have an opinion one way or the other, I'm happy :)
Congratulations on the launch! This is really cool.
I'm not super active in the humanoid robot space anymore, however I did my PhD about 9 years ago in HRI. That was the time of Boston Dynamics, DARPA robotics challenge, and Aldebaran's Pepper and Nao robots.
You mentioned you are building everything open source. What happened with ROS and related projects? Do you build on top of that, or is that all super outdated that a reboot was needed?
Another question I have is: why are you choosing a two-legged human over a four-legged one?
My experiments with two legged robots were mostly bad. Not only did they fall basically all the time but they also had a big drift. So far, I have not seen any large improvements. But again, I might be very outdated.
I always said to my colleagues. The main point stopping robots from picking up is a stable platform. And with the platform I mean walking.
Eh. I think we got a bit nerd-sniped on some things and we ended up trying to build most of our stack ourselves. The control loop is just a pretty simple Rust loop for running ML models. ROS is kind of annoying to work with and the control loop is pretty simple so we just wrote it ourselves.
For two legged - I think it will eventually be quite a bit cheaper, it's mostly a software problem to get them to be stable. RL based control has gotten much, much better. For example, I was talking to Trevor Blackwell a few weeks ago, and he was saying the IMU on the original Anybots robot was over $2k, whereas if you throw a noisy IMU into sim, you can get away with something basically from a cellphone. So yea, personally I'm a big believer in needing to solve the robotics intelligence problem, and once you solve that, humanoids will make the most sense as a form factor.
I have some technical questions about feet.
Human feet have metatarsophalangeal joints connecting the toes to the rest of the foot. But humanoid robots don't have these (at least, the vast majority don't). Why? These joints are very useful.
Also, the bottom of the human foot is soft and has thousands of nerve endings. Can we really expect robots to get anywhere near human mobility performance without this level of compliance and sensory sophistication?
Feet/ankles on humanoids are really hard mechanically. In many ways the kinematic requirements for the ankle are similar to a wrist, but while the wrist of a robot arm is the least-heavily-loaded, the ankle area can be the most heavily loaded. Humans get away with it by having most of the highly-stressed actuators in the lower leg, not the ankle itself, whereas many humanoid robots end up putting more of the actuators in the ankle assembly itself.
In general, I think the best way to think about the differences between human muscles and robot actuators is that human muscles are simultaneously incredible in terms of strength and power density, and also incredibly fragile. Robot actuators are fairly robust, but comparatively poor. Humans are essentially falling apart at all times, but the repair mechanisms in the body do a good enough job that it doesn't matter (although you probably know someone with a "career-disruptive/ending" sports-related injury that shows these mechanisms have limits). Robotics is a long way away from making actuators that can be fixed online in such a process. Even cable stretching in cable-driven mechanisms remains hard to handle effectively, and that's one of the simplest types of mechanism wear.
These are the kinds of comments that make comments worth reading. Thank you!
This is a much better answer than mine
So, most humanoids you see that are relatively cost-effective are just "humanoid" in that they look like humans, but there are significant mechanical differences between them and real humans. It's almost always driven by the cost of manufacturing different components. A good example is the lead screw you see in the knee and ankle on Optimus - while it is more human-centric to have tendon-like actuation, it drives the price up quite a bit. Put differently, humans have a lot of specialization which is not great if you want to mass manufacture something.
For walking, the most important thing is that the robot can be simulated well, so in our case, we tried to model the foot contact with the ground in simulation quite accurately. In fact, we found that force sensors in the foot probably help but they're not necessary in simulation, and we wanted to shave off anything that wasn't necessary. I am not sure how close we will get to human mobility - it's definitely a learning process - but you can get much further in simulation than you'd expect.
Cool.
But this looks like an expensive toy.
The stuff of nightmares is this being adapted by the DoD. I can almost imagine your website as a scene in the prologue of a terminator like movie.
Nightmare 2 is this becomes a companion of some sort. Detroit Become Human goes into this. You have a theme of the robots basically wanting freedom. Which throws out a moral conundrum, if someone buys an AGI enabled bot just to be mean to it, have they done anything wrong.
I like technology , but this feels like step one to a whole lot of weird stuff.
I suspect that these kinds of things will happen with or without my involvement. Assuming that they do, I would rather that a good open source option exists
What ML algorithms do you intend for full autonomy? Multi Modal LLMs for planning that control the robot by generating s.th. like code? Or s.th. that requires more learning from the environment?
When I click "get in touch" on your github I just land on the website where I can buy the robot. Building the hardware for an autonomous robot is orders of magnitudes easier than the control. Do you think anyone with the capability do develop an autonomus robot will buy this and then just give you the code because its open source?
My overall plan is basic joystick control -> VLA with RL -> self-supervised embodied representation -> end-to-end RL -> end-to-end control. I suspect there will be some very good multi modal models coming out in the next few years which we might use as base models, although more likely, we will adapt their techniques to work on data from our own robot.
I agree that the hardware is easier than the software - I am a software guy, personally, but I felt that it was important to do the hardware first so at least we can have a baseline product which we can offer to people. I would personally like to work on this software problem (or rather, build a company to work on this problem), and this seems like the right way to go about funding working on this problem.
I like the K-Sim Gym. Im looking forward to fiddle with it a bit when I have more time. I could see that you get something usefull out of people competing on your leaderboard xD
It's my hot take that the next big ML breakthrough needs s.th. that learns from its own actions in an environment, so this goes in the right direction imo.
On the other hand a lot of big companies struggle with self driving cars even though they predicted to build this years ago. Also probably all big AI companies work on AI for autonomous robots. Where do you intend to do s.th. different to get a shot at competing with them (when they have so much more capital)?
I really do think that building through the open source community is the best way to compete with the big players, even without having a lot of capital. Of course, it doesn't mean we can't execute well, but I do think it's a good way to make a lot of progress without spending a lot of money.
Pretty sweet! Don’t have the time (or budget) to directly invest in the hardware but do you have a list of open source “open” software problems you are looking to solve?
We do! https://bounties.kscale.dev/
We haven't updated it much but it's a good starter point
Amazing, pretty nice list! Two quick suggestions:
a) some of these definitely look they could be done without hardware or with light hardware support from a staff member?
b) if you do a) and are open source completely I bet you don’t even need to do bounties. The increased participation could mean some great community generated solutions quick.
I think it's tricky to manage an open source community effectively while still being opinionated on the product. I don't want to get too sucked into coordinating adhoc contributors - we do have a strong core team of people, and we all live together, which is pretty important. But yea, still figuring out what the right dynamic here will look like. Thanks for the suggestions!
That's fair, especially around firmware and things like OTA updates and lower level controls/safety you'd want to keep that closer to team.
But on the long-tail of autonomy options? Implementing the latest papers, trying cutting edge solutions, I bet a thriving open-source community could be very helpful ala PR2, given that the hardware is already open-source.
Nothing stopping you from picking and choosing from the various implementations to build into a streamlined product offering on the front end.
Yep I definitely agree. Our team is pretty small and bandwidth limited
I basically think our goal is to solve all the boring stuff and make it work reliably, so that other can people can try out the cool ML stuff more easily
I love the idea of humaniod robot and commercially available. I like to think of such expensive things as an investment rather than a toy if I have to buy. Question is what are some good use cases that can be solved with such a humanoid robot ?
To be honest, I don't know that there are many applications today which humanoids are the best form factor to solve. I would view it more as a form factor that is likely to get much more useful over the next few years, and having the hardware in your home lets you try out new techniques as they come online.
Personally, I think the first real use cases will mostly be entertainment. Humanoids have a high "coolness" factor. Also, I think there's a long tail of random problems which you don't want to buy a new robot to solve, but which, if you have a robot lying around that isn't perfect but is "good enough", might be possible to solve imperfectly. For example, I just had a newborn baby, and I was thinking it would be nice if I had a static arm that could hold his bottle for me. There's a lot of tail end problems like that in your day to day life. But I think the really interesting capabilities will come once there's very good end-to-end models running on-device.
Thanks for an honest response. I did some google search and I see that even a simple Robotic arm costs over $15, so K-Bot is at a good price point . If I have to invest purely for learning and trying to integrate with something like Gemini Robotics SDK, do you think it will work ?
Yeah, this is exactly the kind of use case we intend to support. Basically, we want our robot to be the best mass-produced robot for this kind of experimentation.
[dead]
congrats! just sharing also the behind the scenes talk that one of your engineers did at AIE: https://www.youtube.com/watch?v=BS92RdBvI90
>However, a lot of the people who want to buy a humanoid robot today do so because they want a completely autonomous robot to do all their chores,
Not sure your research has been through.
The ones that get the most attention from what I've seen are the ones that look female. And the first comment is always about how easy to clean...
All those lonely men spending thousands on the billion dollar revenue generating onlyfans and webcam sites seem to be the immediate consumer market.
Yes, we are aware of this customer segment. I don't think they have thoroughly considered the implications of what high torque actuators can do to the human body
Every technological leap has it's Chuck Yeagers and Yuri Gagarins that will put it all on the line with early tech for humanity to take that next step - we have to accept the inevitable and hope that luck is on these brave soul's side
You might have a significant portion of those customers that are into that
Castration via motorized actuator? Pretty niche kink, I'd say.
> the demand side, the basic problem with humanoid robots is that they're mostly useless right now ... ... to square this circle, ... we will provide free hardware and software upgrades until we are able to make the robot fully autonomous...This way, we can have some extra cash upfront to kickstart development
Congratulations guys! The technical stuff is above my paygrade, but you have a cracked team and with open source you will have a great chance to be close or at SOTA level at your price point.
However, it looks to me that your core thesis is yes, when the autonomous robots get good enough, even at a medium family car price range they will sell like candies. Sure. But since you also want to have the cash now, to who exactly are you selling? Yes you promise that you will support the full autonomy option, but this sounds weirdly similar to Tesla selling cars promising the FSD, which we all know how that story went.
I'm not saying you won't deliver, I'm just saying you might need to a bit more careful in your story selling/narrative for this. For example, i would be super interested to get one for like 2k if it's not useful now, but paying 10k for essentially promises and possible upgrades is a bit iffy. Hence i would like to at least see some plug in and play current usecases? Even if they are just for fun.
I spent two years on Tesla's FSD team, and I think from a cash flow perspective for funding R&D this model did make a lot of sense - basically, it takes cash upfront for training models, but there's zero marginal cost for distributing the models once you've developed them.
I think this kind of "promise the future, pay now" model does alienates some people, especially when the tech is not ready today. That's why we're open sourcing everything, to avoid the feeling of overpromising on what is ready today. The core idea is that the people who bought FSD early on were very invested in it's success, and that feedback loop is very important for improving machine learning models at scale. The problem happens when actually delivering on the tech takes a long time, but I think we have a fairly clear technical roadmap to make our robot useful. At least, I think there are a lot more intermediate benchmarks for driving value for a humanoid robot than there are for self-driving cars, so I think people who buy it will have a stronger feeling that it is constantly improving.
From a cash flow perspective of course it makes sense to sell the future before you have it as working product. It just needs a great salesman or narrative to keep it going, im not arguing that.
> that feedback loop is very important for improving machine learning models at scale
Oh will you have your own feedback loop with let's say user's data? Or you meant as an example?
> * That's why we're open sourcing everything, to avoid the feeling of overpromising on what is ready today*
I agree here, it helps the today, but I dont think it helps the feeling of overpromising on what is ready today, its more like, even if it's open source , it does not increase the chances of it being ready/autonomous in the future. (im just playing devils advocate here)
I also agree with the intermediate benchmarks for sure, this is more to what i was referring to, it would be nice to see some more short term usecases/fun applications that are realistic to hit today or in the nearer future, that would drive a lot of sales value, at least for me, rather than go from now to full autonomy. Good luck!
> Oh will you have your own feedback loop with let's say user's data? Or you meant as an example?
That's more or less the idea - obviously since it's open source we wouldn't scrape peoples' data without their consent, but I would hope that people would contribute to the project in some form. Like, the core idea of the open source ethos is that building something like this collaboratively is a better / cheaper way to scale data collection / experience than us trying to collect all the data ourselves.
> it does not increase the chances of it being ready/autonomous in the future.
Yea that's true. At the end of the day it's just technical execution, so it's pretty risky. I just prefer that if people sign up for something risky, it's pretty transparent what exactly it is they're signing up for :)
> we're open sourcing everything
Does/will this include the training data, hyperparams, and weights for the models?
People will be reluctant to buy an "open source" robot if the key ML magic to make it work is closed off, e.g. if you charge a subscription for it.
Yeah of course. That's kind of the whole point - I don't think you can really trust a humanoid robot in your house around your family if it is not clear what it's running. Basically, for myself, I would not want to buy a closed sourced humanoid, and I view myself as relatively representative of the early adopter mentality. So personally I think this is the right way to build a great product.
I basically believe that in a world where humanoid robots are actually useful, we will not have any trouble monetizing. Probably we will verticalize manufacturing at some point in the future. I think the bigger risks for our business model are not from people copying us or something, but from not making progress fast enough.
Love it. It's tough to balance keeping the ship afloat vs giving back to the community, I think you've got a good thing going.
Great to hear, thanks for the response!
Congrats on the launch!
Your current market seems to be "niche toys for rich tech people" and the future market seems very uncertain. I am impressed that you were able to get funding for this idea. How do you get around the "solution in search of a problem (SISP)" objection from VCs? In fact, your founding story indicates that you just liked the technology meaning you had to work backwards to find the business case.
I'm asking because I think many of us would like to get funding for ventures in areas of technology that we are passionate about, but for which the future market potential remains extremely speculative. How do you do it?
I have a pretty bad mental model of how most VCs think, but I think good VCs will fund smart people who demonstrate extreme conviction, regardless of how they initially size the market. The opportunity cost for me doing K-Scale is making quite a bit of money at Tesla or Meta, so assuming I am not acting irrationally, either I have extreme conviction or I am a masochist. In my experience, VCs are pretty bad at telling the difference.
Congratulations! This looks really great. What've you found to be the best hands / end effectors these days? When do you think we'll have good, reliable 5 finger hands that are ~reasonably priced?
I'm not convinced that 5 finger hands are necessary right now, but there is a really long tail of hand suppliers that we've been exploring to help get the price down.
I think at volume the price for a good set of hands should settle somewhere around $300-500. Most of it comes down to meeting suppliers where they're at and negotiating mutually beneficial deals. It's not magic but it does require having a good understanding of the hardware in order to negotiate well.
Have you identified any limitations in current grippers based on lack of tactile sensing solutions to unlock truly dextrous manipulation?
Actually yea, the benefit of our parallel gripper is that we get some proprioceptive feedback which we can't get from the current 5 finger hand. I'm not sure how important this will be long term - I think vision can eventually mostly compensate if the ML models are good enough
I give you a lot of credit for your communication and openness. I am afraid full-autonomy is a fairy tale to rationalize the ungodly expense of LLMs, I personally would find taking money on spec for something like fully autonomous robotics to be a little shaky, ethically speaking. But then, I don't believe this is an unsolved problem, I believe this is an unsolvable problem, so more of a philosophical difference than an empirical position at this point.
I think it's a very hard problem, but I would like to align our company's incentives towards solving it
Great. Another problem for us to solve, how defend ourselves from your robots.
Just because you can, doesn't mean you should.
This is awesome! How much of your team's time goes into working on the physical hardware, versus RL simulation environments, versus managing all the training data from the real robot and the simulations?
I'm super interested in learning more about the training process of world and robotics model and the data challenges there.
Thanks!
We're all pretty cross-stack - there are some hardware people and some software people, but the product is quite integrated. Personally, my time has been mostly focused on the RL stack recently, and after there are more robots in the wild, I suspect my time will transition to working on building this data feedback loop.
I try to answer questions pretty actively on our Discord so happy to chat there about whatever you like
I'll hop in there!
My friend and I are so excited about this bot that we're actively looking for AI grants to apply to for funding the purchase! The price is incredible for what you get, but we both work in the public sector :/
We're exploring some options - maybe a rental option in the future. Would like to make it accessible to everyone
Consider the ROS simulators first, and look at the platforms used in RoboCup events. There are better platforms than Talos or HECTOR around.
It can take years to get a basic bipedal platform operational, and in general it takes 3 times as long to tune the software/firmware. Unless you see actual proof of platform operation for more than a minute, than take any claims as marketing hype.
Could also try a cheap servo hexapod or turtle-bot kit first, as stable well-studied platforms are easier to code on. =3
How do you plan on competing with existing Chinese manufacturers. Unitree for instance sells there robots at a reasonable price and already has walking working.
Well, I believe the competition up until now has mostly been on hardware, but moving forward it will mostly be in software. I don't think we will be outcompeted by Unitree on software. And I hope to capitalize on our engagement with the open source community in a way that Unitree has not.
https://lite.berkeley-humanoid.org/ is only $5000. What's the extra $3000 get me?
Yeah, really like that project. The main difference is that it's much shorter than ours, and shorter robots are cheaper. The downside is that it can't really reach normal human spaces. Actually, we made another robot called Zeroth Bot which you can build for $350 if you want: https://docs.kscale.dev/docs/zeroth-01
We are planning to release a similar-size robot later this year (calling it M-Bot) that will be closer in height and price, but our current focus has been on launching the full-size humanoid.
Mechanically, I think Berkeley Humanoid Lite is pretty similar to the 3D printed one we made last year. Our main focus with the K-Bot redesign was to make it not break so much. 3D printed components break a lot and repair time can be quite long. Also, having the wiring routed internally makes a huge, huge difference. So there are some benefits to doing QA on a manufacturing line in terms of quality and consistency.
How far away from robot hands/manipulators that are dexterous enough to repair other robots?
That is a good question. I think mechanically we're probably there today, but on the intelligence level, who knows... If I had to guess, probably 1-3 years
obviously there's more to the kinematics than making the legs longer, but can't they just make the legs longer for it to be taller?
Longer legs means more inertia on the actuators, which means bigger actuators + bigger battery, which means heavier robot, which means...
right? thanks!
I appreciate the humility in your videos, good luck with the launch!
Is there any doc on the hand? It looks surprisingly cheap.
We try to sell it for a fair price while still making money. Actually I think there will be comparably priced humanoids coming out from other companies soon
Sorry maybe I should have been more specific. I had Unitree in mind, last time I checked the humanoid with no hands was $20k and each hand was an extra $20k. Yours seems to ship with 2 hands for extra $1k (surprisingly cheap, which is great of course!). I was curious to read more about what the hands are capable of doing.
Oh I see. Yes, we're currently exploring a few different five finger hand options - we will choose whichever option provides the best value. I actually quite like the Inspire hands and we might be able to get a volume discount
So the non full autonomy would mean little software upgrades? More do it yourself?
Also, what’s the different bt the computes: like what’s the onboard computer running (the 2 options)?
Thanks.
Yeah, the basic robot is just a robot, albeit with open-source software and hardware
For compute - we're exploring a few boards right now, but the base model will be something from Amlogic and the higher end model will be something from Nvidia
nice! For manipulation taskes, is kscale planning to train in sim mostly and transferring to real, or is some imitation learning used, like with gloves and such? maybe both?
Appreciate it
I think we will just focus on making really great, low-cost hardware and a nice SDK, and let other people experiment with different approaches on the intelligence layer.
Hey, just chiming in to say that I think this project is really cool even though it's outside the price range of what I can spend on a cool hobby.
I'm disabled, and one thing I'm really interested in long-term for humanoid robots is disability support work. Disability support work involves a huge variety of individual tasks, as many as a typical person will do in their life, so it's a good fit for an extremely general platform like a humanoid robot. Motorised wheelchairs and dishwashers exist, but a support worker might need to push a wheelchair, do sensitive dishes, do laundry, accurately open and place medications without destroying them, weigh & dose powders, help someone with going to the toilet, cook meals, drive a car, control pets, manage the level of noise/light/smells in the environment to stop someone from being overwhelmed, sanitise surfaces including themselves, navigate confusing interfaces on a phone or computer, help someone drink from a bottle, remember what sort of activities helped a disabled person in the past to be able to do them in the future, help someone with physical fitness activities like punching or kicking a pad, talk to people for someone, carry someone safely in the event of an emergency, make coffee in the morning, monitor intake of various drugs/nutrients/macronutrients, be able to reach and catch someone before they hit the floor if they pass out, help someone walk if they're unsteady on their feet, etc etc. It makes sense to me that it would be cost effective to have one platform which can do all of that with similar performance to a human, rather than automating many of those tasks individually in ways that might not be accessible to some disabled people.
In terms of TAM, absolutely huge amounts of money are spent on disability care (keeping in mind that elder care is also disability care), by both governments and private citizens, and this number is forecasted to continue growing as more people become disabled by COVID-19 and demographic changes increase the elderly population relative to working age adults. As well, there are constantly scandals about how bad conditions are in some area of disability care, almost always due to underpaid, untrained, or unmonitored staff, so there's a lot of demand for both more reliable quality & lower prices; that demand is only going to grow with time. Various government bodies are very large sources of funding that are very concerned with value for money and would pursue any option that could do the job without costing as much - in my country (Australia), there's the NDIS, National Disability Insurance Scheme. They are always looking for ways to consolidate care for less money.
I strongly suspect that any humanoid robot which was good enough to do disability support work would be in extremely high demand in the general population for obvious reasons, as well as being useful as a platform for labour automation, but those are much more speculative. Disability support work is a lot of money for incredibly varied tasks being spent right now. Something to think about.
This is really interesting to read about. To be honest, I know very little about this space, but it's something a few people have approached me about tackling.
I do think that this is a great application of a general purpose robot. I'm not sure what the technical timeline will be, but it would certainly be cool for my parents to have such a robot when they are elderly.
I spend a lot of time thinking about it day-to-day because of my disability and reliance on multiple disability support workers, as well as living with my husband who is also disabled, so if you ever wanted to talk to someone with disability support workers feel free to ask.
This, to me, is the most compelling and humane application of humanoids. Often I think people jump to humanoids taking jobs, but wow, it would be so incredible for elderly folks to have a humanoid that can help them.
Your build guide link is a 404
Which link? I can fix
I'm so glad I sent them a check asap. Super proud that there's georgia (as in 404) in the founder crew
Really appreciate it :)
Looks very cool! $9K is well outside my budget, but very reasonable for even small startups -- props.
Small note: https://www.kscale.dev/why is a 404
Oh thank you! Will fix
Seems like the thing is entirely manufactured and mostly designed by a Chinese company?
https://mp.weixin.qq.com/s/V_WVFSJg3cTPq0Y4gK4cHw
That post says "manufactured", why would you assume "mostly designed"?
Any robotics company that is not thinking about manufacturing from day 1 is not a serious robotics company.
Tao Motors reached out to us about a month ago after our soft launch because they just bought a big factory in Texas to manufacture golf karts, and the supply chain for golf karts and humanoids is actually pretty similar. Their subsidiary invested some money in our company recently and we are partnering with them to scale manufacturing. I discussed this in our launch video.
Nothing against the company but I'm waiting for the tech to get backlash. I have a feeling people are going to want to end the techno autocracy and so-calledb advancements that go on to become weapons while people continue to go back to old ways and learn old skills that prove to more useful.
Yeah I think humanoids are a pretty fraught area. There's definitely been some backlash but overall I have really appreciated the responses we've got from people. Like, I just want people to care about what we're doing - if positive that's great, if negative then we can learn how to do better. As long as people have an opinion one way or the other, I'm happy :)
[flagged]