Yes, it's a similar situation (though with software, replacement of existing workers and augmentation of existing workers are both possible, while in the physical world, replacement is more natural, assuming robotics gets to that point). What are you implying?
I want to see the robots do everything humans can do in my lifetime. This is good for everyone. When no one has a job we will figure out a way to make it work.
>When no one has a job we will figure out a way to make it work.
Mind-bogglingly casual acceptance of civil war when all you had to do was change the order to put "We will figure out a way to make it work" before "No one has a job".
That’s not mind-boggling. If you start putting out restrictions and regulations everything is going to fizzle out like 100s of other fields have done in the past. There hasn’t been any significant paradigm shifting progress in genetic engineering because everyone’s hands are tied when we could probably accomplish so much that would benefit hundreds of millions of people.
An HVAC tech charged me $750 to perform one swage connection on a mini-split install I was working on. $750 for $0.10 of combustibles and 45 minutes of time. Service workers are willingly pushing their customer base into these robotic solutions and I am thrilled at the idea of a robot coming out to do some work for 1/3 of the cost
As much as I am a skeptic of household maintenance robots, I'll admit that the robots will probably 1) actually show up, 2) show up around the time they say they would, 3) won't be drunk, 4) won't ask me to join their MLM.
Though I'm less sure about (4), considering the Worldcoin eyeball-harvesting scammers and the AI-bailout scammers are one and the same.
Your comment doesn't really make sense. A swage connection is done by a tool and requires no combustibles. If you needed combustibles then you need a license to handle those combustibles. And you need special tools to decide how much of those combustibles to add. You also need special tools to actually add the combustibles to the system. And before that, you need special tools to prep the system for those combustibles. And before you can do that, you need special tools to make sure your system is adequately sealed before even attempting to get the system ready to add the combustibles.
Maybe next time you should spend the few grand on all those tools, the hours it takes to learn how to use them, and can do it yourself!
I can't believe I'm seeing this common low-quality "what do tradeys even do" fallacy as the top comment on an HN submission. It makes me want to cry that we are so fucked as a culture that we are scoffing at how much tradesmen charge (not that there aren't conmen in every single sector).
In all parts of society, there is a subjective spectrum between "making work more efficient" and "directly replacing significant numbers of workers in a period too short to adjust to avoid suffering". The closer you get to the latter, the more you are a "traitor". Use of LLMs for software development falls all over the spectrum, case-by-case.
At what point do these services become free, in the Star Trek post-scarcity sense? How is the promise that automation will ultimately provides freedom from struggling for survival upheld? The piper has to be paid, unless the owner class all of a sudden does a one-eighty and decides they can stop raking in money.
The only way that will happen is when we convert from a capitalist to a socialist democratic economic structure. It'll likely need a whole host of bloody wars for that.
Probably should do that before policing/military humanoid robots are cheap enough. If we don't, we're looking at our future as indentured servitude for capitalists for the rest of the lifetime this earth can support life.
>It'll likely need a whole host of bloody wars for that.
Out of curiosity who are we killing or getting killed by? What amount of brainwashing will it take to get all these rubes to kill eachother to make the 1% richer?
Pro-capitalist governments (cycle of corruption choosing the winners, revolving door between politics and business)
Human and robotic military conscripted to fight
Capitalists
Capitalists' workers paid enough to not be "worker class"
The other side:
Union supporters
Workers rights advocates
Socialists/communists
People who lost jobs/poverty class
Workers abused more than "average"
We have seen these types of wars before, and in the USA. Our own worker rights and union history was horrendously bloody. Even had the US military killing strikers and union folks. It's also where the private military, The Pinkertons, came from.
And prior to 1940s with FDR, there was a strong Socialist and Communist groups pushing for their respective changes. FDR appeased some of the demands with his reforms. And those reforms were credited to the success of the late 50's and 60's.
> What amount of brainwashing will it take to get all these rubes to kill eachother to make the 1% richer?
I suppose whatever the amount is we reached it a while ago, given the number of people who have died for oil prices or to distract from this or that domestic political situation.
According to the Star Trek timeline, we first need to experience the Bell Riots, the Eugenics Wars, and World War III before your proposal can be enacted.
the gen ai image on this page has a "shut off" label for the base of the human plumber's neck. To say nothing of the amusing drainage schematics (water flows up?)...
How is this on the front page? Its not a YC company. They have no internet presence besides a Claude website. 0 followers on X/twitter. No public mentions of funding or votes-of-trust. No names to vouch for them.
How is your process for selling the results? do you have a buyer lined up who is loaning the robots for this in some kinda sponsorship deal, do you intend to sell it as a service later on?
That seems sensible enough to me. I'm not sure why people are acting like it's impossible to succeed if you're not a YC company. I'm pretty sure plenty of non YC companies have done just fine
Well, the parent poster has a point. They are asking: why are you notable? Neither your company, nor you seem to have any track record, references, an indication that you know what you are doing or some prior art that'd be indicative of future performance.
That is not to say that your idea or company are bad, but there's very little substance that would indicate that I should care.
I don't care that you trained your own deepseek derivative or that you got contracts. What you are proposing to do is interesting, but a mission to Mars is also interesting. If it's some random person proposing a mission to Mars, I don't care because they have no chance to actually achieve anything meaningful. If it's a person who has credible support and a plan, sure, now I'm listening.
Why should I care about you doing it? How are you proposing to solve the hard problems, which are in the robotics space?
All good, I was just trying to point out what parent was trying to get at, perhaps too bluntly.
I'm not trying to diss anything you do, kudos for trying!
I generally found that the HN crowd is difficult to please and perhaps a better avenue of getting feedback (and support) is to tailor material specifically for HN.
E.g. if you had a blog post talking about what you are trying to do, the problems you encountered and interesting solutions you are looking at, you'd get a much more interesting response? We get continuously assaulted by "the next big business idea" all the time, both here and in our professional lives. It's difficult to stand out in that deluge. Talking about the technical side can be an easier way to engage people in constructive discussions.
I certainly don't like what the parent commenter said, but I find your response to be even more unprofessional and disrespectful, especially coming from the owner of the company.
Hi there, I have some experience with reinforcement learning, PPO, RLVR, and I trained both video game AIs and sim2real robot policies.
In my experience, videos from professionals are not enough to train a robot, because you lack all/most of the edge and failure cases. Let me illustrate with pilots: The average Airbus can land itself via autopilot. Autopilot will fly it to the correct location. You confirm the frequencies, it connects to the airport's guidance system and lands itself. Super boring. Videos of that won't help anyone become a proper pilot. The reason pilots train for years is to handle the 1 in a 1000 case where the autopilot doesn't work. And you suddenly have to take over in the middle of a messy approach.
Translated back to your plumbers, the real training is when the robot is carrying a 100kg metal pipe and then the owner's kids come running and pass through under it. Every professional will give their best to immediately reduce efficiency by making things stationary and SAFE. That's the training data that is truly valuable.
BTW, I am considering bootstrapping a startup that uses video game engines to generate training data to make robot policies anticipate physical effects and counter-steer before they happen. Like putting down a heavy steel pipe before it drops due to gravity. With 2 bugfixes merged into the Unreal Engine, I have some experience there ;) So I might soon be in the business of selling custom training data.
There's obviously a strong incentive for homeowners to participate: they don't have to pay for a service that is often valued at hundreds of dollars.
But what is the incentive for contractors to participate, and how does it overcome the enormous disincentive of "You are training your replacement"?
My guess is that truly skilled contractors in these trades are not hurting for work. So I doubt they're going to be swayed by just a few more extra jobs added to their docket.
Less skilled contractors who are desperate for work might be taken in, but then the quality of their output is likely to be worse, and the value of that training data is likely to be low.
My guess is that people in skilled trades who are used to working with their hands are going to be even more skeptical than white-collar folks about an AI revolution, and more guarded against change in that direction.
> But what is the incentive for contractors to participate, and how does it overcome the enormous disincentive of "You are training your replacement"?
They don't need an army of contractors to do this. They just need a few, much like the AI labs that have a handful of SME staff for each particular subject to guide the corpus of data. See mathematics example.
It is not about knowing if someone can be bought. It's knowing just how much is just enough to buy them.
And then you have individuals whose pride is more than enough. See Terence Tao for a great example of a Useful Idiot.
> Are you assuming these folks are ... or customers are sharing via word of mouth?
Seems fair to me (interjecting: not the parent poster). I rent, too, but have heard countless conversations along these lines among Proper Adults.
Often grown-up #1 complaining about their latest project and grown-up #2 offering someone that served them well in the past. Point being, bad experiences are both so common and costly... that those who are good tend to stand out and stay busy. IME, those listings or the phone book are most suitable when you don't have a relationship to leverage (effectively rolling the dice).
Yes, "truly skilled" contractors get work entirely through word of mouth. All the best contractors I have discovered were through a relationship, and are always fully booked, frequently doing work on my home due to our relationship, and not due to natural availability.
>But what is the incentive for contractors to participate, and how does it overcome the enormous disincentive of "You are training your replacement"?
Seems to be working fine for AI labs and their army of data annotators (for generating training data and RLHF). Is there any reason to robotics is extra tricky?
> My guess is that people in skilled trades who are used to working with their hands are going to be even more skeptical than white-collar folks about an AI revolution
If I was confident that an AI couldn't do my job, I would happily sign up to do this "training" because it's free money and my job is at zero risk.
It is known that the AI labs are going for all economically useful work.
Creatives and all white collar jobs are the easiest to target first, with AI labs basically pitching individuals and businesses:
"AI will free you up with more time to do 'other meaningful things'"
If you find yourself laid off or work has run dry from a previous gig because AI can do it better than you, you'll naturally consider a role which AI hasn't touched...yet
Then you see AI and robotics labs going after the margins of blue collar jobs, which then they will also say again to the blue collar workers:
"AI will free you up with more time to do 'other meaningful things'"
I can't imagine what the world will look like if you repeat this N times, or even in 10 years.
Because you need to pay your rent, and bills right?
Unless rent, bills and all the rest of it doesn't mean anything anymore in this 'AI abundance' age.
I'm seeing a lot of surprised people on this thread. Have a look at the more recent YC batches of startups. Count all the non-AI ones, and subtract from total. And don't pretend to not know what the end game of something like YC is here. And YC is not the only one.
One company doesn't need to do everything. But 100-1000 smaller companies, each taking on a piece of every-day work that people are paid to do and replacing it with AI, is definitely doable and can pay very well if deployed at scale. It doesn't have to be perfect all the time either - just cheap, fast, and good enough. As it always has been.
You didn't really think they are training vision models so they can just drive your car around, or that they are training robots to perform physical tasks so they can just use them in factories, did you?
You’re almost right but missing an entire category of startup: “egocentric” data. A lot of the most high revenue startups right now are, essentially, recruitment agencies for frontier labs to get video of skilled workers. OpenAI and others are spending billions per year with the various startups. We aren’t witnessing specialized startups training for specific sectors, we are witnessing training startups trying to hoover as much video as possible.
Personally, I’m not convinced these startups have much life in them. They’re popular right now because they are low effort as there is a huge demand from the frontier labs, but a year from now, there’ll be more than enough data.
There are at least a dozen in YC and probably hundreds in total.
I think a lot of the founders of such companies know that their companies have an expiry date and no viable exit strategy.
Probably the strategy is making as much money as possible while there's demand from the AI labs and then the founder will figure out their next steps with a bit more money in their accounts.
There will not be, they will find that the videos submitted are all of noise quality after the training starts, and then they would need videos of certain, specific actions that were basically not recorded. and it will go on and on for quite a long time.
HVAC tech here. Good luck. There are so many times where I have to reach somewhere and can't actually see what I'm looking for and have to go off feel alone. And I only know what it feels like because I've felt that particular part hundreds of times before. I think there are some things that could maybe work, but they would be incredibly basic tasks that are probably just cheaper to have a human do. In any case good luck.
Noaidi | 6 hours ago
whartung | 6 hours ago
First pane has a supervisor, a plumber, and a robot, with the supervisor pointing to the robot.
Next pane is the robot watching the plumber work.
Next pane is the plumber watching the robot work.
Final pane is the robot with a bloody wrench standing over the dead plumber.
psychoslave | 6 hours ago
praxulus | 6 hours ago
vkou | 6 hours ago
happytoexplain | 6 hours ago
vkou | 6 hours ago
anon7725 | 6 hours ago
happytoexplain | 6 hours ago
dyauspitr | 6 hours ago
happytoexplain | 6 hours ago
Mind-bogglingly casual acceptance of civil war when all you had to do was change the order to put "We will figure out a way to make it work" before "No one has a job".
dyauspitr | 4 hours ago
theoreticalmal | 6 hours ago
npongratz | 5 hours ago
Though I'm less sure about (4), considering the Worldcoin eyeball-harvesting scammers and the AI-bailout scammers are one and the same.
SuperSixFour | 5 hours ago
Maybe next time you should spend the few grand on all those tools, the hours it takes to learn how to use them, and can do it yourself!
rickydroll | 3 hours ago
happytoexplain | 5 hours ago
gusmally | 6 hours ago
happytoexplain | 6 hours ago
FLeXMurphy | 6 hours ago
black6 | 6 hours ago
nekusar | 6 hours ago
Probably should do that before policing/military humanoid robots are cheap enough. If we don't, we're looking at our future as indentured servitude for capitalists for the rest of the lifetime this earth can support life.
FLeXMurphy | 6 hours ago
Out of curiosity who are we killing or getting killed by? What amount of brainwashing will it take to get all these rubes to kill eachother to make the 1% richer?
nekusar | 6 hours ago
Pro-capitalist governments (cycle of corruption choosing the winners, revolving door between politics and business)
Human and robotic military conscripted to fight
Capitalists
Capitalists' workers paid enough to not be "worker class"
The other side:
Union supporters
Workers rights advocates
Socialists/communists
People who lost jobs/poverty class
Workers abused more than "average"
We have seen these types of wars before, and in the USA. Our own worker rights and union history was horrendously bloody. Even had the US military killing strikers and union folks. It's also where the private military, The Pinkertons, came from.
And prior to 1940s with FDR, there was a strong Socialist and Communist groups pushing for their respective changes. FDR appeased some of the demands with his reforms. And those reforms were credited to the success of the late 50's and 60's.
black6 | 6 hours ago
anon7725 | 5 hours ago
I suppose whatever the amount is we reached it a while ago, given the number of people who have died for oil prices or to distract from this or that domestic political situation.
debo_ | 6 hours ago
maitland | 6 hours ago
psychoslave | 6 hours ago
momojo | 6 hours ago
Can OP or founders comment?
pizza234 | 6 hours ago
momojo | 6 hours ago
thomas_unise | 6 hours ago
I’m the owner of this company
What would you like to know?
Just because I’m not in YC does mean anything
I have deep home service connections and have been doing internal ai app development for the past couple years
I think it’s lame people try and downplay a company because they aren’t SF based or in a YC cohort
I made the calls and got the contracts with these home service company who will document all their jobs for me and my team
nemomarx | 6 hours ago
thomas_unise | 6 hours ago
We are data only. So we have no robots. We capture the entire job and structure the data for training
Our model is to capture hard to get training data for robotics
emdash | 5 hours ago
short_sells_poo | 6 hours ago
That is not to say that your idea or company are bad, but there's very little substance that would indicate that I should care.
I don't care that you trained your own deepseek derivative or that you got contracts. What you are proposing to do is interesting, but a mission to Mars is also interesting. If it's some random person proposing a mission to Mars, I don't care because they have no chance to actually achieve anything meaningful. If it's a person who has credible support and a plan, sure, now I'm listening.
Why should I care about you doing it? How are you proposing to solve the hard problems, which are in the robotics space?
thomas_unise | 6 hours ago
But I am doing what everyone does on here and sharing what I’m up to.
I’m not trying to say I’m special or anything like that.
This just happen to resonate
And I just got this new brand rolling out which is why I have no presence
short_sells_poo | 5 hours ago
I'm not trying to diss anything you do, kudos for trying!
I generally found that the HN crowd is difficult to please and perhaps a better avenue of getting feedback (and support) is to tailor material specifically for HN.
E.g. if you had a blog post talking about what you are trying to do, the problems you encountered and interesting solutions you are looking at, you'd get a much more interesting response? We get continuously assaulted by "the next big business idea" all the time, both here and in our professional lives. It's difficult to stand out in that deluge. Talking about the technical side can be an easier way to engage people in constructive discussions.
thomas_unise | 5 hours ago
I didn’t expect this kind of response
Will get more on the blog and pushed to socials
happytoexplain | 5 hours ago
Sorry, but how?
ranger_danger | 6 hours ago
thomas_unise | 5 hours ago
I think you might have misread my tone?
I was just trying to clear up anything the OP was curious about
Nothing was intended to come off as harsh
fxtentacle | 2 hours ago
In my experience, videos from professionals are not enough to train a robot, because you lack all/most of the edge and failure cases. Let me illustrate with pilots: The average Airbus can land itself via autopilot. Autopilot will fly it to the correct location. You confirm the frequencies, it connects to the airport's guidance system and lands itself. Super boring. Videos of that won't help anyone become a proper pilot. The reason pilots train for years is to handle the 1 in a 1000 case where the autopilot doesn't work. And you suddenly have to take over in the middle of a messy approach.
Translated back to your plumbers, the real training is when the robot is carrying a 100kg metal pipe and then the owner's kids come running and pass through under it. Every professional will give their best to immediately reduce efficiency by making things stationary and SAFE. That's the training data that is truly valuable.
BTW, I am considering bootstrapping a startup that uses video game engines to generate training data to make robot policies anticipate physical effects and counter-steer before they happen. Like putting down a heavy steel pipe before it drops due to gravity. With 2 bugfixes merged into the Unreal Engine, I have some experience there ;) So I might soon be in the business of selling custom training data.
Feel free to email me at hn at hajo me
jawns | 6 hours ago
But what is the incentive for contractors to participate, and how does it overcome the enormous disincentive of "You are training your replacement"?
My guess is that truly skilled contractors in these trades are not hurting for work. So I doubt they're going to be swayed by just a few more extra jobs added to their docket.
Less skilled contractors who are desperate for work might be taken in, but then the quality of their output is likely to be worse, and the value of that training data is likely to be low.
My guess is that people in skilled trades who are used to working with their hands are going to be even more skeptical than white-collar folks about an AI revolution, and more guarded against change in that direction.
FLeXMurphy | 6 hours ago
They don't need an army of contractors to do this. They just need a few, much like the AI labs that have a handful of SME staff for each particular subject to guide the corpus of data. See mathematics example.
It is not about knowing if someone can be bought. It's knowing just how much is just enough to buy them.
And then you have individuals whose pride is more than enough. See Terence Tao for a great example of a Useful Idiot.
clintonb | 6 hours ago
Why? Are you assuming these folks are also good at marketing, or customers are sharing via word of mouth?
I rent. I assume homeowners use AngiesList or Yelp, but those reviews don’t guarantee skill or that someone will be in high demand.
bravetraveler | 4 hours ago
Seems fair to me (interjecting: not the parent poster). I rent, too, but have heard countless conversations along these lines among Proper Adults.
Often grown-up #1 complaining about their latest project and grown-up #2 offering someone that served them well in the past. Point being, bad experiences are both so common and costly... that those who are good tend to stand out and stay busy. IME, those listings or the phone book are most suitable when you don't have a relationship to leverage (effectively rolling the dice).
happytoexplain | 4 hours ago
gruez | 6 hours ago
Seems to be working fine for AI labs and their army of data annotators (for generating training data and RLHF). Is there any reason to robotics is extra tricky?
jerlam | 5 hours ago
If I was confident that an AI couldn't do my job, I would happily sign up to do this "training" because it's free money and my job is at zero risk.
colesantiago | 6 hours ago
Creatives and all white collar jobs are the easiest to target first, with AI labs basically pitching individuals and businesses:
"AI will free you up with more time to do 'other meaningful things'"
If you find yourself laid off or work has run dry from a previous gig because AI can do it better than you, you'll naturally consider a role which AI hasn't touched...yet
Then you see AI and robotics labs going after the margins of blue collar jobs, which then they will also say again to the blue collar workers:
"AI will free you up with more time to do 'other meaningful things'"
I can't imagine what the world will look like if you repeat this N times, or even in 10 years.
Because you need to pay your rent, and bills right?
Unless rent, bills and all the rest of it doesn't mean anything anymore in this 'AI abundance' age.
FLeXMurphy | 6 hours ago
One company doesn't need to do everything. But 100-1000 smaller companies, each taking on a piece of every-day work that people are paid to do and replacing it with AI, is definitely doable and can pay very well if deployed at scale. It doesn't have to be perfect all the time either - just cheap, fast, and good enough. As it always has been.
You didn't really think they are training vision models so they can just drive your car around, or that they are training robots to perform physical tasks so they can just use them in factories, did you?
happytoexplain | 6 hours ago
reticulates | 6 hours ago
Personally, I’m not convinced these startups have much life in them. They’re popular right now because they are low effort as there is a huge demand from the frontier labs, but a year from now, there’ll be more than enough data.
There are at least a dozen in YC and probably hundreds in total.
elzbardico | 6 hours ago
Probably the strategy is making as much money as possible while there's demand from the AI labs and then the founder will figure out their next steps with a bit more money in their accounts.
Aperocky | 5 hours ago
happytoexplain | 6 hours ago
SuperSixFour | 5 hours ago
rickydroll | 3 hours ago
https://s3.us-west-1.wasabisys.com/luminist/SF/AST/AST_1957_...