How Much Will AI Robots Cost?
ฝัง
- เผยแพร่เมื่อ 14 มี.ค. 2024
- How much will the Figure.ai robot cost? We take a look at some other robots and try to work out some pricing models, and also how financing will affect how businesses decide to adopt these robots.
Table of Contents:
00:21 - Figure.ai robot
01:27 - Boston Dynamics robots
02:09 - Unitree (China) Robots
03:06 - Financing
Links:
www.unitree.com/
bostondynamics.com/
www.figure.ai/
VentureBeat - Boston Dynamics starts selling its Spot robot -- for $74,500
venturebeat.com/ai/boston-dyn... - แนวปฏิบัติและการใช้ชีวิต
When mass produced will probably cost the same as a family car
Less. Family cars use far more materials (2+ tons worth) and modern automobiles have just as much tech as a robot will have. Modern vehicles ARE robots. Not autonomous, yet, but they are machines that make human labor (travel and hauling) faster and more efficient.
Once they find the "Model T" of robots and mass produce it, costs will plummet quickly as competition erupts all over the globe.
Except for one small (lol) problem, AI and robots will put so many people out of work so fast, that demand for affordable robots will be undercut by the fact that the majority of people no longer have sustainable incomes to afford them.
This ends badly in every scenario where our entire economic system is not completely restructured. Unfortunately, I see no way this happens without colossal destruction and pain. When human labor becomes essentially worthless for a tipping point of jobs, then only capital and assets have value, leaving the current powers that be in a leveraged position that is impossible to de-leverage without revolution or war.
My bet.... You'll be able to lease these in the 2030s for similar costs as a mid to high end car. A car you'd use for a sales team. However there will almost certainly be a labour tax on them to support UBI or a transition unemployment benefit system. This will slow market penetration of AI robots which will be necessary initially.
Whilst the LLM needed for linguistic interface and vision systems will be computationally demanding. This probably makes large-scale roll out now unfeasible. Improvements in the AI architecture and training will make increasing market penetration possible in the 2030s. Think about the growth of Internet capacity that's enabled mass streaming of video, unfeasible in 2000.
If true, that would be about $4.50 power hour. But estimates to date have been $1.50 per hour. So that would be more like $16,000.
Yeah, apparently figure.ai CEO is targeting $50k. Mostly the $500k-$1m figure is to kind of establish a top-end figure, esp when capturing early "enterprise" sales.
Dont forget the subscription service
This channel is like a breath of fresh air from all the hype and BS in the media. Really dig the nuanced and grounded perspective and I like how you tie it back in to reality and economics of it all.
The CEO of Figure said that he expects that they will eventually be able to get the price down to 30-50k within a few years. The next few decades are going to be crazy...
Not quite sure how that fits with the costs for materials (assuming he's talking about building the robot with the specs listed on the site). But maybe he's anticipating building the whole thing with Figure robots from material extraction...?
Just realized that there's a new test for a robot - bootstrap. Kind of like boostrapping a compiler. Can a robot completely assemble itself. Hmm...
Pretty sure Figure 01 brain is offboard in an air-conditioned rack of GPUs with a link to the robot itself. So, no onboard "brain" is a weakness not a strength. Please correct me if I'm wrong.
@@zvorenergy nobody knows! Hence all the skepticism around the demo. But their stated goal on their homepage figure.ai is fully autonomous w/a 5hr runtime. The specs are huge in gradient fonts so you know they are serious. ::cough::
FWIW I have been very impressed with the results I've gotten from LLMs running on my Mac Studio. lmstudio.ai/ there are a ton to play around with, many/most are a bit more focused eg coding than the full ChatGPT. Macs are nice because you can load the whole model into memory and it's all on a single board with a fast bus. I have a PC w/a 3080 and IIRC 10gb on it and that means I'm limited to a ~8gb model. Vs my Mac Studio w/32gb of ram will quite happily load and run a ~20-25gb model just fine.
I don't have an extra $6k lying around but folks are reporting solid stuff from the maxed out Ultra...
@@ChangeNode Thanks.🤣 Sorry I was just thinking 5 hours is exactly when a worker is required to take lunch
@@zvorenergy no worries. I have so many Qs, like if it's 5 hours how long to charge? Can the robot repair itself, or do you need two to do that, or...? On and on and on.
In another comment someone mentioned something about price and it occurred to me that a solid test for a robot like this is bootstrap, kind of like a compiler bootstrap. Can a robot like this (given parts) build itself? How much of the production supply chain can it do (eg mine, process, etc etc etc)? The math gets really weird if these things can get to the point where they can replicate themselves.
I really like your calm energy and thoughtful content. Subscribed.
This is the most calming AI / Robotics video I have seen ever.
Hey man, just ran into your channel. Thanks for putting out great content!
UBI video pre-thoughts:
Don't do UBI. Instead, expand employment insurance to include training and roll in a government run temp-worker agency. Everyone, anyone that wants money/training anyway, is in the system. Everyone has an employment status, a list of certified skills, and an employee rating. You work, you earn "training time" that you can take whenever you want and train however you want (or get shoved into if you're laid off). Earn new certifications, go to university, travel, or just sit on the couch watching yt, your choice. You are never unemployed; you are either working or on training. That work to training ratio can be adjusted over time as more and more jobs get automated. Maybe now, it's 12 months work to 1 month training. By the end, maybe that's reversed.
That's a pretty easy transition so far, with the benefit that you're not stuck in a job you don't like. Work a few years, then train for something else. It's baked into the system. Next, is to connect the people that have no skills, don't really want to put the effort into new skills, and don't really want to work anyway... with the jobs that nobody really wants to do but still need doing. For this, I propose the "work registry."
As an employer, you have a job you need someone to work at and don't want to bother going through the hiring process, so you just go to the work registry and bid on worker skills/ratings the same way people bid on Google Adsense words. Higher the bid, the more the worker gets paid, give right of refusal to the employee as well. The system tells the workers you "win" where they need to show up. If they show up, you pay the Registry like any temp-worker agency. If you like the employee and they like you, you can agree to extend the job term. Otherwise, at the end of how long the work term was, you punch in your rating of the employee and the employee gets to rate you as the employer. If the employee doesn't show up, they get a zero rating and you'll probably bid higher the next time you want a worker. If the employee rates you lower, then the Registry adjusts your next bid so you have to pay more to compete against other employers in the auction.
But, here's the UBI part. Even if you don't show up for your assigned job, you still get paid and you still earn your training time. You just get a really poor employee rating and that means you only ever get assigned to the worst paying jobs. There's a floor... effectively social security. But, every day, they keep sending you job assignments and every day you get to decide if you want to start showing up, earning that employee rating, and start earning more. Any day you want, you can start. Most likely, some government make-work job that nobody even cares if it gets done, but a job. Or, maybe we'll start "funding' volunteer agencies with paid workers. No barriers, no effort, just start showing up.
Eventually, everything could get rolled into it... social security, disability, retirement, even set it up to make self-employment an easy option. And, as the needed jobs start reducing to zero, the system will seamlessly continue.
And yes, I keep re-writing this and posting, refining as I go.
You are getting close to some of what I'm hoping to cover, in particular trying to work out a rough model for a transition. Esp how this stuff will show up in macroeconomic data and when the #s would get crazy enough to more-or-less require changes vs theoreticals.
The training bit is interesting, we already have a lot of training systems in place and AFAIK most unemployment programs (at least in CA/WA where I've lived) have very explicit programs in place for retraining. Would be interesting to look at how that's worked for folks. I have a friend who transitioned from color prepress/printing to tech recently and it was rough as hell but after ~3 years he's now got a decent job... as a PowerApps lead dev. Very interesting times.
There will be a subscription/maintenance fee. Maybe $300/month
Hey Will, this channel is a hidden gem, thanks for the awesome content!
Great work - look forward to seeing more in the future
Thanks! Any particular topics?
It would make more sense for these companies to charge a yearly salary for these robots with a update / upgrade path. More like a lease structure.
the dreaded subscription model
Well said! I've been of the impression that families will invest in a box that lives at home to do llm inference to keep things private, which might evolve to being the brain for the family to also run the robot custodian. That's only if it's cheap enough to make sense tho. If the jobs are displaced and nobody can afford food, not to mention a house robot then that sucks. I don't think there's a job that's untouchable to AI, but that doesn't mean work will end, it will just force a change from butt-in-seat-to-get-paid office work. Upskilling won't mean years at school anymore, since llm teachers can answer any question at any location. Not sure that there's room for humans to pair program once the Robots are smarter than humans instead of just breaking parity. Seems robots can do it all, so it's more about finding the best way to ride the transition until they suck up everything. The attractor state, good or bad, will be decided during the transition.
Great video!
'In The Year 2525' is now playing in my head. Thanks for that.
lol never heard of that one, looked it up and I don't think that one got a lot of play in the US?
@@ChangeNode I first heard it as an opening song in a movie. It was about the Vietnam tunnel wars. IIRC it was called Tunnel Rats. That flick was uber-nasty.
As mass manufacturing takes over it cost will trend towards the cost of raw materials + energy + labour + profit margin. I think if Tesla can sell them for $10k that will sell a billion. On the other hand if they are $100k they will sell 10000, if it is a $1mil they will sell 1000.
I think the leasing model will end up being a more common option, specially while technology and designs aren't stabilized. Nobody would like to invest a hefty sum in robots that become obsolete in a couple of years.
Jeez... when AGI comes to life I'm totally going to want one as a little friend.
bro we know what you really want.
@@masterroshi8812 Yeah....little spoon at last.
@@masterroshi8812 Hee hee!
Keep believing in fairy tales.
@@dallassegno Either AGI is possible, or the brain is magical. So ye best start believin' in fairy tales, because either way you're in one.
Great video and the economics is crucial.
Absolutely. Kind of think that my next few videos are going to focus more on the econ side. Any particular Qs there you are interested in?
Nice, found a new TH-cam gem.
I kept thinking about cost, as well. Loved your thoughts on financing. Certainly got my gears grinding. (Pun intended) I'm sure that more than a few of us here in the comments have now started running the numbers re: various business models...
I think they need to cost around $10,000 USD for them to be adopted by consumers (middle-class).
Though I'm not sure about the chinese version, I'm fairly certain Spot is made to order, i.e. custom made. The same type of, or more, markup that you get from ordering e.g. custom furniture (around 10x). My guess would be that the chinese copies are made to order as well, and at least they take a hefty margin on them.
Cool channel, good voice, looks cleanly edited.
Thanks!
WRT the model from China, I thought it was interesting esp as more or less cheapest possible version of the hardware. I would assume that there's a minimum 50% markup over COGS on that one and likely at least that or higher for BD. In another comment someone said that the figure.ai CEO is targeting ~$50k.
I'm wondering if there is a version of a robot bootstrap challenge, basically when and how far can figure get to building a robot that can successfully build another one of itself, and how far that can get pushed through the entire vertical chain.
I might get one of these robots as a companion. Keep it in my office . It can brush my suit for me and tell me to think positive.
I predict that the Chinese humanoid robots will be super cheap! Compare with Unitree's robot dog Go2 which is pretty advanced and yet at a very low cost. Also I expect many robots to be wirelessly connected to computer clouds doing the heavy AI lifting.
Unitree is very likely fake. The Chinese are known to put out fake propaganda videos like that claiming to have advanced tech when they don't. If you closely watch the Unitree promo video, you can see it is fake. They are copying Boston Dynamics. It does not have a real robot with capabilities like it pretends to have.
the F01 bot runs an onboard framework that converts text from GPT-4 API to robotic actions. It is important to specify this when considering cost and deployability: the GPT-4 API costs money per query, and the connection requires lightning fast high fidelity internet to maintain that 3-6 second delay on action throughput. It is massively impressive as a chassis and the underlying models will only get better and cheaper over time. I don't think the current GPT-4 cloud infrastructure can handle deployment of this at scale.
Have you played much with local LLMs? lmstudio.ai/
My guess is that they will be able to put the core brain onboard, likely would have to anyways for latency. Which is why IMHO the chip wars are a thing.
@@ChangeNode I've done some projects with the llamacpp repository, which is one of the precursors to lmstudio. It definitely can run an executive llm model as the "agent"/brain on device, I would say the new mac studio can fit a 70B pure density model or a 2x30B MoE, or any denomination of 70 billion parameters, with usable throughput.
It is really a question of how good does the llm piloting the bot need to be for the onboard neural network to correctly parse the instructions.
JAH BLESS, I am a self thaught programmer working at a startup without a degree and I am pivoting to robotics since tech lead said that software is like titanic xd. We have developed heavy ai but the real world applications depends on physical emobdidment of ai
Curious what you think about Devin from cognition ai
At this point I have a request in to check it out so waiting... ;)
There are some existing local LLM tools that do similar but via CLI and not as slick.
Anything in particular about Devin you are curious about?
Awesome its hard to find rational people on TH-cam this guys good
I wonder if the robot can be smart enough to go through a computer with you, look at files, browse the internet and help you scale a current business and then can also work on your laptop 24/7
100 odd grand is the past..
Robots will be relatively cheap as chips. Vertically integrated Tesla will produce its Optimus for 10 grand. How much to recoup R&D and whether they sell or lease remains the question. But in a new competitive field, Tesla could hypothetically sell for $20,000 for 100% return.
Good luck to the rest,
The market will be big enough for niche uses, leaving plenty of opportunity for everyone.. but not at the margins you think.
People blindly state that whenever there is disruptive innovation, new jobs emerge and this will be the case with AI and robotics. However, there’s never been technologies created to replicate all human tasks and think autonomously without human direction, so where will the jobs come from other than a few supervisors of the tech and those with capital/business owners and directors
Yup, that's one reason why people are so concerned. Even if some kind of new jobs come around, they might only require a small % of folks. And even if those jobs do exist, it might take decades for those jobs to come into existence, with potentially decades of turmoil in the meantime.
Very thoughtful. I do remember being pretty sure that most trucker jobs would go away with autonomous driving. It's been ten years and that still hasn't happened.
Yeah, no kidding. Self driving. 3d printing. VR. Crypto. All of these things (except crypto ha) IMHO are nice that the keep moving forward but also haven't exactly lived up to the hype. Which is why I have a lot of skepticism but... dang. That figure.ai demo is pretty wild...
Watch nvidias new chips. Youre not ready for the futurw.
Just watched the latest w/the Blackwell data centers. Very very interesting.
The Optimus robot being developed at Tesla has been costed by an industry expert at about $10,000 cost or less. Elon has said he expects the robot to be sold for $20-$25 thousand. Exactly what the software will cost isn’t know but the marginal cost of software is zero so, in large enough quantities, the software cost might be fairly low ($2-$3 per hour). If the robot lasts about 10,000 hours (2000 hours per year for 5 years), the cost would be about $2 per hour. Even at $5 per hour, the robot would be a no brainer for business so long as software exists that can do the job.
Even if the government added a $5 per hour tax to help pay for UBI, at $10 per hour cost, it is still a no brainer for any business that can get software that works.
I'm editing the video on UBI I'm working on right now. spoiler: IMHO it's going to be a central bank/monetary policy model, not traditional fiscal. Deflationary math doesn't work out otherwise.
@@ChangeNode I believe we will have a UBI and I know how to pay the huge bill. People who are on UBI would also benefit automatically with the overall health of the economy without any begging from the politicians. I also have a major in economics from University.
All I care about is when the Butler version comes out. All this common sense and economic realities thing is all fine and dandy but that doest do my laundry.
Plus wash the dishes and put them away. And dust.
So far I like your content. Isn't it more economical to just develop task specific robots . I have worked in factories and wearhouses. And many tasks only use a small percentage of human motion. If you use robots and mass produced parts to build and AI to help design the robots. The cost is dramatically lower per robot.
As an example take a vacuum cleaner robot.or a go pro camera drone . Or the newer thing in the food you tube space the chefmaker deep. Some ai robot cooking device.
Or in a production environment a robot sewing machine
tl;dr absolutely, but I think the idea is that there's also likely some point where it would make sense to fold in a humanoid robot. In another thread someone said that the figure.ai CEO said the goal is to get the cost down to $50k or less. Plus the idea that a SuperRoomba with an articulating arm might be a lot cheaper. In that scenario the idea presumably is that you can train it once via demonstration and then it just goes, 24/7.
Crazy times.
The production of one bipedal humanoid robot, similar to the Optimus/Figure type, on a mass scale would incur material costs equivalent to one average car. This translates to approximately 30 to 33 robots. When upgrades and services are factored in, consider the financial implications if a small production line were to manufacture 1,000,000 units annually by the year 2027. This scenario presents a significant opportunity for market leadership, potentially by China. Now you can add your math figures, I say we are talking under $9,000 for basic unit.
Yeah. The big one IMHO is if they can get to bootstrap as much as possible vertically. If figure.ai bots can get to the point where they can assemble themselves, esp if they can vertically integrate the resource extraction...
fresh graduate from germany. the situation is grave for robotics software engineer if you are looking for entry level jobs. God!
:(
Honestly everyone I talk to seems to be pretty worried about jobs right now, kind of regardless of field.
Are you only looking in Germany, the EU? Thought about something outside of EU?
I like this channel.
I think we should not only consider humanoid robots. Specialized robots are taking people jobs in manufacturing for years like in automobile manufacturing. The current trend is using robots in kitchen at fast food restaurants. Once all mechanical are in place all you have to do is change the programming to add AI.
Check out my last video, talk a bit about non humanoids and industrial 👍
If AIs can recode themselves (per rumours out of OpenAI) there’s no reason they can’t reconfigure their physical form.
There is a labour shortage so no this a terrible idea.
And get garbage food that's even more garbage. Can't wait.
Robot bartender selling now is $250,000. Would replace at least 6 coffee house baristas (more if your hours go later or 24 hrs). So that would pay for itself in about 2 years easily.
How much would it cost to charge them and how much electricity they consume? Would we face battery shortage if we start producing lots of these robots?
Or perhaps they'll use something like wireless charging? e.g. lay out a wireless charging mat in the workspace. If the robots are on the mat for 30% of the time, they may be able to maintain their charge while working.
15c per kilowatt hour. Based on laptops and electric Appliances, probably about 300w/ hour.
I would imagine the electricity cost would depend on the mass moved. But likely pretty cheap compared to alternatives.
WRT battery tech depends on how many. I'm very curious to see if/when figure.ai can use a figure.ai robot to build a figure.ai robot...
What I'm interested in is home helper/companion models. What would an average consumer be able/willing to pay? (I grew up with The Jetsons. I want my flying car too!)
Could regular people buy a robot to start a business providing a service or making a product? This could be a solution to the employment problem.
The figure.ai CEO says they are targeting a $50k price point. Japan has been experimenting with caretaker robots, could try looking up those. $50k new ~= to a new-ish car. After a few years could see them getting cheaper used.
There are some flying cars coming but the traffic + having them drop out of the sky... oof.
Based on the parts I'd guess $40,000 would be realistic in the next two years, but supply will need to catch up to demand before the price drops down to there. The idea that they will displace workers as free labor is not realistic. Governments are going to start looking for a way to tax robot labor as soon as there's enough of it to tax. I would expect that they will tax robot labor enough to implement UBI so they can pacify the unemployed masses. The mentally challenging jobs will be safest in the long run since processing and power usage will slow the replacement of those jobs.
The dog bot is $80k USD. The humainoid bot is $100k. The lithium battery is completely encased within the torso. And in case you didn't know, they wear out after three years. As an inference, the manufacturer(s) are seriously expecting to just provide these machines to just billion-dollar corporations if not just Amazon only. Furthermore, they're are hoping their client(s) to just fling a few millions on a semi-yearly basis. Because, it's a perpetuation of the Ford design-to-fail mantra. Which has now been concreted and instilled across multiple domestic enterprises in North America for many years now. Why do you think the manufacturing sector is entirely owned by China now?
currently there is no competition and no high volume production in robotics, that is why the prices are high. I'd suggest you to compare robots production cost with electric car production cost. I do not see how robots could be more expensive than electric cars when produced in big volumes.
so, I think after production picks up, there will be a range of robot models with range of prices from $50K to $200K in about 2-3 years, going down to as low as $10K for cost optimized robots, produced by other robots - in 5 years.
Yeah, another thread someone said that figure.ai is targeting $50k. My guess is they will start very high and then drop over time.
Impressed by the quality of your reasoning. Thank you for sharing your thoughts. And regarding your potential next topic about UBI I just wanted say that in my opinion a job guarantee as suggested by the MMT community is a far better and more realistic solution than UBI. Social jobs or jobs in the healthcare sector which we may still not want to be executed by robots could be guaranteed by the government for those who struggle to find a job. Maybe you find the time to do some research about this topic as well and share your thoughts together with your thoughts about UBI. Thank you
Given the sheer number of unemployed people we will have in the future, I think that a jobs guarantee will simply lead to the creation of a vast number of pointless bullshit jobs. The only thing worse than doing a job you hate is doing a job you hate that you also know is completely unnecessary.
I'm firmly in the UBI camp on this, but I would be very interested to hear opposing views. Could you recommend any videos that would change my mind?
I'd add that there's also talk about "UBS" - universal basic services. So the gov't can provide, say, food and shelter. (One way I envision this is for the gov't to buy farmland and let robots run it near autonomously.)
@@marcusmoonstein242 I think wether we do bullshit jobs or not depends mainly on the capability of future governments. There is a potential for bullshit jobs. But according to David Graeber that is already the case for 20-50% of the workforce. Besides that according David Graeber as well the most meaningful jobs are often times also the ones with the lowest salary. Governments could provide money for jobs in environment protection, social work, education, research or renewable energies and revalue meaningful jobs which we would like to do but for which there is currently no money.
UBI bears the risk that it only leads to higher rents in the cities for example and therefore misses its actual purpose.
@@andreschlinke3392 Option 1: here's $1000 per month forever with no strings attached. Option 2: here's $1000 to do a job you may or may not be interested in, and that you have to do five days a week whether you feel like it or not, but luckily some government bureaucrat decided it's a "meaningful" job.
Give me option 1 every time!
PS: a UBI might actually lead to lower rents in cities because many people go to cities to look for work which increases housing demand and prices. With a UBI people can live in cheaper rural areas if they want to, which could lower housing demand in cities.
I've been nerding out on this stuff for a while now. I tend to prefer a UBI to a jobs guarantee because of this chart:
fred.stlouisfed.org/series/CIVPART
Basically the labor participation rate has been remarkably stable. I think the jump from ~1970-2000 is mostly due to women entering the workforce, with the decline 2001-2015 being very interesting. I think that UBI would only be politically feasible if that number dropped significantly (eg to 55-50%). I do suspect that AI/AI robotics could actually get us there but the timeline is very much up in the air. eg is this going to be like VR/self-driving cars/3d printing or more like the adoption of the Internet?
My sense is that there are a ton of people that are unable and/or simply shouldn't be working for a variety of reasons, and more economic certainty would be *huge* for them. Job guarantees would likely hurt traditional employment by focusing specifically on those who could work but can't find jobs vs supporting everyone. Personally I like the idea of a UBI as a floor + a universal health care + enabling more people to take on more gig/part-time work as a way to both encourage labor fluidity + provide a social floor. One thing I'm not a fan of is how some of the more recent govt programs have made companies move away from part-time work toward FTE. I get it, it's a way to force the companies to provide benefits but it really messes up a lot of people that might want a part-time job (eg so they can be home in time to take care of the kids).
I think MMT is very interesting. I personally don't necessarily connect MMT to job guarantees - I am more interested in MMT as an alternative to traditional debt modeling esp as a method for inflation/deflation regulation.
My degree is actually Political Science, and one of the things we covered that I found compelling was the notion that simple, easy to understand universal programs tend to be both more popular and durable than targeted programs.
My suspicion is that the end of this will be a) a UBI and b) managing inflation/deflation vis a v the UBI will be added to the central banking responsibilities, likely paired with a VAT. That would give central banks another tool beside massive piles of debt to manage inflation/deflation. It's all pretty wild as modeling it involves a defacto unit of $X trillions. Which is why I think it will be a break-in-case-of-collapse thing after that labor force participation rate starts to drop precipitously. Part of what I want to do is do a video that sort of walks through all of this. Just have to finish the script lol
The potential for robotics is amazing, but in terms of job losses, AI alone (without robotics) will replace a lot of white-collar jobs. As for blue collar jobs, I suspect that we'll see a lot niche "smart machines" replacing jobs before humanoid robots. e.g. self-driving tractors.
Yeah, this is something I touch more on in the pivot to robotics video. The humanoid robot is kind of a last step - the industrial automation side is clearly the in-between stage.
What about Tesla optimus? I can imagine that will be a lot cheaper to build
AI Explained, another yputube channel, is reporting that Figure 1 will sell for $50K-$150K. That's a wide margin, and I have no idea how it could be that inexpensive, but that's the price point they said the company would be selling it.
My guess is that they will start at a much higher price point and then eventually get it down. But if they announced, say, full GA for mass-market availability of a humanoid robot with the specs on that site and the brain of whatever, say, ChatGPT 5 will offer... the reaction will be interesting to say the least.
IMHO the big one will be if figure.ai can announce the robot version of a compiler bootstrap - how close can they get to fully vertically integrating a figure.ai self-assembling bot? That starts to look like geometric growth constrained only by resources.
Wild times.
These numbers will never work out here in Africa where you get a housemaid at $20 a month. Even at professional level, you get a college-educated able-bodied person at $300 a month. Try get a robot beat that price, I dare you!
Yeah but I don't think I can buy a person from Africa.
@@gagepoler1984why buy? you just buy their time.
@caty863 were can I buy/rent this "college-educated able-bodied person at $300" do you have a link?
Not so much how much they cost it's how much it cost to maintain and keep the things running it will be too expensive
They would have to be like $5000 for the regular middle-class American to afford. Even Message that they would they was thinking that it would cost like $10,000 so I can see that they will be lower as it is the market and everybody starts competing they will be cheaper.
On the manufacturing side yes I can see them being more expensive but for their point of view, they’re saving a lot of money by going with robotics insurance sick vacation time
Elon is on record as saying if they can get the encomies of scale right they will offer Optimus for 10K
On an institutional level It'd be important to start early gaining some experience in how to provide a substantial base welfare because the US is very much behind on that along with constantly being on the edge of social unrest or even civil war
I don't think the average American is about to engage in civil war. We have some extremists, but they are not even a percentage of the population. I agree with needing plans for a universal basic income of some kind. That will cause issues because those same extremists trying to overthrow the government will lose their minds over UBI.
Again - Jacque Fresco predicted in the 80s it would eventually be the case with robots and automation, and left us all in 2017 with something much better and robust and holistic than a "band-aid" solution like UBI - Resource Based Economy.
I think if you eliminate human mental + physical labor the fallback would always be something like a resource based economy. That's part of what's underlying the minimum cost stuff in the video - the atoms will still fall under at least some kind of economic scarcity modeling regardless.
The challenge underlying a lot of the anxiety I think is sorting out how much of that kind of economy winds up looking like, say, Norway at one end and Saudi Arabia at the other end. For what I think are pretty obvious reasons there's a lot of concern there.
You have to take into account that consumer goods are going to get ridiculously cheap too.
Yeah, it’s all deflation all the way down. I touch on this & central banking a bit more in the UBI video
Everything..
I am going to bring Gary vee with me to a garage sale and get one for pennies on the dollar
Actual adult unemployment in the US is around 23%. It's fairly remarkable how statistics can be manipulated to paint a more rosy picture.
Yeah, you can look at several different slices of the data. The convention is to use the lower definition for comparables. If you use the higher number w/the other data slices it shocking. To really dig into it IMHO you have to be consistent w/the reporting and that's a level of nuance most don't have (same with a lot of economic data, from money supply to debt etc etc etc).
Easy answer. $18k. Humanoid capabilities can be build for under $6k and triple it for dev and gpu costs. Source I am an AI for robotics founder. Disclaimer: I am biased.
So COGS is $18k, with markup for distribution, retail, support etc closer to $50k?
Figure & Nvidia might sell it as low as $2K to gain market share!😊 Tesla said $20K! Kelper maybe $10K. Boston Scientific still in the $100K+😂
there will be many options for less than 500k
Yup. Makes the math for jobs even that much harder. 🤔
There’s going to be some sort of tax related to sending people out of work. Don’t be surprised.
when will they be able to tell us when i can give a robot a hug.
Check on eBay... :)
th-cam.com/video/Dkr_4RICL9M/w-d-xo.html
UBI
Too expensive for people to use they are not quite there on being able to do anything really useful
Consider becoming a robot repair technician!
They will repair themselves and each other.
You are definitely not a financial person. Your hypothetical numbers were really super wrong even hypothetically, it's way off.
But the content is interesting and the numbers economic impact as these robots get more capable will be daunting. Thank you for your hard work in putting this out.
Oh, yeah, I'm just trying to outline some very very broad numbers for rough scale scope without completely freaking everyone out including myself lol.
If you have any thoughts on what's right/wrong do chime in, curious as to thoughts...
I'm expecting these robots to cost about the same as a car once they're mass produced, say about $30K - $50K per unit.
In another thread someone said that the figure.ai CEO is targeting $50k. My guess is they will start out a lot more to capture $ and then they will drop over time...
I told my child that he was lucky because he will soon be able to move me in to the auto nursing home where 99 percent of the care is provided by the mechanized building with one or two humans on staff for insurance purposes. Allowing him to spend his life in the meta-verse, with a personal mech-butler at his beck and call.
Not sure if you have seen but search for "japan retirement robots" - they are really, really trying to make this happen. Will be very interesting to see how this pans out w/the addition of more smarts.
Tony Seba predicts humanoid robots will start at a cost of 10 usd per hour of work and then a fasr decline to 1 dollar per hour and by 2035 10 cents per hour. Billions will be deployed around the world
Yup.
When I think about the implications it really makes me just want to go for a walk outside. Touch grass and all that.
A humanoid robot capable of human tools to make copies of itself from raw materials to final assembly and programming has a marginal cost of zero. At that point you’re not dealing with a new technology, you’re looking at a new species.
Would still have to contend with the raw materials/energy/resource costs.
Not that it really matters tho as if we hit anything like the full humanoid robot for real it basically completely breaks economics as we know it anyways.
The Price is Right 2045 will be crazy.
Thanks for the analysis. Contact your Congress person tell them no on tax breaks for intelligent machinery. Tell them you will vote them out of office and replace them with a Figure One robot that can craft meaningful legislation. The newly elected Congressional robots won't sleep they'll be available 24/7 to be responsive to constituent needs. They won't need lobbies either. I'm so happy your analysis appeared. I've been contemplating how to stop labor from becoming horse's.
AI explained says the Figure01 robot is up to $150k.
Yep, I saw that too. I was surprised by the low estimates. If they can sell these robots for $150K or less, things are about to rapidly change.
@@mygirldarby yes I was surprised too. I don't know how they arrived at that since the cost of the prototypes is basically all the expenditures of the company to date. They gave a range of $30k-150k, the median being $90k. AI Ex didn't say that was a retail price either, but the cost of the prototypes.
I would guess manufacturing at scale might bring prices down by an order of magnitude. In which case these things will be easily affordable for middle class households and then a few years later, cheaper versions and third party knockoffs for lower income households.
Which gave me a mental image of discarded cheap plastic robots crumpled on roadsides.
I actually think robots, especially in low income neighbourhoods, will be the subject of quite a bit of violence.
My guess is that early access/initial sales will be a lot more esp if they have COGS that make $150k work but they can capture "enterprise" money.
If they can get to full bootstrap, ie figure.ai robots can capture the vertical integration to build figure.ai robots it's a whole new thing. Both very compelling as validation + they get to reap the economic benefits first.
lol, no way for home use or small busness
Someone in another comment said that figure is hoping to get the price down to $50k or less for a humanoid robot. It would be pretty interesting to see how that evolves over time eg with used models.
There was a startup in the office next to use for a while that was doing a beer-of-the-month subscription service. When they were starting out they were really struggling sorting out the shipping side (apparently they didn't realize how heavy beer could get as they scaled), and they absolutely would have been thrilled to have thrown $50k at it to help with the manual labor side. But at eg $500k-$1m total non-starter.
Oh, and I don't really drink much but yes they were very popular with folks in the office lol.
I work so that I can buy the stuff I made. If I don’t work I don’t buy and the place where I once worked doesn’t need robots to make stuff because there’s no one to buy it. 🤔
That's when folks start talking about UBI, precisely because of that doom loop...
@@ChangeNode Yes, but what is UBI, really? Where does it come from? It comes from taxes that I won’t be paying anymore because I don’t have a job. The essence of it is this, people won’t have income so they won’t be paying taxes so the only source for the taxes to support UBI is corporate. They will have to pay taxes equivalent to the salary that they were paying me before they replaced me with a robot. It can’t work any other way.
Separate to these, Elon estimates his tesla bots will come in somewhere not much above 20k.
Technology should be getting cheaper with innovation
I just want to know the easiest way to take one out. I doubt they put the brain in the head, would be too easy to behead during an emergency situation.
If your robot kills someone who is responsible? What would happen to the robot? Decommissioned? Imprisoned? Barred from recharge for some time?
Ask Amazon.
Nope. Unitree
Yeah, what is the tipping point at which the economy breaks because of the layoffs? Ideally, we could eliminate the government jobs first. They're all technically paid by other people's tax money, so putting them on unemployment first seems to make sense in that I don't think it hurts the economy much. If we remove ten percent of the people who actually pay taxes? I don't know if this country could handle the strain.
You can't be serious about government jobs, 😅. They will NOT be the first jobs affected, not even close.
Classically 10% unemployment -> loss of elections for current government, 25% unemployment -> revolutions/complete government failure modes.
One thing that is very strange (which I think we are seeing now) is unemployment is low but it's much more focused on a lot more openings for lower pay jobs. So if we go from, say, 5-6% unemployment to 3% unemployment but it's all w/big pay cuts. Relatively unique configuration from what I can tell, with Japan possibly being a very interesting model/example.
4:50 Actually, first thing I think of is CO2 emissions.. We’re blowing it, and AI hasn’t been helping
Yeah. The most optimistic version is that AI+robots will help with the labor costs for everything from deploying renewables to fixing the grid to upgrading homes etc etc etc. Potentially even making large scale carbon capture feasible. There's some stuff that's even wilder, like AIs moving us closer to fusion.
The worst version of course is that the AIs just wind up burning a lot of energy w/o proving a lot of value other than recreating Blade Runner.
Interesting times! ::ahem::
Great, early retirement for everyone! :)
The sooner the AI and robots take over all of the jobs - the better.
Poverty will be solved. Also 40-80% of people don't like their jobs.
We all need a few Optimus and Figure 01 robots per person and it will be the end of human labor forever.
Unless you want to work and create something of course. :)
This will create a parallel economy fully independent from humans.
The transition period - the next 15 years - that might be tough though.
Hope we will find a solution as fast as possible.
Deflation in prices of goods and services is what we need. From 100.000 to 10.000 to 1.000 to 100 to 10 to 1 to 0,1 to 0,01 and then it's already ridiculous and basically free.
Things should be cheaper if humans are not making them.
Deflation > UBI
This is very much getting closer to the macroeconomic themes I want to hit in the talks about UBI.
Part of my journey was "hey, let's at least establish a, say, $100/month UBI and then actively encourage deflation" and then I started researching deflation. The big problem is debt, and oh boy is there a lot of debt out there. Deflation+debt looks a lot like an economic death spiral, esp during a transition.
My core is that this is going to be a monetary policy problem, not a fiscal policy problem. And that will mean moving from debt financing to UBI with a rough algo based on jobs. The biggest challenge IMHO is going to be how society views "fairness" over time.
@@ChangeNode
I'm not against a small allowance of a $100 a month as long as you can buy goods & services valued at a few hundred or a few thousand dollars in today's money. ;) Obviously thanks to deflation provided by ever cheaper and cheaper price of work done by robots.
Ideally we will be flooded with ultra cheap (and then free) goods & services provided by robots, and then the whole issue of debt and money, and monetary policies will kinda be - forgotten... :)
That's my best hope for it!
An invisible transition would be ideal!
A social upheaval and stress and rioting transition would be disastrous.
I agree about "fairness" btw. People's perspective will be important.
But for beginning, being grateful for free food and housing and transportation etc. - would be a good start. :)
We should still make companies pay taxes for robots, so we can use that money for common wealth in society and redistribute a part of the gain between humans. Or use that money to refill pension funds for humans
Hey! ❤
TH-cam recommended one of your videos, and I must say, you post awesome stuff. I love it, and I want to see you on top!
By the way, I was wondering if you or your friend need a video editor or thumbnail designer.
I am Mohan, a professional video editor and thumbnail designer with 4.5 years of experience in creating engaging and high-quality videos and thumbnails for TH-cam.
If not, please reply with "NO" and I will remove you from my follow-up list.
Keep creating amazing content!
Best regards,
Mohan
Your whole talk is quite irrelevant compared to Tesla saying their target is below 30.000 USD - case closed. Also, the "brain" of the Optimus is not 6000 USD - it is the same hardware that runs Tesla car Self Driving, any higher AI in the cloud is - in the cloud.
Oh, if the costs can get down that low and the robot has full AGI that's full blown singularity territory. The figure.ai CEO apparently said the target is $50k for that one.
The video is more about the top level costs, similar to building a high end car for big $ first and then getting the price down over time. Mostly it's thinking through an upper limit esp w/financing that's economically viable.
“Several decades” you don’t look 100 years old to me. 🤔 😂
Decade is 10years
Century is 100years
And how many is several decades?
@@ukpauchechi
Boom. No reply. Check and mate. Thanks for playing. LOL This guy is great though. His videos are an asset to humanity.
@@Alice8000 can range from 10 to 90.
Thank you. very insightful.
Yes, Robots and AI will likely replace a lot of manual labour now - if so and unemployment rises dramatically - who will be earning enough money to purchase the goods all these robots and AI are making - will they be selling to other robots? Automation via robots and AI only works if it does not destroy your customer base.
Yeah, that's why I did the UBI video that nobody watched (lol). Take a look at that one and lmk what you think... th-cam.com/video/IIVDLCDeZT8/w-d-xo.html
Looks like dystopia to me
It is, but embrace it. What more is there to life :D
For the reference point , Gaza had unemployment of 31% b4 the current war , yea It gets real bad. Thats why all these billionaires making home in private islands
Yup. My rough benchmark is 10% unemployment flips leadership in a democracy, 20% gets rise of fascism/communism/extremism, 30% gets you government collapse/revolutions/etc.
There was a very fascinating podcast I listened to a while ago interviewing a futurist who had been hired by various billionaires to discuss post-apocalyptic scenarios. Apparently none of them could wrap their head around having their own security forces revolt and taking over the compound in a collapse scenario. "But they work for me!"
the cost will not be 1 million. it will cost like a car or even less.
Yeah, that's the goal. In another thread someone said the figure.ai CEO is targeting $50k. The $1m is IMHO for an upper bound of demand pricing. Will be interesting to see if they start making them commercially available at a higher "enterprise" cost at first to maximize margins to start and then how fast they get the $ down.
Aren't these humanoid robots dumb? Many businesses work like systems, where very specialized parts are what really cut costs. Does having a general purpose machine makes sense for most companies? I really don't think so.
EDIT: typo
I tend to agree - I mean, imagine imbuing a toaster with consciousness. th-cam.com/video/LRq_SAuQDec/w-d-xo.html
AI does not exist.
Honestly I wish that we would just use terms like LLM and ML and never use AI. But I'm sort of resigned to the fact that folks use the term interchangeable. ::sigh::