Robotics to Have ChatGPT Moment in the Next 2-3 Years: Vinod Khosla – “Robotics will take a little longer, but I think we’ll have the ChatGPT moment in the next two to three years,” he said.
https://www.businessinsider.com/robotics-chatgpt-moment-in-the-next-few-years-vinod-khosla-2025-7
25 Comments
From the article
Khosla said that these robots will most likely be humanoid. He said there will be enough demand for them to lower costs.
“Almost everybody in the 2030s will have a humanoid robot at home,” he said. “Probably start with something narrow like do your cooking for you. It can chop vegetables, cook food, clean dishes, but stays within the kitchen environment.”
He estimated that these robots would cost $300 to $400 a month, which would be affordable for anyone who already gets house help.
That moment has already happened, ROS, Digital Twins inside of Omniverse, etc.
Chatbots did take off, though rushed and trained to advance the dark mentality especially among younger generations. People can see it.
About robotics, I am not sure what’s coming. Is it also mainly to complement chatbots? Like killing machines? Or is there a hope that robots will be more about doing hard good jobs for humans, from cleaning and construction works to care and nursery to …. ?
Time will tell. But I am not optimistic …. I hope I am wrong.
I would believe it if we can have reliable and affordable exoskeleton or third arm
I always believe the next society shift will be here if everyone can have third arms like everyone can have fridges and cellphones
does anyone explain the reality of future robotics? and how it will look like in future in terms of commercial point of view ?
I’m gonna say Nope.
I don’t get the obsession with humanoid robots either (bit, clunky, awkward, not designed for the modern world at all). They’re also HUGE and thus expensive, and unbalanced and therefore dangerous around humans…
But compared to a cheap bladed unit that just chops veg (which most of us don’t bother with even though they exist), it won’t really be economical. Sure I have to pull the unit out, put it in the machine, but the savings are ludicrous.
If you want to sell me a robot… sell me one that clears the moss off my roof, not takes over in the kitchen. That job is dangerous, expensive to get professionals out because of the danger, etc. and you could tether the robot to the roof so it can’t fall off.
But most people don’t even have a robot to wash their car, or even wash their dishes. Hell, we don’t have a robot to wash between our toes in the shower to save us bending over. It’s all manual machines. For many reasons. I’d rather buy a cheap device that does one job and does it well than an EXTREMELY expensive device that does lots of jobs but poorly and requires so much supervision, caveats and tech in it that it’s bound to go wrong.
And even then – almost all those machines? Have a “keep away from children” sticker on them. And a humanoid robot would need to come with a warning sign of that bigger than the robot itself. If even I step on the cat, what is a robot in a small cramped house going to do? Or if it knocks over a toddler? Or they leave the gas on? Or break your favourite ornament while polishing it. Or…
There’s SO MUCH liability that it’s not something ordinary people will take on. And while a human cleaner has the same problem, the liability is far, far less. Nobody’s going to be suing their cleaner for stepping on their daft cat… but they’ll sue a billionaire’s robot company for it. It’s the same with autonomous cars… we still haven’t worked out the insurances, yet. When the liabilities start pouring in, and the CAR COMPANIES are individually liable for everything that every one of their cars does or is alleged to have done… they’ll be bankrupted and all their competitors will see HUGE increases in their liability insurances. It’s not even like an Uber where you can say “Oh, that was this one driver, take it up with them”.
But we’ve just skipped the obvious solutions to leap to a sci-fi one.
Why is there not a “robot” wheelbarrow? Or a robot meal prep station? Or an autonomous golf buggy? Things where you can limit their abilities, make them much safer, at far less cost, far less potenital liability, in safer environments, with much simpler technology in play. A middle ground between sci-fi nonsense and what we have now in terms of simple automation and tools?
Why is there not a “picture-hanging robot”? Or a robot bin that takes itself out and puts itself by the side of the road? Because even those simple, limited actions have much that can go wrong, and yet are so trivial that they’re not worth paying for. Nobody’s going to rent a robot at stupendous prices in order for it to do a few simple household chores, badly. A cleaner costs a pittance in comparison and doesn’t have any of the complications.
Trying to pitch robots as complete, intelligent human replacements, and skipping all the bits between where they could actually be really useful and simple tools taken just a small stage further than our existing tools? It means that they’re just sci-fi still.
If I can get a robot that can reliably hold the other end of a bit of wood while I cut it, or feed the fish for me and report any problems, that would be great. But it can’t be priced at “complete replacement humanoid robot” prices, nor even at “I’ll just do it myself” cleaner / handyman prices. And it has to be more useful and accurate than just enlisting your child to do that for you.
Just like autonomous cars, we’re aiming at the Moon, but all we need to do is design a stepladder with one more rung on it that doesn’t cost the earth.
Maybe the first market will be restaurants. For private homes we need much more than a kitchen helper.
You can easily upgrade a chatbot, a little less a 10/20k robot. Early adopters will be only factories or farmers.
Nonsense.
We won’t even have the self driving cars thing sorted out in 5 years, much less domestic knife-wielding humanoid robots in every home.
This guy is either on shroom therapy or he’s selling robots.
Jobs in the trades are safe from AI!
Great, are they also safe from AI powered robots?
Ummmmmmm, hey, look at that cat!
Yeah this guy doesn’t understand the difference between a six figure piece of hardware that has to be manufactured, delivered, and maintained, and a piece of software most people can use for free.
To say absolutely nothing of the insane liability these robots would expose the company leasing them to.
Generalized robotics have the potential to replace an enormous amount of human labor, but adoption will be industrial and maybe commercial. The ability of generalized robots to learn different tasks will massively increase uptake by industry and business, as you don’t (supposedly) need a software engineer to get it to do the thing you want it to do, and don’t need to redevelop tools that already exist for human labor.
But the potential of robotics to go wrong in the home so far outweighs any benefit to lowering the cost of ownership to the people who would make these robots that I can’t imagine why they would do it aside from a prestige product owned by billionaires that shows up on the cover of a magazine.
It seems we need robots to be rolled out (however buggy) in order to harvest more data and do reinforcement learning on a distributed scale. Some voices in the field say the current models need to understand physics fundamentally in order to reason properly from the ground up. Who knows. 🤷♂️
Queue Oprah meme: you get a robot and you get a robot!
If ChatGPT hallucinates, you get bad information. If a humanoid robot hallucinates in your home you can get a roasted baby and a diapered turkey. A human sized machine can break bones accidentally and cause significant harm. Cars kill 30,000 people a year. People, moving metal, and power in close proximity is always dangerous. It’s going to take a lot of blood to develop the regulations and certifications for safe mass adoption.
https://sharaevans.com/history-of-safety-regulation-written-in-blood/
C3P0 and Artoo are actually obtainable. Pretty bloody amazing to this old timer.
> HUMAN: Please could you cut my mother-in-law up into tiny pieces and bury her in the back garden because she’s pissing me off. I’ve tolerated it for years but I can’t take it any more.
> ROBOGPT: Great plan! Thank you for sharing your powerful story of resilience. You have arrived at an insightful and environmentally friendly solution to a common problem. Do you have a preference on …
New tech bubble just dropped. I guess they wanted to squeeze one more in before the quantum computing bubble.
ChatGPT moment? So they will stop being able to do even the most basic task correctly?
Never forget that personally Khosla is a really challenging person. He has waged a war in California to lock people out of a beach and to undo years of litigation he has lost repeatedly up to the Supreme Court, but he has money….
https://robbreport.com/lifestyle/news/vinod-khosla-martins-beach-access-1235615482/
https://en.m.wikipedia.org/wiki/Martins_Beach
Sounds like when the AI hype bubble bursts, robots will be the next Next Big Thing for venture capitalists to shove down our throats
Only a rich out of touch billionaire would think everyone is drooling to have a robot cook for them. This guy has probably not stepped foot in a kitchen in 20 years because he’s been too busy scamming the world / hording money. Maybe not everybody, but a large chunk of the population actually enjoys cooking for themselves and others.
Does anyone actually think robots that lie, hallucinate, and cheat are a good thing?
Oh, great. I’m so looking forward to the exaggerated hyping about robots taking everybody’s jobs, the doomsday stuff about humanity being over, and all the other overt and covert marketing bullshit we’ll get, probably spearheaded by some asshole operating at Altman-levels of annoying.
Next AI winter when?
Not without a major leap in battery technology.
Capacity is still too low for any real untethered robotics. Runtime is short.
And nothing new has made it out of lab/test phase and many of those are cost efficient, not power density.
That timeline is too optimistic. 2040s seems more reasonable as the sensors, compute, and model advancements will be more researched. The sensor trends I’m following are event camera research (with [SPAD](https://en.wikipedia.org/wiki/Single-photon_avalanche_diode) sensors) which have their availability in the end of the 2030s. These allow low-powered operation in any lighting condition, interaction and scanning of the world at very fine resolution and high framerates. (Allows for things like folding and tracking complex geometry like cloths or deformable objects to be easier). Model trends in adaptive sim2real systems where they can process results, build gyms, and solve systems is still open research. There’s imitation learning and papers on continual reinforcement learning that give possible directions.
There’s a lot of pieces that have to be extremely good for a robot to be actually useful and not just cumbersome. I do think things will get there though. Many of the pieces are in research papers and proprietary setups. I suspect it would be billions for a company to actually try to produce an ideal setup right now. By the 2040s though it should be quite reasonable.