Why Don’t We Have Better Robots Yet? | Ken Goldberg | TED
Why hasn't the dream of having a robot at home to do your chores become a reality yet? With three decades of research expertise in the field, roboticist Ken Goldberg sheds light on the clumsy truth about robots - and what it will take to build more dexterous machines to work in a warehouse or help out at home.
If you love watching TED Talks like this one, become a TED Member to support our mission of spreading ideas: ted.com/membership
Follow TED!
X: / tedtalks
Instagram: / ted
Facebook: / ted
LinkedIn: / ted-conferences
TikTok: / tedtoks
The TED Talks channel features talks, performances and original series from the world's leading thinkers and doers. Subscribe to our channel for videos on Technology, Entertainment and Design - plus science, business, global issues, the arts and more. Visit TED.com to get our entire library of TED Talks, transcripts, translations, personalized talk recommendations and more.
Watch more: go.ted.com/kengoldberg
• Why Don’t We Have Bett...
TED's videos may be used for non-commercial purposes under a Creative Commons License, Attribution-Non Commercial-No Derivatives (or the CC BY - NC - ND 4.0 International) and in accordance with our TED Talks Usage Policy: www.ted.com/about/our-organiz.... For more information on using TED for commercial purposes (e.g. employee learning, in a film or online course), please submit a Media Request at media-requests.ted.com
#TED #TEDTalks #robots
Folding the laundry at 3-6 folds per hour is way higher than the 0 folds per hour I am currently doing.
Statistically speaking: you are correct. Technically, you’d have a pile of clothes by the end of the week.
Yes ! once robots can roll the perfect joint. , I'm sold 😅
I was thinking the exact same thing. I haven't folded laundry in years. I'd love a robot that can get 3-6 folds done per hour. If left running constantly in the laundry room, it'd still be faster at folding than I am at wearing the clothes.
Yup I don't think speed is an issue because if you can get enough robots for cheap, so that they're doing different tasks passively, speed is a non-issue. But they'll get faster over time anyway
You don't have to be faster than the dragon, just faster than the Hobbit in your party.
For manipulating objects with our hands, we have a set of sensors in our fingers that robots can only dream of. Pressure, temperature, texture, weight, slipperiness... I couldn't pick up a coffee cup if vision was my only sense.
Proprioception is a really big one too
we can actual feel surface variation that we cannot see with our eyes, eg. finding the start end on a roll of scotch tape
1x seems to know touch is an important component of robots. And think about how foundational it is to our own models of the world: vision, hearing, and smell only give you some information about the world, and they can lie. Touch, however, gives you a very accurate and objective model of what things physically are. A blind and deaf person can still understand space very well. At the very least it's another modality to work with. Relying on one sense for everything is not safe or robust. You want a good allegory of the cave that's at least decent, with multiple faculties to model the world with.
But you can pick up coffee cup even when your hand is numb What we really need, is to mimic anatomy of our hands. And then all you really do to grasp things - is just flex muscles. The rest adapts. It's not a rigid system.
@@tempname8263 let's see a robot open a plastic bag in the produce section, how about simply opening a can of peas or an alkaseltzer packet? How about undoing a knotted shoelace ?
Watching this, I realized my cluttered house might actually be a strategic move to keep robots at bay!
Keep fighting the good fight! ✊
I'm here because I work at an Amazon warehouse five minutes from home by car. Just checking out the competition
You are done, maybe try to become a farmer
You gotta go the warehouse that they make the bots in.... but then the bots take over that so you gotta go to the warehouse that makes the bots that make the bots. There's more problems then just AGI that is not being brought up in media. Just creates an infinite recursion problem that only really makes sense if people do not exist at all.
In 5 years the bots will take your job. In 10 years the world will be like a Mad Max movie.
Dont listen to these people the bots won’t take your job. AI Will replace mostly white collar jobs and the economy will crash before physical bots can compete with humans in physical space. But the social order will change the modern era will be over and the warrior class will rise and take control of most complex societies as AI replaces the bureaucrat elite and government official. AI Will be to the first world white collar dude what the musket was for the samurai but the modern soldier will not be replaced and simply gain more expensive weapons and equipment which will take years to master thus creating a warrior class elite. We know this because self driving cars have failed. If AI can’t navigate a city they can’t navigate a battlefield
what do you think?
11:23 That kid has 5 fingers on his right hand (not including a thumb, which might be out of view). And that Rubik's Cube is 3x4x2.
hey, its hard for robots to draw. didn't you watch the video?
Lmao, bro you got that genius level attention to detail
Okay negative Nancy. A stormtrooper bumps his head and a Starbucks cup sits on a table in game of thrones. And you expect A.I to not make mistakes?
@@TheMrlittletooth I expect TED to not use AI. I don't care how good or bad AI is, I don't want to see AI art here.
Thats an "F" for this presentation. He was doing so good until that point. I don't mind AI generated image/art but blatantly ignoring the error is a big NO for me. There is always a better picture to use, this just shows being lazy.
"upload dates" don't cut it. The dates of the actual talk is what matters the most, with technology advancing as quickly as it is.
when was it uploaded?
@@theJesai Looks like the "upload" date was March 28th, 2024. This Talk is much older than that. When this happens, I believe we need to see in the video description the date of the original talk.
Your excellent comment shows you are more qualified as a presenter than whoever is doing this TED channel on KZhead. I mean come on, TED, this is grade school stuff: put the date on it. Information is useless without an historical context.
0:38 You can clearly see it shows the date in the video "September'23". did you miss it?
@@kalimero86 Your excellent reply shows that I am the least qualified KZhead watcher in the room. Doh! Leaving original comment for historical purposes, despite my extreme instincts to the contrary…
Please always add the date of the talk, thanks!
The key is continuous learning. I'm still "learning" how to make my morning omelette. Each day I get new feedback; try new techniques, gather more information and learn from tiny mistakes. Robots must do the same.
This is the type of presentation I can never forget. Knowledge is shared with humour. Well done Kim!
Robots don't need to be perfect, they just need to cover 3 criteria. 1) graceful failure states. A robo vaccuum sounds great, but if it is going to choke on a string, or smear a mess and make things worse rather than better... Yeah, no thanks. If a folding robot rips a shirt a week then I'm not keeping it around. I feel like this is the biggest hurtle right now. The price of many things has gone down, and reliability gone up... But those rare fails are big fails that undo all of their savings. They don't need to be perfect. They just need to fail better. 2) they need to save time. I keep looking at robot mowers, especially during spring allergies. My fear is that I will spend more time picking one, and programming it, and maintaining it, and replacing it than I will save by just mowing the lawn myself. The argument that learning the skills to maintain it are more valuable than the skill of mowing the lawn is not lost on me. But what I need in life is time savings, not a new hobby. And the savings needs to be well beyond 10:1. It doesn't matter if a robot is significantly slower than me at a task, as long as it saves me time. A robo mower may take 20-40 min a day to cover the same yard I do in an hour a week... But if I don't have to do anything for that hour a week, then it is still a net benefit to me. 3) it has to be a money saver. So many appliance style robots come with a high up front cost, and a yearly subscription, and generally do a worse job than I do at the task. Like time, it has to be in the 10:1 savings range over the life of the product to make it worth it. So either a loss leader with a subscription like printers have moved towards, or a high up front cost for years of little to no maintenance like traditional appliances. Robot mowers that cost thousands, and then cost monthly are never going to break even before they are replaced. It may free up some time so I can work more... But me working for my robot kinda misses the point of the hired help. 😅 But I think we are getting close. Robo vaccuum cleaners are already just about there. And we are on the cusp of many others. I think the big mistake companies are making is designing specific use robots, or building humanoid robots, when neither are particularly easy or cost effective. Something that can use the equipment I already have, but without the fall hazard of a bipedal robot is ideal. Something on rollers or treads, with arms and a mast for cameras is the simple answer to so many of the issues. Then as long as it can push a vaccuum, or manipulate laundry, or reach the dishes and stove... We can add capabilities with processing and software upgrades over time. But the actual form factor and mechanic side of things is there already. Just need to dress it up in a way people will enjoy, and gain more skills.
Robot pool cleaners are also already very popular.
Good points, but here are some subtle counterpoints that could be used by a successful robotics product: 1) If I can shift my time commitment, then that's good enough. 2) if I can narrow my skills it is sometimes good. Rather than learning and acquiring equipment for a variety of household chores I might prefer to be good at only robot maintenance. Maintaining robots at night rather than doing ten different chores might suit me - or I could even hire a single maintenance worker to do it.
I honestly don't think we'll make real progress in making humanoid robots like this until we can give them tactile sensors. Like, we get so much subconscious information about our surroundings from our skin (well also the tiny hairs on our skin). I feel like even just knowing "I am touching something" without having to "see" it would help so much with clumsiness.
This is precisely what I want to see in robotics; efforts to automate away drudge work. Kudos to them.
that reminds me of the shirt-folding board that reduces a significant amount of time with a dollar. We always need to keep the problem-solving mindset while approaching the problem. I am impressed by how persistent the honourable researchers are with getting rid of chores for everybody. Applause the real heroes!
6:55 It's not that people "don't like doing this work", it's that the companies providing the work are mistreating and underpaying them.
I'm so glad to hear an insider's view on the ongoing challenges of modern robotics. ❣️
Wait 5 years ... Robots and A.I. are merging. They'll be everywhere soon.
As an old person, I've been hearing this every few years since the early 1970's.
I was a mechanical engineer at UC Berkeley/Lawrence Berkeley Laboratory for 30 years. I’ve built automation for production lines. I seen Goldberg’s laundry folding robots over the years. I’ve witnessed the progress in technology and software. When he says be patient, the robots are coming he means not in our lifetime for robots of the “Jetson’s” sense in our homes. Expect continuing incremental progress for commercial applications.
They are gonna try and deliver things sooner, Project GR00T, is also using simulations and learning by example just like Pieter Abbeel predicted.
Good talk. But it almost seemed like there was a "laugh track" everytime he said something.
I, too, find that annoying. Why can't people be serious about these subjects? Do they really need to be entertained like a bunch of children?
@@ReadingAdam i think they actually get paid for laughing and there's a guy with a laugh sign that tells them when to laugh
No need for a track as AI revolutionaries think it's hilarious that the disruption they are dreaming of will take millions of jobs in the US alone over the next few years.
@@flickwtchr Right. The same way the switch board got rid of the pony express.
Great session Ken! Keep up the good work.
All the ads i got in this video had the "x gon' give it to ya" song ny DMX in them, and since i was just listening to it, i thought it was part of the TEDx theme. I had to go back and check, and am now slightly disappointed that its not lol
Amazing presentation
The thing that keeps me from buying a vacuum robot is my habit of letting things lieing around on the floor.
some current issues: scale & targeted specificity of training (not very general beyond task, lack gen int) inference can be slow inference can be expensive inferences via foundation models are disjointed --- with foundation models / LLMs that will help knockout the scale/generalization issue, looks like some companies are using hierarchical methods to make progress ie figure01 costs per inference seems to be declining rapidly speed of tokens per second or inferences a second is also making progress. Disjointed nature/statelessness of inference is the last hurdle; research is currently investigate State models & memory units to learn from past activations to inform current inference. Though prob not 100% necessary, as tasks can be broken up into many smaller tasks, this would help with coherence esp of goals across inference and fluidity -instead of each inference popping into existence at that moment not aware of the context, 'thoughts', why of previous inference and then handed the chat history up to that point to seem somewhat continuous. Though there are hacks like having it think aloud it's plans and goals needed and currently doing but some nuance is lost thus not ideal.
this is what i imagine a robot would write if it was to report its own experience trying to fold a shirt
At 2:25 he’s basically showing you Roberto from futurama 😂😂😂
I want 3 robots: One geofenced wheeled robot to crawl around our lawn and dig up the thousands of weeds. A second wheeled robot to chase deer (and maybe rabbits) away from our garden. And a third (also wheeled) robot to crawl around a roof with a rotating brush and remove the moss which grows on roofs in the Pacific Northwest. Somehow I suspect that at least one of those would not be so difficult.
Sort of interesting how he often comes back to belittling the Atlas robot (made by Boston Dynamics). I'm sure that has nothing to do with the fact that Boston Dynamics with their Stretch robot is basically one of his competitors in the warehouse automation business, or at least might try to become on in the future.
Great content 🙂
It seems like some of the robotic demonstrations in recent weeks have shown more progress than suggested in this video. The demonstrations might be staged to look much more consistent and reliable than they are in reality but they certainly seem much more capable. Apart from further refinement, the biggest challenge seems like it’s combining everything into one functioning, untethered robot.
that falling robot was hilarious, like a drunk russian guy trying to walk home
Robots are extremely difficult and to make them viable we will need significant advancements in Technology alongside the adequate funding. Hardware has always been difficult to innovate in, we cannot expect massive innovations in Hardware while all the funding is allocated to High Growth Software companies
And the global economy will crash before that happens. Software will replace white collar jobs but blue collar work will be safe
Cool edutainment video. And technology only gets better, when a group of people work hard together to make it better.
TED, you sure this video is recent? Not from like years ago? I feel like I'm having a déjà-vu watching it. But nothing textually says the date
Because of the recently publicized advances?
lots of generated images in the presentation so it was likely done within the last year, and definitely not much longer ago than that.
@@SirCutRy I feel this presentation could have written in 2020 or even before that. I've been closely following news of AI advances for many years.
You're right, I agree
Talks about the pandemic so yes, it is more or less recent.
With spirit, elegance and humbleness. Opposite of Hilton's hype style.
"These robots, they're coming. Just be patient." *Looks down at the android arm I'm repairing while the AI Jetson controlled head blinks and tracks my face*
brilliant
Let's play "spot the extra fingers"!
Japan had tried so many decades, still nowhere near robot maid
We don’t have to say „HURRAH!!“ every single minute. But maybe every other day, for just a little moment, we should appreciate what an absolute MARVEL our body is! And we should be aware of how fragile and still robust everything is. I am an MD and if I start talking about all the thousands of systems working perfectly every single minute of our lives I won‘t stop for a day 😅. It’s taken millions and millions of years of trial and error, so no wonder it’s hard to replicate. Great talk!
Fine motor skills are a lot more complicated but because we get so good at it thru repetition and feedback that we don’t realize all the information we are processing.
The people makin em don't wanna be replaced.
I'm glad to see Dr. Steve Brule moved on to robotics. It's good for the robots!
Amazon delivery stations could really use those sorting robots, lol. It sucks the way we do it now which is basically completely manual. The only machine is the conveyor belt from the dock to the sorting areas.
4:00 LIDR is not new by any means. Been in service since the 90s. Even Mercedes cars have had this.
Another thing that I think is holding robotics back (Honestly it's kind of a problem with computing in general) is that we've created a machine that operates on absolutes, in a world where the abundance and rate of change of variables makes almost nothing absolute. Like the difference between an analog audio wave and a digital one.
7:52 isn't it a bit weird that as soon as the university gives them something they can profit from, they run off and start a private company? Wasn't their research publically funded up until that point? As far as I know, even private colleges and universities get government grants.
Robots can assemble a car, which is a much more complex task than preparing lasagne. But no robot exists that can prepare a lasagne in my home. That’s because car factories, along with the entire automotive supply chain, have been entirely reconfigured to help the robots. If I was ready to reconfigure my house and the way all the supplies arrive and get stored, then robolasagne would not be an issue. We don’t have multi-purpose home robots because we’ve done nothing to make our homes robot-friendly.
Nor should we. We should have the robot accommodate us, not the other way around. A multi-purpose humanoid robot is the solution.
There is no single robot that can assemble a car though. It takes many different specialised robots breaking the complex task down into many simple ones.
Every single person would have to use the same appliances, same layout, same house, same bath towel.... That would also mean that in order for it to be uniform, no living thing can be in the house unless every single person moves, talks, acts the same way because they would introduce variables. This does not make sense for living spaces, sorry. Robotlasgna doesn't exist because there is no market for it that will out weight the R&D and other investments involved. There is however a large enough market to do that with vacuum/mop/mower robots and those do not need explicit layouts. You're argument is static vs dynamic approaches. Static make sense in sterile environments but living environments are dynamic.
That's where the unmet demand is: Designing robots for a specific environment that you have no control over. For example robots that can inspect roofs and attics, or explore caves.
@@JJs_playgroundI guess that depends on whether you want home robotics within your lifetime or are happy enough for next generations to experience this 😢
I fully trust the big tech CEO's to keep us alive, and care and provide for us even when we will provide no value. Tis is because they allready have a great track record for putting people above profit. 👍👍🔥🔥🔥
I was going to make a topic about this as well. Our society is already quite fragile and we're going downhill fast, I don't even want to imagine how our world will look 50-100 years from now. Who exactly are going to pay these people who can't get a job because there are none available? There are probably going to be a lot more people tomorrow than there are today. It's such a mess, don't get me started on housing, these robots will have to perform miracles like creating money out of thin air. You can count on one thing though, we humans will always be greedy and looking to get more power. Don't get me wrong I really, really like the progress with robots, machines, AI and all of that. They will be able to do things we can't even dream of like for example travelling to other stars but I'm thinking more about our children and their children and so on. It can't possible be a bright future for many. Conflicts and wars are probably going to get a lot more common in the future.
very entertaining and yet in these ai times, reasuring at the same time.🥰
Ken Goldberg forgot to mention that 'tasks we don't want to do' is universally 'tasks we rather not pay people for'
I mean yea but what's your point? The vast majority of people coldn't afford to pay someone to do all the things they don't want to do even if they wanted to.
@@noone-ld7pt I'll clarify my position: Normal people would rather have a human touch and a human mind interacting with them, while corporations would rather not pay you a pension or a livable wage. Robots are meant to replace us in the job market not because they are better, But because they are cheaper. this will lead to decrease in quality and increase in garbage for our planet.
At the end of the video... "Task we don't want to do." Nice euphemismo for: "Task CEOs don't want to pay us for".
The sheer amount of computation going on in the human brain is not something current computers can take over.
Try adding a camera to the “hand” of your robot.
How can I intern I such cool projects
Actually I think that Deep Learning digital AI is the flaw in all those attempts. It has limitations that prevent it to truly grasp things, which would make a DL AI robot randomly unsafe to be around. To paraphrase Gary Marcus, DL is a ladder, and building a better ladder doesn’t help you go to the moon. Something completely different is likely needed. My hunch is that one of the requirement is likely the need of memristors. Another one is that to truly grasp / understand things, the robot may need some kind of conciousness / sentience (like any animal), and this may only be possible in the analog domain (not in the digital domain). This would also mean that it would create a new lifeform, that does not rely on biology, and create new ethical problems.
It may be that new life forms cannot be created unless they are biological. Why do machines even need to be like living beings? Why not use machines for what they are good for and let humans do what we do best?
@@gtdcoder « It may be that new life forms cannot be created unless they are biological » -> I agree, at this point in time, it hard to tell if it is or not possible to create a new lifeform that doesn’t rely on biology. More research and attempts/failure will probably be needed to have more clues on that… « Why do machines even need to be like living beings? » -> Like any tool, some tools are better at some tasks than others… Indeed, all machines may not need that kind of capabilities : robots doing repetitive tasks in a controlled environment do not necessarily need to grasp/understand what they are doing. But if you want a machine to be able to adapt without any human supervision to completely unexpected scenarii in a fully open environment, then some kind of conciousness / sentience (like an animal) may be needed. And it is very likely a requirement to be able to do fundamental research and invent new theories in math, physics,…
This video is from September 2023 (for anyone who is interested when this was recorded).
When you’re talking about eliminating tens of millions of jobs from numerous professions, it’s kind of a big deal. This isn’t one profession getting obliterated. It’s many: manufacturing, legal work, software design, truck drivers, hardware design, medical diagnosis, food service, etc.
Please, a robot housekeeper in my lifetime. We can send a man to the moon, but we are still filling the dishwasher by hand.
We already have that level of tech. People simply refuse to sell it. Probably Military / Gov contract related. Thank the people YOU probably vote for that you don't have it yet.
It will need stainless steel design structures , seperate replaceable body parts, sustainable, energy that last 24hours without charge, movement charger, sensors must focus on heat and cold method light for head light onoff feature to work while no light. Abilities to fix and improve itself, for the owner to seat and relax, but alot of factors including running balance, also must have wheels for friction, walk balance, detection in a lot of things ask the own if this can be introduced to the program that how going to improve much faster.
Its weird how human intention is the reason we can move an object exactly where we want to. I guess we "course correct" as we move something.
You mean intuition?
Any updates on sexbot?
The definition of simple is also always defined from a human perspective. The world's simplest things are so complicated by themselves 😅.
Combing those functions into one robot is important to commercialize.
But hey, looks like they are decent enough at supplying you with imagery :)
I dont want a robot in my house. No my toaster is not a robot.
5:35 Human sensors are also very noisy. The model built by our brain is clear. Look 2 inches to the right of this text and try to read.
Thought it was Jochem Meyer
it's kind of like joel hodgeson doing steve brule
I am doing translation work fairly regularly and have to struggle , even with help from capable chatbots like Gemini or ChatGPT....eventually it's almost impossible to without capable and advanced skills interventions by humans
ChatGPT 3.5 or 4.0? ;-)
@igorthelight Free version 3.5 ...- with 4.0 it's much better?
Before having these robots we need to solve energy sources. Otherwise they will never be widely accessible.
Ten years waiting, then ten years utopia then Terminator.
The key is to live in such gluttony during the Utopia phase that you die as soon as Terminator phase begins.
We really do need better robots and ones that detect and kick out spam bots.
We have the mechanical part . The AI is still lacking .
What I learned today. Why spend money to buy stock photos for overhead projector slides when you can use AI to generate them for free?
As I understand it, robotics has advanced beyond what is presented in this video. Either the speaker is unaware of it, the audience is unaware of it, TED is unaware of it, or it is just to feed the little people so they don't understand how their jobs will be gone within just a few years. My broader comment, which is true regardless of the status quo of tech advancement is that this video perfectly portrays how little regard the AI revolutionaries have for the disruption that they (a tiny fraction of the human population) are pushing on the rest of humanity, most of which are unaware of the coming disruption across just about every sector of the economy, and how soon it could happen. We are assured that robots for instance, are just going to take away jobs that are "boring and repetitive", or "unfulfilling", or "hazardous", or whatever propagandistic narrative they have come up with. One such propagandistic narrative in regard to LLMs is "don't worry, they are still just really dumb", meanwhile the AI revolutionaries promise AGI in months if not years. Oh, yeah, there is supposedly UBI to be distributed by predominantly Libertarians and neoliberal economic types that dominate Big Tech. The same people who scoff at the notion of paying effective tax rates that are lower than what most teachers pay, and oppose anything that has ever smacked of what they consider "socialism". Back to my original point, NVIDIA is one of the companies that is advancing rapidly in regard to developing a world modeling technology designed to interface with robotic embodiment whether it is humanoid, etc. So, look up just that and you will be up to date more than this video uploaded recently.
I'm sure your layman's perception is much more accurate than the guy who's been working in the field for decades and we're just so much smarter now than we were decades ago. Or maybe it's actually hard, like the guy who knows about this stuff says.
3-6 folds per hour is 72-144 folds a day. It is ready. Make consumer product now! I do laundry every week, it will have enough time!
Because of a great movie by James Cameron :D jk it probably started way before with H.G. Wells, that fuels the fiction against progression. Yes we haven’t quite figured out progressing further with robotics. But we have time..
There seem to be some big gaps in the simplified theory of robots taking everyone's jobs. If people don't have jobs and are poor as a result then the companies that replaced them with robots will have fewer people to sell their products to. There might be a difficult transitional period but in the end, there could be some sort of UBI (or similar) so that the capitalist cycle can continue. Here in Australia the government has given hundreds of dollars to most citizens so as to stimulate spending during financial crisis. Before that it seemed absurd to think that a government would do such a thing but it opens your mind a little to realise that the machine has more parts to it than just the rich getting richer - which is still the ultimate goal but it’s more complicated and probably doesn't work so well if the poor are too poor. My unqualified opinion is that while we need to keep trying to predict the future of both robots and of AI as well as how both will impact society, the truth is that it will probably play out quite differently. It might even be a self-defeating prophecy where our most popular guesses cause the technology to pivot in other ways that we are less prepared for.
Oh god, the AI art being used for the pictures was just so creepy looking.
yea, normal pictures would have so much more professional.
Why Don’t We Have Better Robots Yet? "We do, contact my company for a quote". :/
This is great but it’s pretty traditional and won’t make progress quickly. We already have AI (Sora) that can generate realistic HD video conditioned on images and text. Why not take video from the robot’s cameras, condition on a task like “POV of robot folding a t-shirt” and use the video as reference to move the actual robot. All of the complexities of physics and uncertainty in sensors will melt away as the robot simply follows the video, and the video subsequently follows reality as new images are fed in to Sora. I’m really looking forward to this new technique as it removes the need for creating complex simulations and training on each task. Sora (and AI video generation in general) will allow robots to do anything that can be generated in a video. Honestly it may be the last step towards completely capable robots and I’m very excited (and terrified) at the prospects. I hope your lab or someone also explores this idea.
This could only work if the AI is powerful enough so it hasn't only learned "realistic HD" (whatever that's supposed to be) but actual physical processes underyling the movements of the objects in it plus the illumination effects. Now, I don't say a dumb-but-huge attention-based black box model couldn't get to that point - that seems to be more or less how animal and human vision works too. But it's a brute-force approach that's only "efficient" in the sense that it can exploit the massive parallelisation of SIMD architectures. The problem with such approaches is that it's basically impossible to know how reliable the model actually is, specifically what the underlying assumptions on the operation domain are. The model will always be influenced by biases in the training set, which we likely will never understand properly. What actually happens in real-world use cases is a complete roulette then - it may work just fine in 99% of cases, but what it does in the odd failure event is impossible to predict and could range anywhere between "sews back the rip in the shirt" (because it has actually seen that in the training set) and "burns down the house". I'd much rather we'd be content with not making progress quickly, but making progress steadily and based on proper human-understandable models that allow us to stay on top of what's actually going on and what the robot can and cannot do.
It often makes glitches that would take large amount of manual work to fix, if you wanted to use such video professionally. Same as any robot, if it sometimes makes a mess and needs fixing, that quickly undoes all work savings.
Humans don't rely in their environment as it is but as it should be. Seen videos of people walking against glass windows? Same intelligence as a fly except that humans realize that the cause of their accident _should_ be a glass window.
The better q is: Why don't we have better humans yet?
The amount of really strange AI images being used in his presentation is really off-putting...
Because of IT-Security issues and our wonderful world of policies because we humans liked to use every invention to harm others in the past.
i rather the robots do the repetitive tasks than AI taking over artists' careers that we were promised we have free time to do when the said robots take over suck repetitive tasks. making art is NOT repetitive, it's freggin' ART FPS!!!
It's not that humans don't want to do these kinds of work, it's that the companies needing that work done most of the time treat their workers badly and don't want to pay them properly. And those exact companies would want to replace the employeed as soon as they could with robots as they then wouldn't need to pay the employees (robots don't get salaries) which means they would maximize their financial wins. They would not need any holidays, sick days, vacation days... In the end it is bad for the economy though, as those companies don't really pay much tax, but the employees do. they pay tax on their income, they pay tax on anything they buy that they need for living, they pay tax for the vehicle they use to get to work... And let't be real here: not everyone is suited for every job, and with replacing manual labour by robots this means a whole lot of people who are good at that end up without a fitting job. Which is bad for society, bad for econonomy... But everybody usually only looks as far as his hands can reach and doesn't think about the "big picture" and all the consequences.
Again has nobody heard of Skynet I really do not want a home robot thank you very much
This aged poorly already. This video is from September 2023.. So the research he's showing off, is actually outdated already, considering the pace of advancement we've seen lately. In the months since this talk, robot companies have started using multimodal LLMs to control their humanoid robots. And it's showing a significant improvement.
It's not outdated, it's just a more traditional approach. It is IMO a big mistake to abandon such approaches and use LLM black boxes for everything, because these are way less interpretable. They may fail less often, but when they do it's likely to be much more insidious.
Nope still up to date. Those are only trained to walk on uneven terrain which is a lot easier but still far from matching the dexterity of the human hand. Level 5 driverless cars hasn't even been solved yet and these home robots are orders of magnitude more complex.
Actually, I'm confused as to how he's making fun of Atlas in a warehouse, while Amazon just signed with Agility Robotics to deploy the humanoid Digit in warehouses.
@@GH-uo9fy That's not accurate. Have you seen a video explaining Nvidia's robot simulation platform? The one used to train that 4 wheeled robot to stand up and open doors? It's more than just walking on uneven terrain. And several companies have been demoing robot hands that are in my view, good enough for now. Also, Mercedes just released Level 3 self driving, without a requirement to watch the road while it's activated. From here on, that's going to progress rapidly. So I wouldn't use driverless cars as an example like that, anymore.
@@leftaroundabout Have you seen how they're using the LLMs for this purpose? They don't handle everything, just the decision making, or like a conductor. The actions the robot can make, are still trained with a more traditional approach. Traditional methods for programming robots, can only get us so far. LLMs are the missing piece, that enables robots to have more common sense in their environment, and that aspect will only improve from here.
TED talksht: WHY DONT WE HAVE A BETTER ROBOTS?!🤔🤔🤔😅 OpenAI-Figure01: HOLD OUR BEERS. 🍺🍺🍺🍻🍻🍻😄😉😂😂😂
Better robots are coming. Just wait, and it won't be long.
not the robots fail, the creators fail
Even better question: why are we worried that robots are not as good at doing some things that humans do perfectly well seeing as how we already have 8,00,000,000 people available to help out?
Lots of ai generated imagery... lol on the 3x3x4 cube
yep, very obvious.
This video would have been great a yeah ago. It is fairly outdated now with all the AI and hardware improvements in robotics
Combine robot with AI, and everyone will had a lot of free time. Yayyy🎉 *They didn’t mention massive job loss in AI/Robot videos for a reason 😂.
If you didn't notice he used the word "training" (a lot), which is AI/Machine learning.
He talks as if Lidar is recent (1962) but also seems to refer to a pandemic (I assume 2020). When was this talk presented??
yes this video is just obsolete
I think it's pretty up to date. Sure Lidar was first developed a long time ago but only recently got into broader usage. The pandemic was only referenced in relation on how much online shopping has increased since then. What we currently get presented as AI progress isn't really AI and not helpful for this sort of problems
You want to know why we don’t have butler robots at sale for a commoners household right now? The answer is simple: the politics. Particularly the safety politics and a large swath of high-tech being consumed by military/safety R&D. We have drones due to Afghan war and US Army investment in such a technology. We have internet due to Cold War and US Army investment in nuke defence system. If we seriously think about robot butlers at home, we better start to think how this technology can be applied in existing battle scenarios in modern day conflicts. IMHO.
If you ever saw a spider making a web or a little fly cleaning his wings i wonder if we ever can make a computer and robot that small doing the same thing😮 i think there must be some real (not ai) conscience involved...the spider Knows what its doing...how can you say that from any computation, how comlex...
so right now, any robot not going to grasp my livelihood😂
Because Humans are not set to distinction yet