Surprising absolutely no one, I hope. Credibility seems difficult to generate for Tesla events. Maybe the secret sauce for Robotaxis is a human driver somewhere watching the cameras. Like driving Uber but from the comfort of home, and it's easy to hit the fridge or bathroom between rides.
I mean, you definitely need people available to intervene even for a L4 or L5 autonomy, because they will get stuck (Tesla is not serious about robotaxis until they start staffing up a team to do that on a full-time basis). But actual driving? This link is way too high latency for that to be safe. The robot needs to be maintaining its own SA, and just calling the human when it doesn't know what to do.
It's amazing to me just how far people will move the goalposts for Elon's perpetual grift.
We've gone from "you'll be able to nap on your morning commute in your self driving car within 18 months" to "they will always need a human to intervene".
Elon Musk has displayed incredible prowess at manipulating modern internet media. He launched "tesla shorts" right when a few big short sellers announced their positions (I believe one of them also put out a report about TSLA being insolvent aside from income from pre-orders, and was proven correct by Musk years later) and SEO-ed them into the ground.
I would assume that several of the pro-elon accounts on most social media are actually either bots or shills. You don't need many shills to get real people interested.
Errr, I am saying that Elon's claims are obviously BS until we start to see Tesla doing something like what Waymo has had for years(1): a team of people ready to intervene and fix things that are outside the training set of the ML.
I happen to know a senior person at a autonomous delivery robot company, which employs a team of people for just this purpose, because even delivering pizzas around a college town in a small little robot needs this. For things like (actual example for them) a sofa that was being thrown away and was just left on the sidewalk, and so a human needed to confirm that it was safe to move around it. And so far as I'm aware, Tesla isn't doing this, which is why I think that their autonomous taxi idea is nonsense.
1: Personal experience from being driven in a Waymo, I hit the assist button when we got stuck by some double parked cars in a parking lot. By the time someone answered the car had already extricated itself, but it didn't start that until after I hit the button.
Reminds me of a plot of device of the 90's movie Shooting Fish where they were scamming businesses selling an AGI computer but were actually controlling responses with a human in another room.
I mean, shoot me down here, but is it that bad of an idea?
If you're going to have an assistant or a taxi driver, and you start off at the base position of "AI is totally unreliable", then having a fully remote gig-worker remotely piloting your robot...
I mean, it doesn't seem like a massive stretch from what Uber does.
...and heck, having a 'remote robot body' is pretty cool tech.
I guess. As long as you don't use it to pretend its just AI for the meaningless purposes of generating hype about your AI that really isn't actually any good.
yes, operating any kind of heavy machinery over a shaky wireless WAN with hundreds of milliseconds of latency and multiple percentage packet loss is, in fact, a bad idea
It seems like there's considerable demand for human labour "below the API" that you don't have to talk to. It's kind of sad but people seem to get comfortable with it very quickly.
> Allowing random people to pilot robots in your house with children around? Absolutely horrific.
Your risk analysis on this is completely wrong. If there is some vetting here this is fine. No different than a babysitter or a handyman off the Internet
You're saying then that tools exist to scan and ping all babysiters and handymen across the globe, fingerprint them for version, lookup zero-days, apply them to matching staff, exploit that to monitor children remotely, and take control over home assistants function to shepard children out the door to a "party van" ?
That's the ecosystem that surrounds most actual IoT devices - I can't see home robots being any different.
Hopefully, they will check for overemployment before allowing them to remotely control, otherwise we end up with one operator controlling several cars. I don’t know if it is good or bad
To be honest, the autonomous control of the robot seems like the easier part of the equation. (doing it safely in a room with guests, unguided... thats another matter). The physical limitations and packaging are a big challenge, and I dont think I saw Optimus lift anything remotely heavy.. just pull a beer tap.. a decision that probably speaks volumes about current limits of the technology.
To apply my first point to reality: put an Optimus in its current state/capability, on a commercial 0-turn lawn mower, plug Optimus into the mower's power takeoff, and have someone in another country remotely pilot the mower. That right there is worth every commercial lawn service having at least one on their crew TODAY.
The appeal of hot swapping an operator real time on the equipment you already own, whether it's a push lawn mower or a huge mining truck, provides enormous value right out of the gate. Especially in tasks where the Optimus can handle 90% of the task autonomously but needs to step aside or oversight for the last 10% of the job. Compare to a business model that requires purchase of all new very expensive and unique equipment.
I've worked in robotics for over 10 years, at state of the art labs and high quality startups.
There are really only two hard problems in robotics: Perception and Funding.
Perception, especially around a bunch of people, with depth, mapping, understanding traffic and gestures, all in real time etc etc will be a huge problem for these machines for a while.
Funding though? I doubt that's an issue right now.
That surprises me. I thought motion planning and motor control would be harder - old memories of Asimo falling helplessly trying to climb stairs, the clunkiness of a robot aligning itself perfectly with a drawer before executing a scripted-looking action to pull the handle, the obvious recorded sequence Atlas uses to get up from a fall. I know Boston Dynamics does impressive acrobatics, but it's all legs and no arms.
Are kinematics and planning solved now? I want to move into the field so I'm trying to learn.
There are others. Controls is hard! You need investment and to solve difficult engineering problems. But we have a pretty good idea that those things are solveable and can demonstrate success b/c they are engineering challenges, not things we fundamentally don't have an approach for yet.
"What is the orientation and 6 higher order velocities of my body?"
etc.
I've been told that a perfectly observable / estimable system is trivially controllable. It's one of the reasons I believe perception is upstream of everything - interaction dynamics alone are nearly impossible to just wave away with models.
I don't even work in perception. But I know that everything is fine until you try to go online with perception in the loop. Then you are behind the perception team's debugging nearly all the time.
I'm also a roboticist. Perception and funding are hard. But don't forget battery energy density, and the power-to-weight ratio and energy efficiency of actuators. Also very very hard, and Moore's law helps not at all.
Autonomous cars are in a nice niche since they store vast energy for actuation anyway, it's OK to be heavy, and the controls are relatively simple. They are limited by perception and decision making.
Humanoids are way more limited by energy storage and actuation. Animals are absurdly efficient.
Tangential question: are there any actuators out there that mimic the animal muscle tissue, i.e. swelling laterally in order to shorten a tendon and pull a joint? This seems like a very elegant method compared to servos with all sorts of slack and rigid positioning. I'm not a roboticist so I'm not familiar with state of the art in actuators.
Battery density is only an issue if these things are spending most of their time moving long distances. If you are targeting a drop in replacement for a human worker who is spending most of their time at a workstation, it can be plugged in while working. Even in a scenario where the robot can not be connected to power while working, that's easily solved with redundancy - get two robots, one works while the other charges. Obviously better battery life is a nice to have, but it's not an impediment to large scale adoption the way other big robotics problems are.
And how much more expensive is the twice as efficient power system that hasn't been developed yet?
Nearly all robots in actual use have tethers, it's really not a big concern. Further there are other methods of providing power, such as induction. For any situation where long range mobility is really a concern, you probably don't want a humanoid robot to begin with.
A problem that can be solved by spending 2x is not the type of fundamental problem I'm referring to (easily solved by "Funding" or otherwise known as a system design constraint and part of everyday engineering albeit very difficult and skillful engineering)
The leaps that would be required to make a mannequin with motors intelligently interact with crowds (in groups no less) at a publicity event cannot be solved with 2x funding jumps, and I'm arguing they are largely perception-, sensing-, mapping- and self-modelling- based.
What stood out to me is that the speed of the movements makes it clear the autonomous balance control just isn't there. You can simulate this: try to move your upper body while standing up without changing the balance or stance of your legs and hips - unless you move slowly you can't do it, whereas if you put any force or momentum behind it you'll feel yourself straining pretty hard to stay upright.
When you see the bits and pieces of behind the scenes for Boston Dynamics it's clear that's where a lot of the magic actually is (and also if you look at how say, Atlas moves) - by necessity it looks much more "natural" because to get any power or speed behind the motions the whole robot needs to actively compensate the movements (obviously having enough power behind the drive system to actually do it is also critical).
I would imagine latency would be an issue if companies were considering teleoperation using staff in a country with cheaper labor. For example I work with people in India and China and they regularly complain about the several hundred ms of latency they get when using their American VDIs. That off the shelf lawn mower is going to be hard to control safely with all that delay, and there's also the risk of connection drops and the like. You would need a specialized mower with collision detection/etc. to handle this, and at this point you might as well discard the robot and just have a remotely operated mower instead.
However there are cases where this can work well, say in a factory handling dangerous chemicals with the teleoperator in an adjacent room. Or maybe it's doing some sort of task where delays and connectivity loss are acceptable.
Let's see, New York to Mumbai over the Earth's surface is maybe 12,500 km, assume a direct fiber optic cable where light travels noticeably slower than in a vacuum at 200,000 km per second... So a minimum of 62.5 ms one way with the best terrestrial equipment.
While one can play network games at 125 ping, it relies rather heavily on tricks that only work in a virtual environment. (Back in the '90s I used to play with 300 ping, no latency compensation, uphill both ways.)
Realistically it's in the several hundred range. I just did a ping using Vultr's Looking Glass from New Jersey to Mumbai and got around ~240ms on commercial fiber. For people working from home in India (with cable/DSL overhead + distance from the IX) connected to servers in LA I regularly see 300-400ms.
Also keep in mind in a VDI or teleoperation setting there's not only network latency but additional delay from the video encoding, compression/packetization, and decoding on the other side plus a bit of buffer. Honestly I think cloud gaming is a good test case for this - and in my experience that only works well when you have fiber and have the game server in the same city as you (basically <10ms).
Don't get me wrong I think this guy's idea is incredibly stupid. But, have you ever operated a mower? They're not fast. A few hundred ms of latency on a mower is no problem at all.
Maybe that's the realistic future of 'self-driving' cars. A teledriver-assisted automous car. It just moves the cab driver from behind the wheel to behind a screen somewhere else.
Waymo vehicles are not driven remotely. Remote assistants give the autonomy stack suggestions for how to proceed rather than drive the vehicle. This doesn't require a low latency connection and the robot is still capable of stopping when the situation changes or proceeding as soon as it's able to without a control handover.
Yes, they are not usually driven remotely, but an operator can take the wheel in an emergency situation. Most of the interventions are "this plan or that one?" decisions from the teleoperators.
That still isn't really "autonomous," but it's a lot closer than anything Tesla has done. My question, though, is how frequent the interventions actually are.
It also doesn't scale, which is the big problem. Waymo works but excruciatingly mapping out the city and its routes. It's not a generalized autonomous driving algorithm.
Which is probably fine, but it does mean it will never make it to a lot of areas.
There's a company in vegas that pilots a rental car to you so you can be picked up anywhere and when you hop out, it drives off. You rent by the hour or something like that
"This is awful! This is nothing like the Hell I visited two weeks ago!" Bill Gates responded. "I can't believe this! What happened to that other place, with the beautiful beaches, the beautiful women playing in the water!?"
"That was a demo," replied St. Peter.
also ED-209 from robocop, "You have 20 seconds to comply."
Or Teslas "battery swap" event which actually helped them earn them a huge amount of subsidies from the government. Watching after understanding Elons m.o. it is clearly fake or how else would they have a planned demo on stage with cameras that aren't positioned to show anything that's going on and then that tech they developed was never heard from again. They just distracted the crowd with a video of someone filling up at a gas station
I was assuming that the robots at the Sphere also had some humans behind it if nothing more than to help "guide" the AI pieces. My assumption, at least.
That event was a huge disappointment. It's clear that Elon didn't consider it to be that important and didn't put any real effort into it.
There was nothing an investor could look at and get excited about, it was the same thing as he announced 5 years ago. Just now his self driving cars have been eclipsed by Waymo and cruise seems to have caught up to what they can do with their demos.
And why show the robots at all if they were just remote controlled by employees.
It's all about controlling money that might go to more honest ventures.
Yes, humanity has engineers who are going to the moon, creating robots, investigating brain interfaces, improving public transport with buses and tunnels.
And there will always be monorail salesmen who try to soak up those investments, taking away from others.
Yes that is usually the way things go for rich people.
They have an asset and take out a loan against the asset (e.g. stocks). Say $10M. This does not count as income. Then they spend the money on what they want and refinance the loan with another loan, say $20M on their grown assets. Then you spend more money, again with no income, so no taxes. When the asset goes up, you refinance. When the assets crash, banks can have the assets. Again, you've spent the money without income. One of the many tax "avoidance" schemes (this one is for income, others are for inheritance tax or VAT avoidance) of very rich people (those with no real income, not professionals with very high income).
You can read about it e.g. in the leftwing-marxist-propaganda Forbes Magazin:
I think it was fairly obvious from the show that the Optimus bots were remotely operated. Not like they tried to hide it at all. Just listen to the responses of the bots, they practically admitted that.
The cars, however, were almost certainly running the latest FSD (or some near future unreleased version).
But they did try to hide it. It's in some videos where one of them was trying not to admit they were remotely controlled and that "probably some AI is used"
AI is a pretty broad term - its been used for generations to mean something intelligent seeming that isn't human. Certainly, computers running precompiled instructions also falls in that category.
Those are definitely exaggerations for marketing purposes.
I find it incredibly silly that 20+ years after the first more advanced "AIs" in games like Half Life, we're still far from a point where I can fire up a game of Dota 2 with bots and have the bots behave constantly well at an intermediate level, so that I can turn Dota 2 into a solid single player experience (I don't have the time nor the patience, anymore, for managing a team of toddler brains for 30-60 minutes).
Yeah, I wouldn't classify mechatronics or bipedal locomotion as AI. It gets blurry when considering algorithms in control systems, but I wouldn't classify those as modern AI/ML systems.
What does everyone think about 1X's NEO? [0] They began from the idea of compliant robotics,[1] which seems to me to be a requirement for safe operation in proximity to humans.
Did Tesla make attendees sign a hefty liability waiver, since Optimus is not a compliant robot, or did they address the inherent problems some other way?
Each robot had several human escorts, and the robots were limited to slow walking and a few slow hand gestures. The only danger would be if one fell over.
1x NEO looks awesome and far more advanced than this version of Optimus. I'm bullish on 1x. Tesla has a manufacturing advantage though. There were 50 units of Optimus at the event and I expect that there are only a few fully working units of NEO made so far. Also, Optimus has been improving quickly. It's possible Tesla could catch up in a few generations.
We had people swearing up and down that the Cybertruck would never be released because it was stock fraud. They already seem to be at a run rate of $1.5 billion in revenue per quarter despite launching in a limited fashion because of manufacturing ramp up.
I haven't really been tracking this, but aren't they selling at a price much greater than they originially announced? I would classify it as "mixed success"
The base model isn't entering production until next year but is quoted by Car and Driver as starting at $62,985. The original price was supposed to be $39,900. However there has been >20% inflation since then, and the base model should be eligible for the full $7,500 tax credit applied as an instant rebate at purchase time, which was not available at the time of announcement.
All that to say: the price actually paid will be $55,485, vs the inflation adjusted original base model price of $48,912 (possibly higher if there is more inflation before the release next year). So yes, the price you'd actually pay has gone up 13% in real terms over the announcement price. But I think the extra 4 years of delay (for the base model vs the announced availability date of 2021) is the bigger issue.
But ultimately all that matters from Tesla's perspective is that they are selling. My understanding is that even the current top end expensive models are selling about as many units as all other electric trucks combined.
> The original price was supposed to be $39,900. However there has been >20% inflation since then
That price was announced with a significant lead time, so at least 5-6% inflation was built in already.
> and the base model should be eligible for the full $7,500 tax credit applied as an instant rebate at purchase time, which was not available at the time of announcement.
It wasn't available that exact moment, but it existed.
> All that to say: the price actually paid will be $55,485, vs the inflation adjusted original base model price of $48,912 (possibly higher if there is more inflation before the release next year). So yes, the price you'd actually pay has gone up 13% in real terms over the announcement price.
I'd put the inflation-adjusted price at $46k, and not use the $7500 to reduce the difference, making $63k a 36% increase. Or I'd apply the $7500 to both and get 44%.
The $7500 existed only for other manufacturers. It had expired entirely for Tesla. No Tesla model was eligible for any more credits at that time and there was no particular reason to believe that Tesla would ever become eligible again in the future. Also, the credit back then did not directly reduce the price paid at purchase time the way it does today. I think it is fair to apply the credit when comparing to the announced price. Tesla certainly takes the credit into account when they set pricing and buyers account for it when purchasing.
Also it was announced at the end of 2019 for 2021 and the inflation rate in 2019 was 1.8% so there's no way they accounted for 6% inflation. 4% at most.
Whether it'd be a true success remains to be seen, at least in terms of making a profit on total investment. My point was that it certainly wasn't vaporware, even if it didn't sell well.
Elastic tendons seem like a pretty reasonable solution to this problem, but can it be applied to all the joints (like ball-socket, etc.) I wonder?
I think the chassis of the robot should also have compliance, humans certainly do have squishy spines. I mean imagine you're on the street and you have to share sidewalk space with these things. Running into it would hurt.
>Each robot had several human escorts, and the robots were limited to slow walking and a few slow hand gestures.
More importantly, the robots were limited to doing no real work. They just feebly pick up objects and place them somewhere else, which I am pretty sure doesn't require AI.
For example, the vid shows the robot pouring hot water into a glass with a massive funnel strapped to it. Why not have the robot fill the kettle, place the teabag itself, etc? It seems like the kind of thing that should be developed before walking and talking and telling jokes.
What if the refrigerator, microwave, etc. could interface directly with the robot. For example, the refrigerator has some type of robotized shelf that is able to bring a rack of orange juice to the front before the robot comes over to grab it? What if the microwave is able to focus the microwave beam on the food to cook it evenly?
It also irks me how the robots are just humanoids. Like for example, why have a head with two eyes. Does it need to wear a helmet? Does it need exactly 2 eyes at exactly human-like placement to achieve stereopsis? Why not have 3 eyes? Did the designers think about the form of the machine at all, or did they just produce robots in the form that is associated with the most hype and thus will bring in the most investor capital? Is this really the ideal form for interfacing with humans? With other robots?
I am just very skeptical of these companies that want to go from zero to doing everything. By the time they accomplish a robot that can do "everything", who is to say that they will even be able to privatize it? The "everything robot" might just be built out of general-purpose components and software at that point. Why not just make a machine that does a limited set of tasks well and then build from there?
>It also irks me how the robots are just humanoids. Like for example, why have a head with two eyes.
the risk-averse, cowardly, snivelling product design is really one of the most odious things about the whole Tesla shitshow. they had the opportunity to completely redesign the automobile from scratch, but chose to meekly clone the exact same bog standard sedan design everyone else converged on 50 years ago, clinging to some form response about safety despite the front of a Tesla crumpling like paper in any collision anyway
Though they can't move away too much from the teardrop design, it's one of the most aerodynamic shapes (and why so many production cars look like a decorated teardrop). Still agree they could have been bolder with other parts of the silhouette, it's unimaginative "futurism" coupled with some strange need to be branded Apple-esque.
Hey Chris, thanks for your continued diligence. This is actually a recent and related discussion, as the duplicate status suggests the same article has been posted twice (e.g. slightly altering a URL parameter to escape the HN dupe-detector).
P.s. If you found the content in one article to be better than another, it would be helpful to steer folks towards the more informative one. In this case, the above Bloomberg article is pretty substantial compared to the one you've linked as a "dupe". Take care.
Dupe means duplicate discussion. There is an earlier discussion on this news story with plenty of upvotes and comments. Stop splitting up the threads, especially when it's a news story already developed from multiple days ago. Got something to add? Share it over there. Even maybe suggest the link over there in some cases even as a replacement article option. Stop splitting up threads and forcing us to repeat ourselves over and over. The discussion is over there.
Tesla was not trying to hide this. The robots were telling anyone who asked that humans were helping control them. Unlike the robotaxis, which were explicitly advertised as autonomous.
They’re prototype mechanisms that don’t have any sort of significant AI yet. Them having AI was never claimed. They’re shells, with (as shown) extremely fluid control systems, with the smarts to drive them still in development, all of it for a whopping three years now.
How is stuff like this not considered fraud? This seems much worse to me than Musk's usual Tesla lies in which he is predicting some future capability. At least there is an argument that Musk believed it those at the time he said them or that they were optimistically possible despite being impractical. This seems to be material misrepresentation of the viability of one of the company's core R&D projects that Musk claims "will be the biggest product ever".
Were you at the event or do you have a source for that claim? I have seen video of one of the bots refusing to give an answer to that question and there were multiple articles in the wake of the event that couldn't come to any definitive conclusion so I'm skeptical of this claim of immediate transparency.
Is this the video you saw (posting below, for you and anyone who hasn't seen it).
While ignoring the question the first time, he did confirm that they were being assisted by a human. Perhaps not as clear as 'remotely operated', but that's about how I took the answer. YMMV.
No, the one I saw ducked the question multiple times so that video is more informative even if "assisted by a human" is still somewhat vague. I guess they just left what to disclose up to the individual operator making the level of deception dependent on who you happened to ask. Such a bizarre way to handle an event like this, but I guess that type of haphazard approach should be unsurprising for a Musk run company at this point.
I didn't go, but I saw the video that someone else linked. Also, during the event DirtyTesla was telling people that he was told they were remotely operated. I don't think he had insider knowledge, just answers from workers.
I did see that a few other people seemed to get more vague answers, notably including MKBHD. It seems strange that there wasn't consistent messaging.
Tesla (Elon Musk) is under investigation for fraud for insinuating robotaxies are just around the corner, and you should buy the already available hardware for 100k that'll yield you 30k/year in the near future.
When it became public knowledge, more value was written off Tesla stock than Enron, Theranos and Nikola combined.
Didn't OpenAI demo facetime with a H.E.R. like chatbot/AI friend in April?
Where is that.. it's not available to anyone I know and was that actual real tech running or just faked demos? The tech playbook is hype even if it's not real and or really exists ... hype it up .. make them pay for the promise of something they think exists but it doesnt. Similar thing here Musk following the technology/startup playbook ... hype hype hype make people think it's real .. boost stocks as Open AI boosted it's subscription revenue Im sure in April of a promise of something that may or may not exist.
I subscribed twice and felt screwed not going to subscribe again until i see a live in person demo and or a trusted tech news source saying its available to all with a video showing them using it. Is there a recent such tech news report saying and showing such? Be good to see it!
I did startups and played this hype and create fake content/news to boost metrics and saw results. And yet i signed up twice lol
Fraud is a material misstatement of fact to induce someone to do something. It's really not clear that there's any actual inducement here; this isn't exactly the centerpoint of a pitch video to investors. (As it turns out, Tesla's stock went down like 9% the next day after the event, with most of the analysts' reactions to the announcement essentially being "okay, so he doesn't really have anything that he wasn't already pitching").
Did you know that many demos on events like Google IO are scripted or hardcoded? I'm saying that not to advocate for such practices, but rather to provide context.
For what it's worth, it was quite clear to many that the robots are teleoperated and it still serves as a demonstration of the hardware.
>So now one of the things we wanted to show tonight was uh that Optimus is not a canned video. It's not walled off. The Optimus robots will walk among
you. Please please be nice to the Optimus robots.
So you'll be able to walk right up to them and um they'll serve drinks at the bar and uh you'll directly - I mean that's it's it's a wild experience just to have humanoid robots and it's they're there they just in front of you. Uh so yeah with that um let's party. https://www.youtube.com/watch?v=6v6dbxPlsXs&t=1355s
So he didn't actually say autonomous. I think you'd have a job proving fraud there.
Elon's a master at saying things that people read as something more than he actually said. I knew without seeing it that there's 0 chance that he (or anyone) actually said they're autonomous. But also that he didn't say they're not.
That was obvious to anyone with any experience with real-world robots.
Nice piece of machinery, though. Boston Dynamics' humanoids were clunky electrohydraulic mechanisms borrowed from their horse-type robots. All-electric is now possible and much simpler. Schatft was the first to get this working, and they had to liquid-cool the motors. Don't know if Tesla has to liquid cool. They do that in the cars, so they certainly understand liquid-cooled electric motors.
I suspect that body balance and possibly walking were automated. It's hard to balance a teleoperated robot manually, and robotic biped balancing has been working for years now.
While I didn't hear anything, I think the walk control would have to be local circuitry. I can't see how a remote operator could sense balance and do all the leg movements.
But the problem is the majority of people buying and hyping TSLA are not working or have close experience with robotics to see what’s real vs human controlled. That’s intentionally deceptive.
Well, maybe they shouldn't buy the stock of the company they don't understand? Or at least check experts' opinion beforehand? This sounds like an obvious hint in case of pharma or defense industry, why automotive should be different?
I agree. Also, especially in the extremely crowded and noisy context - what would have been the chances to have the demo working so well?
In fact, even if the robots worked very well autonomously, you would still have wanted a way to ensure that the demo is successful - the same way Steve Jobs did with the iPhone demo, Larry Ellison did with the Oracle servers demo, etc.
So many stories like that in the history of famous product launches.
The one thing that bothers me a little is that if you look at the robots dancing, they are only moving the upper body; their feet are always on the ground. I would have liked to see them having enough ability to dance and move the legs too… then, again, maybe the gazebo they were in was just too space-constrained, or it was just too risky to do that in the demo - given the crowd, and all the chaotic party context. When you set up a demo, you have to account for the edge cases where your product glitches, not just for what it mostly does very well.
Anyhow, these are all AI issues (as opposed to mechanical ones), and, at the pace AI is evolving, it is not hard to see how these types of issues get ironed out over the time horizon leading to the launch.
The Optimus demo did do a great job at actually making people see a world in which robots just roam around and interact with humans everywhere. .
> these are all AI issues (as opposed to mechanical ones)
Actuation is still a massive problem in humanoid robotics. We have over 650 muscles. A humanoid today can't even approximate that. Sure, a robot might not need that many actuators to be extremely useful. However, to be general enough to be able to interact with any human environment, the number of required actuators will not be trivial.
Add to that gearing, couplings, driver electronics, encoders, thermal management, calibration, noise, maintenance and other per-actuator requirements and the picture quickly becomes overwhelming.
This is an area that is still looking for a significant breakthrough.
- Small, powerful 3-phase servomotors are cheap and easy to obtain. Mass production of drone motors has advanced small motor technology considerably. Tiny motors use to be either toy-grade junk or expensive Swiss precision. That's improved.
- Motors with built-in encoders are, at last, available. Encoders used to be fragile plastic boxes stuck on the end of the motor. Also, thermal sensing inside the motor is common, so you can tell if you're overheating it.
- Permanent magnets are small and powerful, and have such high coercitivity that you don't have to worry about demagnetizing them if you over-drive the motor. The main limit on motor power is cooling. You can way overdrive a motor momentarily, like muscles.
- Motor controllers are now small and cheap, They cost about $1000 per motor two decades ago. The power semiconductors are small. Controllers can be programmed to use very high power levels briefly, monitoring thermal sensors.
It would be nice to have good linear actuators. Linear motors do exist, but never really became a big thing.
I'm totally with you on the evolution of motor tech because of drone and also personal mobility (scooters and hoverboard motors are a steal for what they can handle).
While high torque motors got way cheaper, especially with MIT Cheetah "clones" getting easily available, they're still at least 200-500 a pop (depending on the torque needed for each articulation) from what I could find.
I might not know where to search for the real gems though. Where do you search for cheap powerful servomotors?
It's impressive how well the hardware seems to work now, though the software is still clunky. You can see how well the hardware works under human nervous control in the recent MIT bionic foot https://youtu.be/1tD7qd68i3o?t=36
My kids have been around lots of robots of all kinds. The very first comment they made while watching the event was: "The robots are being remotely operated. There's no way that's autonomous."
Nice looking machines. Far from being practical outside of a highly controlled environment. This does feel like progress though.
245 comments
[ 3.9 ms ] story [ 250 ms ] threadWe've gone from "you'll be able to nap on your morning commute in your self driving car within 18 months" to "they will always need a human to intervene".
Incredible
I would assume that several of the pro-elon accounts on most social media are actually either bots or shills. You don't need many shills to get real people interested.
I happen to know a senior person at a autonomous delivery robot company, which employs a team of people for just this purpose, because even delivering pizzas around a college town in a small little robot needs this. For things like (actual example for them) a sofa that was being thrown away and was just left on the sidewalk, and so a human needed to confirm that it was safe to move around it. And so far as I'm aware, Tesla isn't doing this, which is why I think that their autonomous taxi idea is nonsense.
1: Personal experience from being driven in a Waymo, I hit the assist button when we got stuck by some double parked cars in a parking lot. By the time someone answered the car had already extricated itself, but it didn't start that until after I hit the button.
If you're going to have an assistant or a taxi driver, and you start off at the base position of "AI is totally unreliable", then having a fully remote gig-worker remotely piloting your robot...
I mean, it doesn't seem like a massive stretch from what Uber does.
...and heck, having a 'remote robot body' is pretty cool tech.
I guess. As long as you don't use it to pretend its just AI for the meaningless purposes of generating hype about your AI that really isn't actually any good.
yes, operating any kind of heavy machinery over a shaky wireless WAN with hundreds of milliseconds of latency and multiple percentage packet loss is, in fact, a bad idea
Driving at 60mph with shaky internet connection? Absolutely.
Piloting a robot to fold laundry? Maybe not.
Allowing random people to pilot robots in your house with children around? Absolutely horrific.
Your risk analysis on this is completely wrong. If there is some vetting here this is fine. No different than a babysitter or a handyman off the Internet
That's the ecosystem that surrounds most actual IoT devices - I can't see home robots being any different.
To apply my first point to reality: put an Optimus in its current state/capability, on a commercial 0-turn lawn mower, plug Optimus into the mower's power takeoff, and have someone in another country remotely pilot the mower. That right there is worth every commercial lawn service having at least one on their crew TODAY.
The appeal of hot swapping an operator real time on the equipment you already own, whether it's a push lawn mower or a huge mining truck, provides enormous value right out of the gate. Especially in tasks where the Optimus can handle 90% of the task autonomously but needs to step aside or oversight for the last 10% of the job. Compare to a business model that requires purchase of all new very expensive and unique equipment.
I agree but it is frustrating watching Elon like Michael Copperfield but thinking it is real like a 4 year old.
I don't see a clear advantage of Tesla against other competitors if he will launch it in a couple of years.
There are really only two hard problems in robotics: Perception and Funding.
Perception, especially around a bunch of people, with depth, mapping, understanding traffic and gestures, all in real time etc etc will be a huge problem for these machines for a while.
Funding though? I doubt that's an issue right now.
Are kinematics and planning solved now? I want to move into the field so I'm trying to learn.
IIRC, that wasn't a control problem but a mechanical failure of a gearmotor shaft.
There are others. Controls is hard! You need investment and to solve difficult engineering problems. But we have a pretty good idea that those things are solveable and can demonstrate success b/c they are engineering challenges, not things we fundamentally don't have an approach for yet.
"Where can I put my feet safely?"
"What is the orientation and 6 higher order velocities of my body?"
etc.
I've been told that a perfectly observable / estimable system is trivially controllable. It's one of the reasons I believe perception is upstream of everything - interaction dynamics alone are nearly impossible to just wave away with models.
I don't even work in perception. But I know that everything is fine until you try to go online with perception in the loop. Then you are behind the perception team's debugging nearly all the time.
Autonomous cars are in a nice niche since they store vast energy for actuation anyway, it's OK to be heavy, and the controls are relatively simple. They are limited by perception and decision making.
Humanoids are way more limited by energy storage and actuation. Animals are absurdly efficient.
Yay, twice as expensive.
And power tethers on robots suck so hard. Try it sometime, you’ll hate it.
Nearly all robots in actual use have tethers, it's really not a big concern. Further there are other methods of providing power, such as induction. For any situation where long range mobility is really a concern, you probably don't want a humanoid robot to begin with.
The leaps that would be required to make a mannequin with motors intelligently interact with crowds (in groups no less) at a publicity event cannot be solved with 2x funding jumps, and I'm arguing they are largely perception-, sensing-, mapping- and self-modelling- based.
If you want to make a energy-unconstrained robot into a fully functional android, you have much bigger, fundamental problems.
When you see the bits and pieces of behind the scenes for Boston Dynamics it's clear that's where a lot of the magic actually is (and also if you look at how say, Atlas moves) - by necessity it looks much more "natural" because to get any power or speed behind the motions the whole robot needs to actively compensate the movements (obviously having enough power behind the drive system to actually do it is also critical).
However there are cases where this can work well, say in a factory handling dangerous chemicals with the teleoperator in an adjacent room. Or maybe it's doing some sort of task where delays and connectivity loss are acceptable.
While one can play network games at 125 ping, it relies rather heavily on tricks that only work in a virtual environment. (Back in the '90s I used to play with 300 ping, no latency compensation, uphill both ways.)
Also keep in mind in a VDI or teleoperation setting there's not only network latency but additional delay from the video encoding, compression/packetization, and decoding on the other side plus a bit of buffer. Honestly I think cloud gaming is a good test case for this - and in my experience that only works well when you have fiber and have the game server in the same city as you (basically <10ms).
Teslas also do it, and truly necessary interventions are rare these days.
I sort of hope that it's not that often, but I also thought the amazon store was automated.
That still isn't really "autonomous," but it's a lot closer than anything Tesla has done. My question, though, is how frequent the interventions actually are.
Which is probably fine, but it does mean it will never make it to a lot of areas.
"This is awful! This is nothing like the Hell I visited two weeks ago!" Bill Gates responded. "I can't believe this! What happened to that other place, with the beautiful beaches, the beautiful women playing in the water!?"
"That was a demo," replied St. Peter.
also ED-209 from robocop, "You have 20 seconds to comply."
(112 points, 1 day ago, 108 comments) https://news.ycombinator.com/item?id=41831009
(89 points, 2 days ago, 92 comments) https://news.ycombinator.com/item?id=41815567
There was nothing an investor could look at and get excited about, it was the same thing as he announced 5 years ago. Just now his self driving cars have been eclipsed by Waymo and cruise seems to have caught up to what they can do with their demos.
And why show the robots at all if they were just remote controlled by employees.
Yes, humanity has engineers who are going to the moon, creating robots, investigating brain interfaces, improving public transport with buses and tunnels.
And there will always be monorail salesmen who try to soak up those investments, taking away from others.
And Boring Company salesmen...
They have an asset and take out a loan against the asset (e.g. stocks). Say $10M. This does not count as income. Then they spend the money on what they want and refinance the loan with another loan, say $20M on their grown assets. Then you spend more money, again with no income, so no taxes. When the asset goes up, you refinance. When the assets crash, banks can have the assets. Again, you've spent the money without income. One of the many tax "avoidance" schemes (this one is for income, others are for inheritance tax or VAT avoidance) of very rich people (those with no real income, not professionals with very high income).
You can read about it e.g. in the leftwing-marxist-propaganda Forbes Magazin:
https://www.forbes.com/sites/johnhyatt/2021/11/11/how-americ...
When you die, you never pay back the money, the strategy is called "Buy, Borrow, Die" for a reason:
https://smartasset.com/investing/buy-borrow-die-how-the-rich...
Who is working on this?
The cars, however, were almost certainly running the latest FSD (or some near future unreleased version).
I think "AI" did control their walking. Although calling that AI is probably a stretch.
Nobody reasonable considers ifs and dependency injection to be AI.
I find it incredibly silly that 20+ years after the first more advanced "AIs" in games like Half Life, we're still far from a point where I can fire up a game of Dota 2 with bots and have the bots behave constantly well at an intermediate level, so that I can turn Dota 2 into a solid single player experience (I don't have the time nor the patience, anymore, for managing a team of toddler brains for 30-60 minutes).
Did Tesla make attendees sign a hefty liability waiver, since Optimus is not a compliant robot, or did they address the inherent problems some other way?
[0] https://www.youtube.com/watch?v=bUrLuUxv9gE (also remote controlled for now, while being trained)
[0] https://www.youtube.com/watch?v=Sb6LMPXRdVc
[1] https://en.wikipedia.org/wiki/Soft_robotics
1x NEO looks awesome and far more advanced than this version of Optimus. I'm bullish on 1x. Tesla has a manufacturing advantage though. There were 50 units of Optimus at the event and I expect that there are only a few fully working units of NEO made so far. Also, Optimus has been improving quickly. It's possible Tesla could catch up in a few generations.
All that to say: the price actually paid will be $55,485, vs the inflation adjusted original base model price of $48,912 (possibly higher if there is more inflation before the release next year). So yes, the price you'd actually pay has gone up 13% in real terms over the announcement price. But I think the extra 4 years of delay (for the base model vs the announced availability date of 2021) is the bigger issue.
But ultimately all that matters from Tesla's perspective is that they are selling. My understanding is that even the current top end expensive models are selling about as many units as all other electric trucks combined.
That price was announced with a significant lead time, so at least 5-6% inflation was built in already.
> and the base model should be eligible for the full $7,500 tax credit applied as an instant rebate at purchase time, which was not available at the time of announcement.
It wasn't available that exact moment, but it existed.
> All that to say: the price actually paid will be $55,485, vs the inflation adjusted original base model price of $48,912 (possibly higher if there is more inflation before the release next year). So yes, the price you'd actually pay has gone up 13% in real terms over the announcement price.
I'd put the inflation-adjusted price at $46k, and not use the $7500 to reduce the difference, making $63k a 36% increase. Or I'd apply the $7500 to both and get 44%.
Also it was announced at the end of 2019 for 2021 and the inflation rate in 2019 was 1.8% so there's no way they accounted for 6% inflation. 4% at most.
If Tesla's jacking up the price because of that rebate, they get NO credit for it. 63k is what they are charging and what I will judge them on.
And I don't think the exact moment it applies really matters.
I think the chassis of the robot should also have compliance, humans certainly do have squishy spines. I mean imagine you're on the street and you have to share sidewalk space with these things. Running into it would hurt.
More importantly, the robots were limited to doing no real work. They just feebly pick up objects and place them somewhere else, which I am pretty sure doesn't require AI.
For example, the vid shows the robot pouring hot water into a glass with a massive funnel strapped to it. Why not have the robot fill the kettle, place the teabag itself, etc? It seems like the kind of thing that should be developed before walking and talking and telling jokes.
What if the refrigerator, microwave, etc. could interface directly with the robot. For example, the refrigerator has some type of robotized shelf that is able to bring a rack of orange juice to the front before the robot comes over to grab it? What if the microwave is able to focus the microwave beam on the food to cook it evenly?
It also irks me how the robots are just humanoids. Like for example, why have a head with two eyes. Does it need to wear a helmet? Does it need exactly 2 eyes at exactly human-like placement to achieve stereopsis? Why not have 3 eyes? Did the designers think about the form of the machine at all, or did they just produce robots in the form that is associated with the most hype and thus will bring in the most investor capital? Is this really the ideal form for interfacing with humans? With other robots?
I am just very skeptical of these companies that want to go from zero to doing everything. By the time they accomplish a robot that can do "everything", who is to say that they will even be able to privatize it? The "everything robot" might just be built out of general-purpose components and software at that point. Why not just make a machine that does a limited set of tasks well and then build from there?
Sorry https://blog.comma.ai/a-100x-investment-part-2/ has me coping and seething at the AI space
Boston dynamics went back to building arms on wheels.
the risk-averse, cowardly, snivelling product design is really one of the most odious things about the whole Tesla shitshow. they had the opportunity to completely redesign the automobile from scratch, but chose to meekly clone the exact same bog standard sedan design everyone else converged on 50 years ago, clinging to some form response about safety despite the front of a Tesla crumpling like paper in any collision anyway
Discussion: https://news.ycombinator.com/item?id=41815567
P.s. If you found the content in one article to be better than another, it would be helpful to steer folks towards the more informative one. In this case, the above Bloomberg article is pretty substantial compared to the one you've linked as a "dupe". Take care.
While ignoring the question the first time, he did confirm that they were being assisted by a human. Perhaps not as clear as 'remotely operated', but that's about how I took the answer. YMMV.
https://x.com/zhen9436/status/1844773471240294651
I did see that a few other people seemed to get more vague answers, notably including MKBHD. It seems strange that there wasn't consistent messaging.
Didn't he say "this is not a pusher?"
Where is that.. it's not available to anyone I know and was that actual real tech running or just faked demos? The tech playbook is hype even if it's not real and or really exists ... hype it up .. make them pay for the promise of something they think exists but it doesnt. Similar thing here Musk following the technology/startup playbook ... hype hype hype make people think it's real .. boost stocks as Open AI boosted it's subscription revenue Im sure in April of a promise of something that may or may not exist.
I did startups and played this hype and create fake content/news to boost metrics and saw results. And yet i signed up twice lol
OpenAI is just very comfortable lying right around the time google is doing a big public announcement.
Where is Sora? Everyone and their dog has video gen out now, Sora is conspicuously absent.
For what it's worth, it was quite clear to many that the robots are teleoperated and it still serves as a demonstration of the hardware.
>So now one of the things we wanted to show tonight was uh that Optimus is not a canned video. It's not walled off. The Optimus robots will walk among you. Please please be nice to the Optimus robots. So you'll be able to walk right up to them and um they'll serve drinks at the bar and uh you'll directly - I mean that's it's it's a wild experience just to have humanoid robots and it's they're there they just in front of you. Uh so yeah with that um let's party. https://www.youtube.com/watch?v=6v6dbxPlsXs&t=1355s
So he didn't actually say autonomous. I think you'd have a job proving fraud there.
The 'fake it till you make it' fraud will just make everyone building so-called AI companies look bad and heavily faked with events like this.
But there is still time for the Theranos of AI to reveal themselves. (It is not Tesla Inc.)
[0] https://news.ycombinator.com/item?id=41805764
Nice piece of machinery, though. Boston Dynamics' humanoids were clunky electrohydraulic mechanisms borrowed from their horse-type robots. All-electric is now possible and much simpler. Schatft was the first to get this working, and they had to liquid-cool the motors. Don't know if Tesla has to liquid cool. They do that in the cars, so they certainly understand liquid-cooled electric motors.
I suspect that body balance and possibly walking were automated. It's hard to balance a teleoperated robot manually, and robotic biped balancing has been working for years now.
You can see a Tesla upper body controller here. Not sure if it's what they used for the event, but probably something like that. https://x.com/TroyTeslike/status/1845047695284613344
Reaction of the stock market to Tesla's "demo" was very negative. TSLA stock dropped about 10% immediately after the demo and has been flat since.[1]
[1] https://www.bing.com/search?&q=TSLA+stock
In fact, even if the robots worked very well autonomously, you would still have wanted a way to ensure that the demo is successful - the same way Steve Jobs did with the iPhone demo, Larry Ellison did with the Oracle servers demo, etc. So many stories like that in the history of famous product launches.
The one thing that bothers me a little is that if you look at the robots dancing, they are only moving the upper body; their feet are always on the ground. I would have liked to see them having enough ability to dance and move the legs too… then, again, maybe the gazebo they were in was just too space-constrained, or it was just too risky to do that in the demo - given the crowd, and all the chaotic party context. When you set up a demo, you have to account for the edge cases where your product glitches, not just for what it mostly does very well.
Anyhow, these are all AI issues (as opposed to mechanical ones), and, at the pace AI is evolving, it is not hard to see how these types of issues get ironed out over the time horizon leading to the launch.
The Optimus demo did do a great job at actually making people see a world in which robots just roam around and interact with humans everywhere. .
Actuation is still a massive problem in humanoid robotics. We have over 650 muscles. A humanoid today can't even approximate that. Sure, a robot might not need that many actuators to be extremely useful. However, to be general enough to be able to interact with any human environment, the number of required actuators will not be trivial.
Add to that gearing, couplings, driver electronics, encoders, thermal management, calibration, noise, maintenance and other per-actuator requirements and the picture quickly becomes overwhelming.
This is an area that is still looking for a significant breakthrough.
- Small, powerful 3-phase servomotors are cheap and easy to obtain. Mass production of drone motors has advanced small motor technology considerably. Tiny motors use to be either toy-grade junk or expensive Swiss precision. That's improved.
- Motors with built-in encoders are, at last, available. Encoders used to be fragile plastic boxes stuck on the end of the motor. Also, thermal sensing inside the motor is common, so you can tell if you're overheating it.
- Permanent magnets are small and powerful, and have such high coercitivity that you don't have to worry about demagnetizing them if you over-drive the motor. The main limit on motor power is cooling. You can way overdrive a motor momentarily, like muscles.
- Motor controllers are now small and cheap, They cost about $1000 per motor two decades ago. The power semiconductors are small. Controllers can be programmed to use very high power levels briefly, monitoring thermal sensors.
It would be nice to have good linear actuators. Linear motors do exist, but never really became a big thing.
While high torque motors got way cheaper, especially with MIT Cheetah "clones" getting easily available, they're still at least 200-500 a pop (depending on the torque needed for each articulation) from what I could find.
I might not know where to search for the real gems though. Where do you search for cheap powerful servomotors?
(from https://spectrum.ieee.org/prosthetic-leg)
Yup. Exactly. The term for this is "Telechir":
https://link.springer.com/chapter/10.1007/978-3-642-93104-8_...
My kids have been around lots of robots of all kinds. The very first comment they made while watching the event was: "The robots are being remotely operated. There's no way that's autonomous."
Nice looking machines. Far from being practical outside of a highly controlled environment. This does feel like progress though.
Musk actually used this argument in his stock price manipulation trial, and the jury bought it.