Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Figure 03, our 3rd generation humanoid robot

figure.ai

Post reply on HN
Print view View original

111–120 of 417 posts

Re: Figure 03, our 3rd generation humanoid robot

#111

Post by refulgentis »

jppope wrote: post #105

Correct me if I'm wrong here but humanoid consumer robots are basically smoke and mirrors right now because the unit economics are so bad. Even the baxter robot was targeting small business because the lowest price point you could get was $40,000 for something that would have to work 24 hours to do something a human would do in an hour or two. Having a robot do chores at home is an even worse financial position. Anyo…

I do, I think.

It is breaking news if there is a $40K robot that had a 12:1 efficiency ratio.*

Because production is so nascent and small, cost doesn't mean too much, no ones scaled yet.

At only $40K capital investment, even a guaranteed 12:1 efficiency ratio would be an absolute no-brainer financially for many, many, wealthy people and certainly businesses. I do 1-2 hours of chores a day if I'm lucky. If I had the equivalent of a robot vacuum working 24/7 it'd do a much better job than me.

* The whole thing is written up and shown in a way that makes you think we're on the second refining release of a breakthrough**. I don't think they've gotten to the breakthrough yet - we would have seen > 0 videos from outside the company by v3.

** Really, the whole thing has an audience of one: Musk. (c.f. focus on fingers which was recently reported as the major pain point for whatever he calls their robot not making it to production; aping of Musk-y things like the factory itself a product)

Re: Figure 03, our 3rd generation humanoid robot

#113

Post by MountDoom »

lossolo wrote: post #104
ipnon wrote: post #98

Earlier quoted context omitted.

Now the question is if this is GPT-2 and we’re a decade away from autonomous androids given some scaling and tweaks, or if autonomous androids is just an extremely hard problem.

https://www.figure.ai/company "Building Figure won’t be an easy win; it will require decades of commitment and ingenuity." "Our focus is on what we can achieve 5, 10, 20+ years from now, not the near-term wins." At least it's not Musk's forever "next year".

> At least it's not Musk's forever "next year".

The problem with the principled approach to high-uncertainty projects is that if you slowly execute on a sequential multi-year plan, you will almost certainly find out in year 9 that multiple of the late-stage tasks are much harder than you thought.

You just don't know ahead of the time. Just look at how many corporations and research labs had decades-long strategies to build human-like AI that went nowhere. And then some guys came up with a novel architecture and all of sudden, you can ask your computer to write an essay about penguins.

Musk's approach is that if you have an infinite supply of fresh grads who really believe in you and are willing to work crazy hours, giving them a "next year" deadline is more likely to give you what you want than telling them "here's your slow-paced project you're gonna be working on for the next decade". And I guess he thinks to himself that some of them are going to burn out, but it's a sacrifice he's willing to make.

Re: Figure 03, our 3rd generation humanoid robot

#114

Post by drcode »

natemcintosh wrote: post #23

I would love to have something like this for all the chores around my house. But I also have serious reservations about the increased level insight into my private home life this could provide to the manufacturer. Look at cases like the ring camera security violations [1]. A moving robot could be an order of magnitude more invasive of your privacy. If I were to purchase this, I would want serious privacy guarantees.…

I would want the robot to be programmed to immediately leave any room a member of our family enters, always cleaning only the empty rooms

Re: Figure 03, our 3rd generation humanoid robot

#115

Post by kibwen »

ipnon wrote: post #98
HAL3000 wrote: post #78

All of the examples in videos are cherry picked. Go ask anyone working on humanoid robots today, almost everything you see here, if repeated 10 times, will enter failure mode because the happy path is so narrow. There should really be benchmarks where you invite robots from different companies, ask them beforehand about their capabilities, and then create an environment that is within those capabilities but was not u…

Now the question is if this is GPT-2 and we’re a decade away from autonomous androids given some scaling and tweaks, or if autonomous androids is just an extremely hard problem.

For LLMs, the input is text, and the output is text. By the time of GPT-2, the internet contained enough training data to make training an interesting LLM feasible (as judged by its ability to output convincing text).

We are nowhere near the same for autonomous robots, and it's not even funny. To continue to use the internet as an analogy for LLMs, we are pre-DARPANET, pre-ASCII, pre-transistor. We don't even have the sensors that would make safe household humanoid robots possible. Any theater from robot companies about trying to train a neural net based on motion capture is laughably foolish. At the current rate of progress, we are more than decades away.

Re: Figure 03, our 3rd generation humanoid robot

#116

Post by throwaway-0001 »

sksksk wrote: post #26

The use cases in their videos are interesting, I suppose the world we live in is build for humans, so it makes sense to build a robot that is human shaped. So we don't need to buy new washing machines and redesign our house to get a robot maid. The hotel reception use case seems ridiculous though, if you get rid of a human receptionist, you lose the human element of the check in process, which people like. If you're…

I personally don’t care about human receptionist. I rather get it faster

Re: Figure 03, our 3rd generation humanoid robot

#117

Post by robots0only »

ihumanable wrote: post #88
the_gipsy wrote: post #62

Earlier quoted context omitted.

Yea, the selected short clips are extremely suspicious.

They've got an hour long video of it sorting packages if you want a longer clip https://www.youtube.com/watch?v=lkc2y0yb89U They've shown the "putting dishes in the dishwasher" bit before, it seems to be getting better, but I imagine it still has a high failure rate. I wonder if this company started off or has some founder that's really interested in the "handling deformable stuff" space. They really seem keen to pro…

Here you can see another much simpler robot folding clothes for far longer: https://www.youtube.com/watch?v=gdeBIR0jVvU (there are more videos from other companies as well)

To answer your question -- folding clothes is easy, because clothes easily deform, do not break, fall smoothly when you drop them and most importantly are easily resettable task. Just through the well folded cloth up and voila start again.

Re: Figure 03, our 3rd generation humanoid robot

#118

Post by jcims »

ipnon wrote: post #98
HAL3000 wrote: post #78

All of the examples in videos are cherry picked. Go ask anyone working on humanoid robots today, almost everything you see here, if repeated 10 times, will enter failure mode because the happy path is so narrow. There should really be benchmarks where you invite robots from different companies, ask them beforehand about their capabilities, and then create an environment that is within those capabilities but was not u…

Now the question is if this is GPT-2 and we’re a decade away from autonomous androids given some scaling and tweaks, or if autonomous androids is just an extremely hard problem.

I don't know if I caught your comment in my peripheral vision or what but GPT-2 is exactly where I conceptually placed this.

Neural networks for motion control is very clearly resulting in some incredible capability in a relatively short amount of time vs. the more traditional control hierarchies used in something like Boston Dynamics. Look at Unitree's G1

https://www.youtube.com/shorts/mP3Exb1YC8o

https://www.youtube.com/watch?v=bPSLMX_V38E

It's like an agile idiot, very physically capable but no purpose.

The next domain is going to be incorporating goals and intent and short/long term chains of causality into the model, and for that it seems we're presently missing quite a bit usable training data. That will clearly evolve over time, as will the fidelity of simulations that can be used to train the model and the learned experience of deployed robots.

Re: Figure 03, our 3rd generation humanoid robot

#119

Post by ikeashark »

modeless wrote: post #73
ikeashark wrote: post #46

How does Figure 03 know if the washer is actually on? I know it has tactile sensors on its fingers but if it tried to press the button and missed how would it know? I know it's probably not by sound.

Why not? It has microphones.

I assume that with whatever onboard audio model they use, it can only handle conversational type audio.

Re: Figure 03, our 3rd generation humanoid robot

#120

Post by renewiltord »

causal wrote: post #14
dpcx wrote: post #2

Maybe it's just me, but everything about this announcement feels very I, Robot... and not in a good way. > allowing the entire fleet to upload terabytes of data for continuous learning and improvement Ugh. Edit: Yes, I meant I, Robot the film. U.S. Robotics and the like.

Yeah tech companies have a weird fixation on using dystopian literature as their entire branding playbook

It's just that sci fi authors try to see into the future and have to write things interesting. There's two ways:

- novel idea or technology

- counterintuitive effect of technology

I think the second is easier written as "what if Good Thing was actually Bad". So that's what you get. The former style is perhaps still available in books like Children of Time by Adrian Tchaikovsky.

But the latter style is much more readily written and consequently has dominated sci fi as more authors enter the field.

The Torment Nexus view is mostly driven by context blindness. "oh my god, they'll scan the mother's blood to perform eugenics if they have sequencing technology and it will be horrible". Well, advanced societies do that a lot: Down's is scanned for using a Maternal Serum Alpha Foetoprotein test. "oh my god, they'll use ultrasounds to find undesirable genetics, torment Nexus" but Nuchal Translucency tests are fairly routine in advanced societies and we're fine with them.

This might appear like a fixation on dystopian literature to others. "omg gattaca this MSAFP". It's just generic technoluddism because almost all near future tech is explored via sci fi in the "what if Good is Bad" genre.

Post reply on HN