Yeah. I didn't want to be rude, because real people building things they are passionate about is a great ideal. However, it is REALLY starting to feel like 1999. There is a new paradigm with clearly amazing potential, but we do not have all the pieces in place to execute on the forward thinking ideas, much less the businesses. Just imagine if that robot arm destroyed a ton of super expensive machinery. It was an extremely fair prompt, very clear and very simple instructions. Also, 90 Million to tweak an AI model to not achieve a task in an unvalidated market? IDK if that is too harsh and I just don't understand? There is 100% chance that myself or Dream-1 is out of touch, but I am concerned I can't tell which is which.
Approximately the same result as your run. But it was interesting to see that the idle arm seems to have these rhythmic movements across all the prompts I saw on there. Also that one arm, the left one most of the time, is just idling.
I wonder if the entire thing would be better if the first step was for the LLM to try to understand the final goal of the user, rather than just the simple instruction.
For instance, the LLM may have an understanding that the ball is easier for it to pick-up than the elephant (which it seemed to struggle with, which also made it a strange choice for the demo).
How much of an LLM understanding the physical limitations of the robot would help the task be accomplished in a way that satisfies the goal, without following the exact directions.
Interesting. It actually moved the ball next to the elephant, but I guess that accomplishes the goal of the elephant and ball being next to each other. But probably not how must humans would interpret that prompt. And it continued to move both the elephant and ball after that for some reason.
Hopefully this is like the first GPT models and it gets better over time.
This is all very cool and something I would like to play with. However I also can’t mentally bridge this over to the actual automation, manufacturing, and warehouse work I’ve been involved with throughout my career. Everything about this looks optimized for getting-started tasks to play around with, but I’m left feeling that it’s not the right fit or even a good starting point for real automation work.
Everything about this page strikes me as having been optimized for investors, not customers. Why would they cite a literal toy as a leading example of automation work? It doesn’t make any sense unless they’re maximizing for what investors remember seeing in headlines, not what actual automation customers would be interesting in paying for.
I think the product they’re selling is their equity, and their target customers are acquirers. The demos and the model feel like marketing material for that highly specific goal, but they forgot to include enough about how it could be useful to anyone.
Obviously AI written text. At least there is real content here, however I do find the order and the way in which it was written and portrayed to have a heavy AI smell. That is, it's hard to read and follow the clean narrative, it isn't written the way a human who was trying to convince another human would write, and is less effective because of it.
It's weird to spend all this time making a product and then just throw up something you had a LLM regurgitate. Any amount of human written content here would have been better, even if scribbled on the back of a napkin to show that the product had anybody care about it. A blog post, post-LLMs, can exist entirely in a vacuum away from anybody believing in it.
Building products is a different skill than writing about them. And if you write just "back of the napkin" people will complain about it even more than about an LLM version. Writing polished content is hard and takes a lot of time and skill.
People used to complain about back of the napkin posts because they signaled low effort. The same is true about LLM blogs. A back of the napkin blog signals more effort than slop.
A back of the napkin post means at least one person had interest in its authorship to put down their drink and write something, the LLM made blog posts are a show that zero people had interest in its authorship enough to put down some words.
I didn’t say it was fake, I said nobody cared enough about what they were making to write a paragraph saying why. Being excited about your product, what if can do and why you made it is very different to promoting “make a blog post about this in a boring corporate style” and going to lunch.
This post feels really weird, DreamLabs isn’t some well known project like Anthropic or OpenAI and Dream-1 wasn’t really an awaited release.
But somehow this was pushed to the front page of HN before receiving a single comment with a title reminiscent of a major LLM release. Like I can only find 1 other reference historically to this lab and nobody noticed it or cared.
It does seem weird. In my experience HN generally shows little interest in robotics research, so I'm surprised to see this up top. I'm biased but I think Physical Intelligence's recent announcement is more impressive, with greater diversity of robot embodiments and more impressive real-world tasks in real workplaces. https://www.pi.website/blog/api-beta
This one as I read it is a bit different, the Physical Intelligence API requires finetuning for the task/robot while the Dream-1 model offers an API that can do zero-shot (with great limitations of course but still cool).
If you look at what they actually demonstrate, zero-shot embodiment generalization is limited to controlling two-finger grippers on two stationary arms. The embodiments are all very similar, and when they're slightly different (even just replacing two fingers with five fingers) they can't do zero-shot anymore. But there's no real reason to do zero-shot embodiment generalization in practice, as it's trivial to collect a small amount of data on a new embodiment to get significantly improved performance. Zero-shot embodiment generalization, especially in such a narrow domain of similar configurations, is not an important problem to solve and doesn't necessarily indicate progress on the important problems.
Zero-shot task generalization is more interesting, but has been demonstrated by many other companies at this point, so it's not new. And it only becomes useful with high success rates on long-duration tasks, so demonstrating short zero-shot tasks with low success rates in toy environments is uninteresting.
What's hard and still unsolved in general-purpose robotics is doing real useful tasks end-to-end in real uncontrolled environments with high reliability. I guess there's just a lack of understanding generally about what's actually hard and useful in robotics.
I discovered the Robocurve benchmark after someone on X was saying that Kimi K3.1 Pro was performing particular bad on it.
Anyway I found the model particularly interesting because I think it's the first time I see an API for a random bimanual robot. Also HN removes "Introducing" so it's just the name of the model.
Sorry I don’t mean to specifically point you out, I believe you’re being genuine.
However this comment got me thinking. Whats up with all the YC backed benchmarking companies? I know there’s only a few, but what is the business model for companies like Robocurve exactly?
I came across this recently and was initially surprised too. It makes sense when you consider that any company trying to build models (or an agent/value-add layer on top of a model) needs benchmarks. The best funded will do the full vertical - build their benchmarks and models together - but many will outsource the benchmark. This is particularly valuable when the dataset is difficult to compile (robotics being a prime example) - it makes more sense to let the benchmark vendor focus on that.
Sure it’s useful, but I’m asking who actually pays for this. Sort of a conflict of interest if you’re providing public testing for the companies paying you for the tests no?
Perhaps it’s purely for VC so they can judge which model is best? Also seems a little too conflict of interest-y, but that might just be too cynical.
You think that claims of an actual honest to god working robot is somehow so out of sync with this audience that we need to be so cynical we call it a conspiracy? Am i literally the only one who's wanted a Robbie of my own since i read i, robot as a kid?
There are PR agencies that take care of your content to be trending.
They do use fake accounts to upvote and manipulate, but you as a customer don’t have to know that.
Occam's razor would suggest that mods simply thought that it would generate insightful discussion, and manually pushed it up, something that dang regularly does and a core feature of HN -- and that is assuming it was pushed up in the first place.
I'm certainly not a mod or affiliated with YC and I've had random things I posted briefly end up on the front page. Then if people keep upvoting it, they sometimes remain there...
Wouldn’t it just be easier to ask their team to upvote the post? They probably all have legitimate hacker news accounts already and are happy to upvote their work.
dumberquestions 13 hours ago
https://dreamlabs.ai/playground/?play=a05f0e641559490e#sim
wand3r 12 hours ago
mapkkk 11 hours ago
Approximately the same result as your run. But it was interesting to see that the idle arm seems to have these rhythmic movements across all the prompts I saw on there. Also that one arm, the left one most of the time, is just idling.
pedalpete 10 hours ago
For instance, the LLM may have an understanding that the ball is easier for it to pick-up than the elephant (which it seemed to struggle with, which also made it a strange choice for the demo).
How much of an LLM understanding the physical limitations of the robot would help the task be accomplished in a way that satisfies the goal, without following the exact directions.
ajjenkins 12 hours ago
Hopefully this is like the first GPT models and it gets better over time.
Aurornis 13 hours ago
Everything about this page strikes me as having been optimized for investors, not customers. Why would they cite a literal toy as a leading example of automation work? It doesn’t make any sense unless they’re maximizing for what investors remember seeing in headlines, not what actual automation customers would be interesting in paying for.
I think the product they’re selling is their equity, and their target customers are acquirers. The demos and the model feel like marketing material for that highly specific goal, but they forgot to include enough about how it could be useful to anyone.
hdz 12 hours ago
Folding laundry seems like one but then quickly falls apart the more I think about it. Not sure what else.
ethanj8011 13 hours ago
drum55 13 hours ago
jampekka 9 hours ago
Routine8502 8 hours ago
drum55 6 hours ago
fn-mote 13 hours ago
drum55 6 hours ago
besterman23 13 hours ago
But somehow this was pushed to the front page of HN before receiving a single comment with a title reminiscent of a major LLM release. Like I can only find 1 other reference historically to this lab and nobody noticed it or cared.
smallerize 13 hours ago
modeless 12 hours ago
GaggiX 12 hours ago
modeless 12 hours ago
Zero-shot task generalization is more interesting, but has been demonstrated by many other companies at this point, so it's not new. And it only becomes useful with high success rates on long-duration tasks, so demonstrating short zero-shot tasks with low success rates in toy environments is uninteresting.
What's hard and still unsolved in general-purpose robotics is doing real useful tasks end-to-end in real uncontrolled environments with high reliability. I guess there's just a lack of understanding generally about what's actually hard and useful in robotics.
jere 11 hours ago
fwlr 12 hours ago
schleck8 10 hours ago
cyanydeez 10 hours ago
soundworlds 9 hours ago
There are huge numbers of new posts - the ones that reach front page are definitely not the only ones with that amount of upvotes.
nine_k 10 hours ago
GaggiX 13 hours ago
I discovered the Robocurve benchmark after someone on X was saying that Kimi K3.1 Pro was performing particular bad on it.
Anyway I found the model particularly interesting because I think it's the first time I see an API for a random bimanual robot. Also HN removes "Introducing" so it's just the name of the model.
besterman23 12 hours ago
However this comment got me thinking. Whats up with all the YC backed benchmarking companies? I know there’s only a few, but what is the business model for companies like Robocurve exactly?
computomatic 11 hours ago
besterman23 11 hours ago
Perhaps it’s purely for VC so they can judge which model is best? Also seems a little too conflict of interest-y, but that might just be too cynical.
dmix 10 hours ago
that would be an interesting business to run.
halJordan 6 hours ago
tgrowazay 13 hours ago
jelloroll42 12 hours ago
CamelCaseName 12 hours ago
jelloroll42 9 hours ago
walrus01 12 hours ago
jelloroll42 9 hours ago
computomatic 12 hours ago
bertt 11 hours ago