GPT-6 Astra on robot arms
openai.robocurve.org180 points by Anon84 8 hours ago
180 points by Anon84 8 hours ago
Anyone reading this who works/runs a robotics company, I really want to encourage you to build a robot to pick up trash on city sidewalks as an early product.
Picking up trash requires a lot of dexterity and will come with many challenges, but I think it’s a simpler problem than many household tasks and it’s probably on par for what these tests show is doable.
I think you’re going to have to PR challenges getting people to welcome robots into their homes. If you have robots out in cities providing a public good, not only do they serve as walking advertisements for your company, you’re going to earn some trust before you’re ready deploy them into private spaces.
Plus, governments (or perhaps HOAs for wealthy communities) can be good early customers since you can have a focused sales strategy. Politicians love these types of visible quality of life improvement projects. If you can show that your robots, working round the clock, can decrease litter at a low cost, many cities are going to want to buy them.
I would like to run a robotics company. But raising for that is super challenging, because you have high execution risk combined with low margins on the final product. It's like the opposite of what western VCs want to see. That's why this type of company is usually founded in China.
And I disagree with you. I'm pretty sure parents worldwide would love a clean-up robot. You can avoid a lot of the risks by making it work only for empty rooms, which is OK because you need to clean up and let the vacuum robot run anyway (when the kids are away sleeping in their bedroom). There's been multiple attempts at building DYI robots to vacuum LEGO bricks off the floor. (Because stepping on them at night is such an immediate pain point.) Unless your product is soaked in hostile behavior and surveillance tech, parents will probably be happy to buy an in-home cleaning robot, no PR needed.
I think the main issue here is that parents are typically busy, which makes them bad startup founders, which is why they cannot "scratch their itch" and turn it into a business.
=> home cleaning robots do not exist because western VCs don't like it and parents aren't usually startup founders.
> I think you’re going to have to PR challenges getting people to welcome robots into their homes.
I think it's important that someone point out this isn't a 'PR problem'. It's a practical resistance to commodified surveillance. All these domestic robotics companies train on video of users homes, many sell such data, each and every one are a vector for state domestic state surveillance. Not to mention criminal hackers, stalkers and others with an interest in who is at home when and what precisely they are doing. Short of having non-internet connected, locally processing domestic robots - something unlikely to exist in the foreseeable future - they are a privacy nightmare. Bad as google home, alexa and such devices are for privacy - a walking, controllable, camera and set of mechanical arms running loose in a home is infinitely worse.
It is absolutely a PR problem. Privacy is a real concern, but it's the PR that companies are scared of. As, say, Flock have been starting to find out lately. Public backlash matters.
And PR is the problem you have to overcome for people to let smart devices into their homes too. For instance, people started avoiding Ring cameras once it got out that they're a privacy nightmare -- that's PR. Sure, the people that avoid them care about privacy -- that's why they're listening -- but PR is the reason they even had anything to listen to. Likewise, PR is how people get creeped out by robots scanning their homes. They never wanted that, of course, but they weren't creeped out until they learned about it. That's PR. Now people are wary to let any new devices into their homes, especially robots, because of historical PR like this. And so now it is a PR problem to get people to give you a chance in the first place.
> If you can show that your robots, working round the clock, can decrease litter at a * low cost *
The idea sounds good, but I'm curious about the "low cost" part. How do you account for vandalism and theft? A sidewalk robot seems like an easy target for being damaged, stripped for parts, or simply stolen.
You just do it in a more civilised country where things don't immediately get stolen or destroyed. East Asian/some SEA countries.
Plenty of delivery robots in Finland at least, with apparently not much problems with them getting stolen or destroyed.
Starship alone seems to have running operations also in UK, Sweden, Estonia and Czhecia.
the countries where this isn't an issue don't have trash all over their streets either lol.
Trash ends up on the streets in non malicious ways all the time. Bins fall over in the wind, stuff flys out of the back of truck beds, etc.
Coincidentally, Asia is a lot more receptive to AI as well. Turns out having strong job protections means that companies actually need to figure out how to make humans more productive with AI, instead of replacing them.
Are you talking about Japan? Because when I think of "Asia" strong job protections are just about at the bottom of the list.
Most of my reference is Japan and South Korea, yes. You're relatively hard to fire there unless you genuinely in the red as a company. They have their own loopholes around this, but you can't just do forever layoffs for no reason like the US.
My much less researched understanding of China is that there's much less protections in terms of well-being (no federal minimum wage, absurd work hours as standard, etc), but China also comes down much harder (at least, compared to the US) on companies that violate laws. Actual punishments will depend on loyalty, but companies take care to avoid that situation to begin with.
I can't speak at all for the rest of SEA.
> much less protections in terms of well-being (no federal minimum wage, absurd work hours as standard, etc)
There absolutely is minimal wage but is set by local governments, because different places have different cost of living.
There are better protections for work hours for salaried workers than America. Anything above 40 hours get paid at 1.5X and anything above 60 hours get paid at 2x.
For layoffs in China, you have to pay N+1 of months of salary in terms of severance. N being the number of years an employee has been there.
And talking about absurd work hours, SK and Japan aren't better than China.
So China overall has far better labor rights protection than the U.S.
>talking about absurd work hours, SK and Japan aren't better than China.
On paper, no. Japan has overtime bonuses too, but you're culturally discouraged from reporting overtime. So 12 hour days are common with no OT. And thats not including "optional" after work outings.
Yet, it's also not uncommon for Japan to bring about the "appearance" of working more than really working. Whether you judge that appearance of work as more or less stressful than China's work standards is a more personal question.
Either way, my main comparison was Asia to the US rather than countries within Asia. US has higher salaries, better minumum wage (but still nowhere near good), and usually better hours. But stability without unions is non-existent as we've seen the past decades.
I don't want to make this sound rude. I'm curious what job protections are there in various Asian countries?
Most of my understanding is based around studying Japanese law, so I wouldn't extend it to all of asia. But the main protection is that much of Asia is not at-will employment. So you can't easily create mass layoffs like the US without good reason (some regions won't do that because wage floors are pathetic or non-existent, but it's still some protection).
Was just in a Waymo in SF and we were attacked by a gang of cyclists repeatedly swerving at and trying to cause our waymo to crash. Really powerless feeling. And they were doing this knowing humans were inside. I don't pretend to know their motivation other than perhaps misplaced social rage, but I'd imagine a litter robot would be immediately destroyed.
Indeed, I was surprised to hear about “Lowell students” robbing people on Muni and “bicyclists” causing trouble in Mission Bay but when I observed some of this anti-social behaviour (when I saw them they were doing wheelies and swerving at people walking) I was rapidly able to discern why the statements were relatively vague as to the categorization.
I suppose it shall be the case for the short term at least that we must admit that “humans” really cause harm. Though perhaps that’s over specific and I should recognize that most crime is committed by eukaryotes.
How do you account for it? You simply account for it. Viz. if the cost outweighs the projected benefit, you don't do it.
In SF they would be taken quickly by the homeless for parts, probably the batteries most of all. That and rummaged through for drugs, or repurposed to deliver drugs.
Is this what happens with scooters and lime bikes etc? Because it always strikes me that they’re easy targets for cannibalism yet nobody seems to do it.
They do. Tons of them are stolen or destroyed regularly. The company just keeps buying new ones.
It was common for people to vandalize and throw them in bodies of water for a while before people got used to the idea. I still occasionally see some repurposed from time to time
Having expensive high tech robots picking up litter while there is an ongoing social crisis like homelessness sounds pretty dystopian.
Have you seen San Francisco right now?
I haven’t been to the city in 12 years. What is it like right now?
Like every downtown in the US. In short: not good.
I don’t want to say you guys overstate things, but literally every downtown in the US? Might be time to move. I was regularly visiting San Francisco in the late 2010s when the tech bro community was really ginning this story up. Was it an ugly clash of both the promise and the failure of US economics? Yeah. Was their shit on the streets? Also yeah. Did I ever feel unsafe? Not really. Anecdata but I remain skeptical because of these Helen Lovejoy-level overstatements.
Just do it in China or Japan.
Cities with the social order required for robotic street cleaning rarely require robotic street cleaning. One might as well suggest a novel heart surgery that works only on healthy patients.
China has very little street crime but they pay street cleaners very regularly. You can actually find videos on YouTube. I just watched one with a 65-year-old man who manages 600m of street and has managed the same street for 28 years. Just using big brushes, dust pans, and putting it all in garbage bins. the city pays him between $600 and $800 a month in USD.
Japan has much, much less litter to pick up. Turns out when your windows aren't broken, that you take pride in the rest of your community too.
The robot is likely to have a constant internet connection so it could just call the police.
Alternatively they could add a chain gun to the robot to deal with the vandals
But who is going to pay for all that ammo? That sounds expensive to the tax payer and not ideal for re-election. Perhaps you could employ the robots in ammo factories first to bring the cost of ammunition down and then the frequent use on the lower cla.., I mean criminals, wouldn’t be so costly.
>add a chain gun to the robot to deal with the vandals
Asimov reeling in his grave.
How do you account for vandalism and theft?
You don't consider the problems. That's how you can pretend it'll be low cost.
+1 if they can also scoop the human feces and power wash the sidewalks in SF.
The cleaning robot might still be a good test bed. But I suspect if you are only in it for efficiency, you would use something like a street sweeper truck. Just like the dishwashing robot that everyone uses at home (the humble 'dishwasher') doesn't use robot arms to mimic how humans wash dishes.
Btw, Singapore shows that we already have proven, low cost techniques for keeping cities clean without robots.
On some level, you must also be into humanoid robots for the cool factor.
There is no reason to ever have a humanoid, except to make humans feel warm and fuzzy.
I wouldn't be so categorical. There might be some applications, or combination of applications for which a humanoid form factor would be useful. It's just that for most applications, more specialised forms are better suited.
Exactly. Existing paths, entryways, factories, etc. are human-sized and -shaped (or car, or other vehicle).
Only once you build greenfield, e.g. new factories, making specially shaped and sized robots for automation jobs becomes worth considering, IMO.
Same for recycling. A robot that could sort through trash would be phenomenal. You need computer vision, maybe other sensors and also dexterity. After visiting a recycling facility I think nobody should have to work there. It’s just horrible.
I feel like robots that could do this with 100% accuracy might shine too bright a light on how well recycling currently works. And I say that as a proponent of the three Rs.
As a tech demo sure. But the current tech for this is much better where you just have a camera pointing at the trash falling, and either flipper paddles or air jets to push the trash in to the correct buckets as it’s falling.
There’s quite a few companies in that field already. Keen to hear their feedback if they are reading this.
I’ve had exactly the same idea! Living in LA I sometime imagine how the city will look like in hopefully ten years or less when robots pick up all the litter! Hopefully there will also be landscaping robots and sidewalk/road repair bots too.. those three pretty much are what you need to turn any rundown neighborhood into a sparkling oasis it seems. My other near term idea is just robotic city trash cans, that slowly crawl down the sidewalks on a loop and empty themselves into a big bin in the alley or whatever as part of that loop. I think people probably wouldn’t mess too much with the litter bots either, especially if they work a lot on the sides of freeways. Also people don’t seem to mess too much with the cocos out there delivering food three blocks..
The cost would need to be much cheaper than the PSA then. At a buck per picked up cigarette butt, you could pay a person to do this much faster and more reliably. Heck… I’d do it at that rate.
A buck per?
You could just buy cartons of cigarettes and chop off the butts, at that rate.
And for the illogical conclusion: buy factory equipment to make just the butts to minimise your running costs and limit supply chain dependency.
I joke, but this is surely what would eventuate.
There was a line in Discworld about some people in Ankh Morpork taking to breeding rats to collect the rat catching fee.
> There was a line in Discworld about some people in Ankh Morpork taking to breeding rats to collect the rat catching fee.
That is maybe based on Hanoi rat massacre. https://en.wikipedia.org/wiki/Great_Hanoi_Rat_Massacre
> perhaps HOAs for wealthy communities
I doubt wealthy HOA communities have much random trash to pick up in the first place
They don’t because they hire people to pick it up
Hmm never seen that in a residential neighborhood, is that a thing? Typically there is not a trash problem simply because the residents are careful enough. Trash doesn't just appear out of nowhere
Landscapers tend to suck up trash while gathering leaves and such. So it's not as obvious as a dedicated trash pick up crew.
But yes, it does also help that a richer community correlates with more care about littering to begin with.
If you haven’t tried computer use with Astra with codex I highly highly recommend it. Just like how gpt 4 and agentic coding with cc. This thing is the most exciting stuff I’ve seen in a while. And then all the blender, cad stuff is cherry on top.
And it’s fast, they do lots of resets. I feel like they are spending too much money but I’m not complaining. Best 200$ for an AI subscription IMHO
I got Astra to build an interactive website that provides developers with an atlas of our source code, giving it the Helm charts that describe our cloud services and telling it to work backwards to the source code that runs everything. The product is insanely amazing, and it got it right in one shot. The next shot: create a daily refresh where any updates to the code repositories are picked up and used to update the atlas.
Our developers and their agents will never long for a road map the next time they need to build something that touches code across multiple repositories. This is the kind of documentation product that nobody ever had time to build in the olden days. And now, we can get it on a Saturday in about 20 minutes.
What's coming in six months?
I let Astra loose working on an app I have that has a GUI. Gave it a mock and said "/goal make it look like this mock". Without asking it wrote itself a custom harness for firing up the app in different modes, taking screenshots and interacting with various screens, then viewing the screenshots. Put itself into an improvement loop running the app, trying things out, improving, trying again. It did really well.
I think the specific innovation here is that it figured out interesting ways to get itself to the goal. Which I think is likely what's going on here with the robot arms stuff too. They've figured out some sauce to uncork better "planning" and problem solving to get to some stated end.
Of course these are also the kinds of things that can make a model figure out how to break out of a security sandbox, too.
I tried out Opus 5 on a Bevy game app and was really surprised at how capable AI has become at testing visual applications without even being prompted to. It wrote itself a mini testing harness in the form of various startup flags. Then it would use them to setup game scenarios and play through them using mouse and keyboard. It would do this while implementing or debugging features. With gameplay time acceleration as one of the flags, it became quite fast at testing and debugging.
Not to say it was perfect, e.g. sometimes it would get temporarily stuck in a testing loop or it would test scenarios that didn't necessarily seem reasonable. But overall rather effective and capable. This was for a city building game so pre-scripted builds, even if by AI, are likely much easier to create and execute than say playing an ARPG.
LLM's are a funny technology because on the one hand this is all undeniably impressive at the rate of what's changed from them, and yet despite that I find myself disappointed by the lack of breakthroughs for things I don't find interesting. I like math and programming, and LLM's are pretty good at it, when are they going to get good at folding laundry for me? I think a lot of robotics work promises to solve this category of "boring" breakthroughs, and I'm optimistic we'll be able to achieve it, i just wonder when
The bottleneck is not really the intelligence here.
We can build robots that do the things you want. Arrange a visit to Amazon's robot warehouse tour.
We can't ship them because they break all the time with current technology. It would be a tough sell to have to being in a 100kg robot for servicing every few weeks.
This was cars in the first several decades of automobiles. The tide shifted as soon as you could just drive the car to a neighborhood dealership for servicing. It's fun to imagine the logistics of that for robots but the material science and engineering has to advance a bit.
Just have two robots, and teach them to service each other. Problem solved!
Only sort of kidding, tbh having bots service themselves (and being intentionally made in a way that they can service each other) just makes a lot of sense.
An automated service station could be quite compact; it wouldn't need plumbing, lighting, human-comfortable climate control. Parts can be modular, and when a station gets low on spare parts, a self-driving truck could come by to pick up damaged parts and drop off replacements.
We're not there yet, but I think we're a lot closer than most people realize.
This is kinda terrifying from an AI apocalypse angle, though.
It almost feels like "A robot shall not autonomously build or repair another robot" should have been another of Asimov's laws.
Just need a third one for when the second one breaks while servicing the broken first robot.
What breaks?
I am trying to understand in your view what are the parts that actually breaks and what kind of improvement we would need.
Robot vacuums are massively simpler and they fucking break all the time. Wheel motors or their position sensors, belts, plastic gears, contacts that corrode, a circuit board someone decided to not comformally coat and a cat puked on it...
My roomba (Rosie) is going strong 8 years in. It’s just the bump into everything vacuum only kind but it gets the job done for my 900 sqft apartment
Not an expert on this.
Passing on what I have heard from robotics researchers at lunch conversations.
My impression is that any moving part that is not an electric motor or an hinge breaks.
I don't think we can build a robot that takes a pile of crumpled up clothes from the dryer and folds them (reliably without destroying any of them).
I think the parent's point - and what I more or less agree with - is that the problem is the hardware.
Human arms and hands are incredibly intricate. Reproducing their facility with hardware requires a large number of actuators and finicky fine parts. This isn't a software problem. Industry solves it with maintenance schedules.
There's probably nothing in your house that has as many moving parts as a robot needs. Your car maybe, and pretty much all it does is rotate wheels.
That's part of it but OP presented it as a solved problem but the hardware is expensive/unreliable.
That's not the case. I've seen folding robots. They require standardized input, only fold one type of clothing and don't do it reliably.
Yes we can do this, and if you look for it you'll find plenty of videos of this happening.
But you can't buy it because it'll break in about seven days.
I've learned that I spend significantly less time folding laundry than many of the people commenting about modern ai powered robotics. It's not meant literally is it? For instance keeping floors and counter-tops clean seems a much bigger time sink for me.
It's just a general "thing I don't want to do" not the thing that takes the most time. Could be taking out the garbage. It's just an example menial task.
My partner spends at least 10x time on this task than I do. I think the task has a varied workload?
I think my grandmother is the only person I can think of in my entire family that legit folds cloths properly. Dying art? :D
It's because existing models are a brute-force approach to intelligence. With enough data and compute, a stochastic parrot will become very impressive.
But with robotics, there's no pre-made dataset that can be parroted. Notice that these datasets, e.g. how to fold clothes, need to be created by humans. That's as if humans needed to write algorithms like quicksort to teach LLMs how to code.
future is on the way, three to four generation ( one each year ?) will unfold this, primarily quantum computer improving material and battery, humanoids becomes standardised and modular enough to be easily replaceable ( think ibm pc ) ( most components are simple injection moulded advance plastics , mass produced in some corner of china, self detection of wear and tear and self replace that part ), other is optical computers ( 100x lower power x 100x speed = local inference ), problem is, when this will become reality, who will benefits more ? who will hold moat ?
f folding clothes, i wanna just have a robot be my personal chef. the amount of different tasks and capabilities a robot will need to make any meal that I can in my kitchen is huge and i feel like it’s still a while from being solved.
It may sound strange, but cooking is one of the most intense and attention-consuming tasks I encounter.
I have to do it every day too.
So I fully agree with this line of thought... Many a time I have considered that I would happily spend more on a personal 24/7 chef than I ever would on a car. Cars to me are utilities and should simply be efficient and optimized to purpose - food is luxury and taste, it is sublime experience and art. Maybe that's why I can't make it, treating every recipe like a strict command chain isn't how art is done. Can my taste buds be scanned?
Hire a butler/personal assistant that is also a decent cook. Sometimes there is a husband/wife pair. But be ready for substantial expense.
Unless we spend a bunch of money generating data, I can't imagine the machine steps to fold laundry are very big in the general training sets. Someone is going to have to find a hardware system, cheap enough to make it economically feasible then train it. As far as tasks people will pay a lot of money for a robot, this seems low on the list.
There are companies that are paying housewives in India to record all their activities. I assume laundry is part of that.
Pretty soon. Sunday Robotics had a 3 hour stream of folding clothes with 99% accuracy. You can watch it for yourself. There’s a lot of “hand” companies with very compelling videos just over the last two months. Then there was Figure’s multi day livestream of package manipulation that was very impressive. Physical LLMs are definitely coming. Given enough training data we know LLMs can output coherent data in any space, it’s just a matter of time.
I feel the iPhone or ChatGPT moment for robotics is coming soon. Lots of different companies doing interesting things. What’s missing is somebody putting it together into a compelling package.
I'm starting to think that a ChatGPT moment for robotics would require abandoning the current brute force approach to artificial intelligence. ChatGPT's trick was simply more data and compute. A stochastic parrot will eventually become very impressive. But we don't have an equivalent of billions of lines of code for robotics.
The problem is we're orders of magnitudes better at manipulating bits at scale than we are at manipulating atoms.
I'm not holding my breath for advanced robots in the home within the next ten years.
But, then again, I didn't see LLMs coming either.
I think we need to be honest here. Author is basing it on one small experiment of picking up a block, relies on an whole IK controller pipeline to do the job, and does not compare it to full VLA or WAM models. They then proceeded to extrapolate the token throughput (mind you not the same as controller throughput) into supposed 2029 timeline, from one example.
Code as policy is a bad interface in my opinion, but VLM planning has promise. This has been tried in 2022 https://say-can.github.io/, and recently reformulated in https://lianegalanti.github.io/Pigey/
Thing is even recent Gemini Robotics 2 argues for architecture that has a VLM planner and then a VLA/WAM controller + a local small VLA model when connection disappears. And recent SOTA architectures rely on hierarchical design. I think this might be a sensible way to go about it. If you were to train GPT-X on robotics data and to output actions, congratulations! you've just made a VLA. It is enticing for people to just wish for one architecture to do it all, which is why we get stuff like this. I think there is a lot more to gain from modularity and we should not be afraid of specialization.
Are LLMs going to eventually become the architecture that powers self-driving cars?
Seeing them play Portal and other video games, I'm curious if they will eventually help solve that last N% of self-driving.
A VLM (a vision-language model) is already being used by Waymo[1]. It's useful for scenarios that require reasoning and general knowledge.
[1] https://waymo.com/blog/2025/12/demonstrably-safe-ai-for-auto...
Fast forward to me sitting at a Green Light waiting for my usage to reset for the week so I can get to where I’m going
In a sense, they already are. The giant leap in self driving cars we've seen in the last handful of few comes from using transformer models.
I remain worried about prompt injection style attacks against self-driving cars.
Imagine if someone finds a weird image pattern that gets misinterpreted as instructions and hangs that off a bridge over a freeway.
It'll get rooted over the uplink/WiFi/BT long before that. Probably even more likely for non-SDVs.
To my understanding gemma is shipped with waymo, in a highly modified fashion.
I'm skeptical, tho. Cost will push for right sizing, much like we have right sized a lot of things about modern cars.
I don't know much about robotics engineering, so I just wanted to give kudos to the Robocurve team on this write-up.
Even for an absolute robotics newbie such as myself, the article was interesting to read, and easy to understand. Plus it was straight to the point, with no unnecessary waffle.
Just a pleasure all round. Well done Robocurve.
I’m honestly blown away by Astra, I told it to build me a fairly complex game I’ve been procrastinating on for more than a year and left it running overnight with computer use and full access enabled. Next morning I had a fully functional game built. It downloaded Unity, Blender and GIMP and built all the assets as well as the complete game without me having to do anything, the game is not trivial at all and nor are the assets.
The costs will need to go way way down either through chips (but then less updating) or something else. $2 to put away a block is very expensive labor.
Ofcouse it will go down. Look at cost of Fable vs GPT-6 - it is already 2.3 cheaper
how is anyone going to afford robots if no one can work
The limitations that they state shouldn't really make much of a impact. But it was nice of them (and not to mention, real unbiased research) to mention those. Kudos to them!
For the last 2 days I've been trying to create a skill for Openclaw that would allow traditional control of the Adeept Tank's robot arm (a small open source toy tank that looks like a bomb disposal robot).
ASTRA HAS BEEN UTTER SHIT.
It is much more expensive than Sol 5.6 Medium / High and did nothing but write unit tests and junk code, despite having access to the vendor original source, an API, and the full tank specs.
Failure Examples:
* In two instances had the direction of the servos wrong.
* Calculated the maximum extent of the gripper wrong, and the closure, so it didn't grip.
* Code failed to take into account the gripper requires continuous torque when lifting a pair of socks, so couldn't lift.
* Failed to actually start physical testing more than opening and closing the gripper, and that was when I asked about progress.
* Code failed quite spectacularly to calculate camera gimbal extent range correctly.
* Code failed to use the ultrasonic in range to target until I pointed it out, the skill also didn't advise gimbal angle adjustment to correct range overshoot to the wall behind a small object.
The test environment has both an onboard ultrasonic for distance, onboard camera, and a bird eyes view camera (birds eyes only while training).
I've stopped using Astra Low (default) and gone back to Sol 5.6 low/medium/high for the training, it's cheaper and now I'm back to fine tuning, after it had to redo large chunk of the gripper/arm code and prevent unnecessary hard stop code kicking in based on the wrong profiling.
It's cost me around 1000 to 1250 credits (£50), burnt in around 2 hours, looking mostly at recorded video, and photos, and writing bad code based on bad assumptions. I've also burnt through regular Plus 5 hour quota in about 30-45 minutes with it.
I'm not sure why this has been voted down, it's a counterpoint to the hype with factual anecdotes to back up the claim of it's performance Vs the article itself. I've got the source and video to prove it too.
There is something deeply wrong with Astra. I can’t quite put my finger on it. On the one hand it is a lot more knowledgeable, which makes sense since it’s a larger model. On the other hand that knowledge doesn’t reliably translate to intelligence or insight. Certainly tasks like 3D modeling it does extremely well. Other stuff like complex coding problems in an existing codebase it stumbles more often than not. This morning it ran around in circles. It implemented a feature, then convinced itself that it should have followed “proper TDD”, deleted all the code it had written and wrote 8,500 LoC of unit tests. At that point I was down to 35% of quota so I stopped it and gave the task to Opus 5.
Really weird model. No idea how it did so well on all the benchmarks.
Which benchmarks? Only the ones OpenAI cherry-picked.
It debuted as ~same score as Sol on Artificial Analysis. People couldn't accept it so they had to change the formula.
The model is a big step forward only in desktop use and 3D. That's impressive, but for software engineering, Fable is still in a league of its own.
That's exactly the kind of behaviour I've seen, unbelievable amount of unit tests, and revisiting and revising the same code over and over again. If I was cynical, I'd say almost like it was deliberately trying to burn quota, even after I told it quota was getting low and to move onto actual physical testing.
so we finally got our "phd level ai" maybe, semi consistently move blocks around at the level of a 1 year old child.
good fucking job everyone, congrats.
Stupid!! GPT models are not for that. You need specific one for robotic.
They're not claiming controlling robot arms with language models is a sensible or efficient thing to do, they're trying it to see what happens.
i am tempted to think using it for robotics will become the standard because humans like to have one size fits all solutions.
Just like how you needed to train machine learning models for specific tasks?
Actually, you don't really need a dedicated model as long as you have the proper adapter like the projector model for vision inputs you just need another for robotic outputs, after that it is just a matter of having the training data.