robots0only

The next token prediction machine is writing 90%+ code at multi-trillion $$$ companies, its already solved millennial problems -- but yes its definitely not living up to hype.

What do you think Tesla has been doing this for so long?

This is incredibly sad, with google you can atleast use the argument that search was free (which is still weak IMO). But with OAI collecting indiscriminate data on paying customers is just incredibly sad.

I work at GDM and this is not at all true, 3.7 is markedly different (and better) than 3.6.

Any robot that does this reliably is easily more than a decade away.

From your early point -- both 1) and 2) are true. True human level dexterity is ver far (few decades surely), it would require further advancements in hardware, learning approaches etc. Recent approaches provide a glimmer of hope and maybe we can have some intermediate robots -- to be honest even waymo's and tesla's are robots and we will see much more of such robots with vision, working with humans etc. in narrow settings - chinese dancing robots are examples of this.

here is a real video of a unitree robot playing ping pong https://www.youtube.com/watch?v=tOfPKW6D3gE

how do you know this is a better model? I wouldn't take any of the numbers at face value especially when all they have done is more/better post-training and thus the base pre-trained model capabilities is still the same. The model may just elicit some of the benchmark capabilities better. You really need to spend time using the model to come to any reliable conclusions.

This is probably very similar to what happened!

the problem here is that text as the communication interface is not good for this. the model should be reasoning in the pose space (and generally in more geometric spaces), then interpolation and drawing is pretty easy. I think this will happen in some time.

How are you defining dextrous? I think it can be somewhat challenging but not dextrous -- the robot doesn't need to be very precise (few cms here and there do not matter), there are no forces involved, motions are all pick-place. Dextrous tasks would be things like shoe-lace tying, origami folding etc.

Locomotion and manipulation are pretty different. The former we know how to do well -- this is what you see in unitree videos. Manipulation still not so much. This is not at all like GPT-2 because we still don't know what to scale (and even the data to scale is not there).

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.

+100!!! Please don't fall for the HYPE.

The current best neural networks only have around 60% success rates for small horizon tasks (think 10-20 seconds e.g. pick up apple). That is why there is so much cut-motions in this video. The future will be awesome but it will take time a lot of research still needs to happen (e.g. robust hands, tactile, how to even collect large scale data, RL).

In all of these posts there is someone claiming Claude is the best, then somebody else claiming they have tried a bunch of times and for them Gemini is the best while others find GPT-5 is supreme. Obviously, all of these are subjective narrow experiences. My conclusion is that all frontier models are both good and bad with no clear winner and making good evals is really hard.

so their way to differentiate against frontier labs is to try writing research blog posts (not papers). It will be interesting to see how this plays out. I don't think that anyone serious about developing frontier models would be putting anything useful out there for others. We already see this with all the incumbents -- Google, OAI, Anthropic, xAI, DeepSeek and other chinese labs.

This paper was just too overhyped by the authors. Also, the initial evals were very limited and very strange. This blog post does a much better job at a similar observation -- goes into details and does proper evaluation (also better attribution): https://jinjieni.notion.site/Diffusion-Language-Models-are-S...

and so is the safety margin for a humanoid. The consumer market is huge only if the robots are highly reliable and work very well both of which are not true at the moment. Things will change but it will take quite a bit of time and much more research.

Claude is extremely poor at vision when compared to Gemini and ChatGPT. i think anthropic severely overfit their evals to coding/text etc. use cases. maybe naively adding browser use would work, but I am a bit skeptical.

The 1 million robot number that Amazon keeps on using is a quite nuanced. It includes more ~800K robots that simply just move stuff in a 2D plane. I think the number of robots that actually manipulate things is far far less (probably less than 500) (but really no human wants to just move things from A to B).

Also, I completely agree with what you said. Cars (w/ no self-driving) can be thought of as primitive robots (just like robots of today). For good or bad, we will move towards more and more automation.

ohh wow, that's bad, just tried this with Gemini 2.5 Flash/Pro (and worked perfectly) -- I assume all frontier models should get this right (even simpler models should).