[NalNezumi]

I have high respect of Tuomas and his work around SAC for RL in robotics.

But this is slightly unconvincing, most because of the author

>They spend thousands of computer hours practicing movements inside simulated worlds and inventing their own solutions.

This is exactly what almost every other picking startups have been doing for the last couple of years.

I can think of at least a dozens, some even making their custom gripper hardware. It still relies on sim2real transfer and then there's a bottleneck of things such as representing deformable objects. And that's still just scratching the surface of it.

I can definitely see that they have the right team. But the claim made by this author is far removed from the actual demo he describes. I've seen same demo for years, last one was in CoRL by Google (Gemini) and even then you could see clever robotics guy (some Boston dynamics engineers) that came by and gave it a clever task it failed on.

[Joel_Mckay]

Universal Robots ( https://www.universal-robots.com/ ) force sensing collaborative platforms were very advanced years ago, but like most bot firms small market demand made retail consumer pricing unsustainable.

>I can think of at least a dozens, some even making their custom gripper hardware.

The simplest solution sometimes is more robust in practice:

https://news.uchicago.edu/story/balloon-filled-ground-coffee...

Too many edge-case failure modes in an uncontrolled setting. Building platforms that could seriously harm people by just falling over is an inherent design risk. =3

[theteapot]

> But the claim made by this author is far removed from the actual demo he describes. I've seen same demo for years

The article describes multiple demos. Are you referring to the chicken nuggets one? That sounded pretty impressive to me. Is there publicly available videos of this?

[NalNezumi]

I'm describing all of them.

As for chicken nugget here's for example one (company) showing same capacity 4 years ago

https://youtu.be/6SbpfN5ed38?si=srtdZCdKOdPZ_wRn

They today have similar system that can quickly sort dumplings (more sensitive than chicken nuggets) ob conveyor belt.

No sim2real even needed. That haptics sensor is dirtcheap; camera based haptics sensor are today even available as open-source hardware that you can assemble for cheap.

If we don't limit to company demos we can dig up demos from I think almost a decade ago, and at least ~5 years ago for company demos.

[ghostly_s]

nothing remotely like the described demo in the video you linked.

[pixl97]

One thing the LLM market has done is any AI work that can run on a GPU has a vastly larger amount of computing power.