robitsT6

Right, if people want to talk about how they are worried about a future with super intelligence AI, that's I think something almost everyone can agree on is a worthy topic, maybe to different degrees but not the issue in my mind.

I think what it feels like I see a lot, are people who - because of their fear of a future with super intelligent AI - try to like... Deny the significance of the event, if only because they don't _want_ to wrestle with the implications.

I think it's very important we don't do that. Let's take this future seriously, so we can align ourselves on a better path forward... I fear a future where we have years of bickering in the public forums on the veracity or significance of claims, if only because this subset of the public who are incapable of mentally wrestling with the wild fucking shit we are walking into.

If not this, what is your personal line in the sand? I'm not specifically talking to any person when I say this. I just can't help but to feel like I'm going crazy, seeing people deny what is right in front of their eyes.

I can't really comment on the expenditure or the general strategy of Google's robotics teams, I know they closed down every day robotics, but that seemed to be specifically about the hardware.

Their software efforts still seem to be going strong, and I don't really think it's fair to say that demonstrating transfer learning in an embodied model none the less, is unambitious - nor does that seem like it's reflected in the results - but to be honest, I'm more or less a layman when it comes to this so I'll defer to you and keep what you're saying in mind.

To that end, if you're still feeling up to it, maybe you could tell me what your thoughts are with efforts like this from Google:

https://diffusion-rosie.github.io/

Seems to compare (somewhat) to Mimicplay, in that it is attempting to create more data for "cheap", even if it's not "real" control data.

Right but is there any reason that this architecture won't work with more diverse data? Fundamentally it seems like their research is benchmarked around pick and place, so it makes sense to me that they would want to prove out transformer models could work in this space. Knowing transformers, it's probably safe to say that with more diverse training data, it will be able to scale to more complex controls in more complex embodied robots.

While I would love to see them working on robot chefs, I can appreciate that they want to start small. And regardless, it doesn't seem like there are any constraints outside of data for this architecture to work in more and more domains.

Reading the paper, they mention the robotic control is handled by RT-1:

The low-level policies are from

RT-1 (Brohan et al., 2022), a transformer model that takes

RGB image and natural language instruction, and outputs end-effector control commands.

For those that don't know, RT-1 (Robotic transformer) is previous work from the team that converts natural language to custom control code.

You can read more about Rt-1 here:

https://ai.googleblog.com/2022/12/rt-1-robotics-transformer-...

Maybe I'm missing something, but this sounds quite generalizable.

I'm sorry, I like riffusion as much as the next guy, but in what world is any riffusion example better than the literal first example on this demo page?

This isn't a very compelling argument. First of all, they aren't a "mish mash" in any real way, it's not like snippets of images exist inside of the model. Second of all, this is entirely subjective. Third of all, entirely inconsequential - if these models create 80% of the video we end up seeing, is it going to matter if you don't think it's a tasteful endeavour?

But there have been quite a few scientific papers that have used discoveries from AlphaFold already. There have been many scientists who have been stuck for years, who are suddenly past their previous bottlenecks. What gives you the impression that it hasn't helped us?