glitch83

joined 1 year ago
[–] glitch83@alien.top 1 points 11 months ago

Take a look at flux.jl. Last time I checked, they were building in support. You may have some success if you join that community over pytorch or tensorflow.

[–] glitch83@alien.top 1 points 11 months ago

I’m guessing inputs are the pose (x y z, rx ry rz) and the outputs are your joints.

Two things: unless you are just trying to solve for the final position, it may be smart to add the initial pose. So if the trajectory matters for purposes of avoiding obstacles then knowing the start position and maybe the obstacles in configuration space could be helpful input to the network. That being said, some sort of sequence model where you encode the goal with an mlp and then a decoder for the trajectory would be cool. Id you just want the final joints then an mlp is fine.

Some people use reinforcement learning in this setting too but I tend to think that’s overkill.

As others have said too: you could always use a traditional algorithm like some variant of RRTs to solve this too as it’s mostly a “done” problem for traditional robotics.

[–] glitch83@alien.top 1 points 11 months ago

Controversial opinion: OpenAI never was a leader. Sure it did some cool things but it neither reached AGI nor became profitable. It was doomed to failure from the beginning based on the non-profit's mission.

That being said, I'm still very bearish on AGI in general. I don't think we're as close as we think we are and the chaos is natural since we don't actually know how to get there. Success in AI is an illusion.

[–] glitch83@alien.top 1 points 11 months ago (1 children)

Honestly I'm not sure what you are really asking. Could you define Machine Learning in the way you see it? I feel like the answer is too obvious to be the answer you're looking for.

[–] glitch83@alien.top 1 points 1 year ago (4 children)

Sounds a little like you're overqualified. Fine tuning is something you could probably pick up from a udacity course if you'd like.

[–] glitch83@alien.top 1 points 1 year ago

Maybe? Vision has been around a lot longer than NLP in industry. It’s permeated into some challenging areas like embedded and edge spaces due to privacy and requirements. If the foundation models can’t run on the edge then I can imagine foundation models only affecting a small portion of vision applications.

[–] glitch83@alien.top 1 points 1 year ago (1 children)

Probably to support models with options?

[–] glitch83@alien.top 1 points 1 year ago

Generally pretty hard.

[–] glitch83@alien.top 1 points 1 year ago

The responses are insane. LLMs are out of control…