In 2016-2018, I would read the occasional article on decentralized computing and edge computing for ML. In a conversation today we realized that that these ideas are barely discussed now. Is it that in the age of LLMs, decentralized computing is a lost cause? Are there any experiments on open source GPUs that I can read about?
Ni! Hi tarunima… This doesn’t directly concern open hardware, but I’ve been watching closely a project called Bacalhau (website, forge), which kind of fits your description. Cheers
This is not my area of expertise, but I have noticed some things.
Companies like SeeedStudio still seem to be investing in edge computing hardware from what I see in their emails, and projects like TinyML are still mentioned quite a bit, but it does feel like LLMs have somewhat crowded out other aspects of ML.
I think there are (or have been) efforts to create open source GPUs, mainly using FPGAs.
It seems like an FPGA would be the most cost-efficient way to do this type of development unless a project has major financial backing.