bitfucker

joined 8 months ago
[–] bitfucker 5 points 3 months ago

Research and development is tricky because you will never know how much more progress you will need before reaching a satisfying result.

[–] bitfucker 2 points 3 months ago (2 children)

*Rant for the beginning of the article ahead

Why in the name of god did they try to bring LLM to the pictures. Saying AI/ML is good enough for predictive maintenance tasks, but noooo, it has to be LLM. If they want to be specific then don't be misleading, I think what they mean is the attention layer/operation commonly used in LLM to capture time series data. I understand that the Recurrent style neural network and LSTM has its limitations. And I agree that exploring attention to be used in time series data is an interesting research but LLM? Just no.

[–] bitfucker 3 points 3 months ago

I really like 8 but also understand that not all unfinished projects are caused by ADHD. So 2 it is.

[–] bitfucker 4 points 3 months ago

Yeah, that is the only weird part about it. I think AMD is cooking up something right now behind the curtain. Whether that something is good or bad we don't know yet but so far it does not look good.

[–] bitfucker 55 points 3 months ago (3 children)

Did he taste good?

[–] bitfucker 16 points 3 months ago

Yep, I really hope that AMD is cooking up something. A shame that it's not open sourced but if it means going up against nvidia dominance now, I'll take their side just for that cause only.

[–] bitfucker 2 points 3 months ago (1 children)

You're not wrong. IIRC they already planned to stop or layoff their Dart developer in favour of Golang (also a language by Google)

[–] bitfucker 20 points 3 months ago

Man, I never thought of it as a physical shitpost. Now it all makes sense.

[–] bitfucker 1 points 3 months ago

I don't know what a co-op is, but I see the moderation system at BlueSky and think if that can be generalized then maybe that is the answer to a lot of centralization problems. So a node representing a peer can do basically anything that node operator/peer wishes including delegating his node operation to another node or a quorum of node. And if the peer wishes so, they could also take their node back anytime. For example in communication, we already have decentralized web of trust from PGP or other Public Key Infrastructure.

[–] bitfucker 3 points 3 months ago (5 children)

I think a better solution is to truly emulate what really happens in the real world via peer to peer networking, and appointing / trusting certain individual as an admin/moderator for the node. That way a node can choose to become independent or have a quorum system or fully trust a single other node. That is my idea anyway, I haven't dug much deeper into this idea.

[–] bitfucker 1 points 3 months ago
[–] bitfucker 10 points 3 months ago

You deserve a dedicated wikipedia page ma dude

view more: ‹ prev next ›