@RippaSats What I’m saying is that companies can’t simply pick a LLM, deploy it and be done with it. By the time you are done (and it takes a long time, trust me), there are already a few newer models that significantly outperform yours. You can innovate, but you cannot statically deploy AI

“the evidence to date suggests that noncompetes suppress wages, reduce competition and keep innovative ideas from breaking into the market.”

https://t.co/lslMyTfWEI

Unless you live and breathe artificial intelligence, you will not be able to catch up with the evolution of large language models. Whatever you are building now is already DOA.

Especially if you have no specific competence in #AI, and you think this is just another technology. https://t.co/4CW6HPXoGE

Oh no! What completely surprising bad news! >

“The number of active users on the Mastodon social network has dropped more than 30% since the peak and is continuing a slow decline”

https://t.co/uNly7fHMt1 https://t.co/3OWKgeaNyV

@camenduru If you find a way to speed up Tortoise TTS (maybe by updating it to Torch 2), please let me know. It’s a shame it has been abandoned.

@camenduru Do you think this could work with a non-human subject? For example, to make talk an object?

I assume not, but worth checking.

Amazing new release from @liuliu, enabling checkpoint merger on iOS platforms!

It’s mind-blowing to think that we can do this on the Tube on our iPhones with intermittent or no Internet connection instead of playing Candy Crush.

(at least, I hope so) https://t.co/BKIsDyDvhl

BTW, if you are working with #stablediffusion or other #generativeAI technologies, you should really follow @camenduru. His/her Stable Diffusion Colabs and Hugging Face Spaces are the absolute best around.

#VSCode inside Hugging Face. And it’s very fast, even without an assigned GPU. https://t.co/552GNXpgA7

@CarolineSalen It’s called Historical Figures App.