NFT artists kind of hide behind the oldschool art world to justify it…

I think it shows how stupid that world is.

The point of art is to move people, to make them think and feel something by conveying the artist’s thoughts, not be a store of value. And while I’ve been to museums with gorgeous pieces, if you’re paying millions for a painting, at some point most of that value is the gratification of hoarding it. That same money could buy you an incredible experience in today’s art landscape, but it’s not about the experience, is it? And NFTs are like the perfect deconstruction of that.

That being said, OP I am downvoting your post because eyeballs are exactly what crypto bros want, no offense :P

Stadia.

…Is he gonna hire Phil Spencer?

Will it go as well as X AI?

Grok is the laughing stock of the LLM world, and he also completely lied about his promise to make them open source. The latest Qwen release from Alibaba (open, locally runnable weights, tiny, Apache, freaking smart) has rendered his obscene hoard of H100s basically obsolete, with a miniscule fraction of the money.

That’s the strange thing, it’s completely open with tons of news outlets constantly calling it out, not an “taboo secret” like of like old facist governments. And in all liklihood, it will stay that way.

The filter bubble and American apathy is just that powerful, I guess?

The plan is to have a disturbing deficit, and interest payments are getting really big now.

This is like climate change, it’s not apparently hurting anyone tomorrow so… no one cares anymore.

Llama 3.1 is not even a “true” distillation either, but its kinda complicated, like you said.

Yeah Qwen undoubtedly has synthetic data lol. It’s even in the base model, which isn’t really their “fault” as its presumably part of the web scrape.

And… this curbs immigration? By making the country less desirable to live in or something?

How useful would the training data be

Open datasets are getting much better (Tulu for an instruct database/recipe is a great example), but its clear the giants still have “secret sauce” that gives them at least a small edge over open datasets.

There actually seems to be some vindication of using massively multilingual datasets as well, as the hybrid chinese/english models are turning out very good.

You can use larger “open” models through free or dirt-cheap APIs though.

TBH local LLMs are still kinda “meh” unless you have a high vram GPU. I agree that 8b is kinda underwhelming, but the step up to like Qwen 14B is enormous.

It turns out these clusters are being used very inefficiently, seeing how Qwen 2.5 was trained with a fraction of the GPUs and is clobbering models from much larger clusters.

One could say Facebook, OpenAI, X and such are “hoarding” H100s but are not pressured to utilize them efficiently since they are so GPU unconstrained.

Google is an interesting case, as Gemini is getting better quickly, but they presumably use much more efficient/cheap TPUs to train.

I don’t think Qwen was trained with distillation, was it?

It would be awesome if it was.

Also you should try Supernova Medius, which is Qwen 14B with some “distillation” from some other models.

All the other brands went along

(My 2020 G14 has 3 A ports and ethernet, but still…)

Yes blocking major backers is bad, I agree with that. The mod behind this kind of sounds unpleasant too.

What does being on Lemmy matter?

Discord is like the antithesis of Lemmy, a siloed off, inefficient, unscrapable, private, proprietary and dangerously monopolistic echo chamber. I’ve seen it swallow too many of my niches, and from my experience, it turns people into jerks.

Hence what I’m getting at is that this may not have happened without all that nonsense in the unoffiical discord.

You mean a few github accounts and a bunch of jerks on Discord, which you are bringing up on Lemmy?

If anything this is just another repudiation of Discord as a whole. I hate how its eating the internet like mad cow disease.

Yeah, it’s great for instant feedback. You can ask it “does this dialogue above feel out of character or stilted?” or ask it about character growth, and it will generally give an interesting critique in-line with your writing in the notepad.

You have to keep in mind that it can go off the rails (repeatedly querying it is good to check for this), and has a sycophancy bias, and is kinda dumb, but its still quite useful.

For me it’s not efficiency, if anything it takes much longer running the LLM notepad because I will randomly ask it about passages/chapters and revise stuff.

It’s kinda fun having an assistant to just bounce ideas off of on a whim. You can’t get that with beta readers, as they don’t just sit there while you write (and the LLM is much faster), and I don’t feel like I’m being charged for every response with an API model, especially if it ingests the entire story every time.

It’s also “smart” beyond me. For instance, sometimes I wanna come up with a name for a character, city or whatever, and I can ask it “what’s an interesting name for this engineer character’s brother, from X city in the story, maybe something mythological and fire themed,” and it will crank out tons of examples and start a little conversation about it. It takes me places I never would have googled, much less known off the top of my head, because everything is on the top of an LLM’s head.

Those diabolical Chile preservationists…

I suspect Lemmy is going to dislike this, but local LLMs are great writing helpers.

When you’re stuck on a sentence or a blank paragraph, get them to continue it, and rewrite it once it jogs your mind. If you’re drafting ideas for characters or chapters, you can sanity check them or sometimes get ideas. They can reword and clean up your writing and improve it beyond what self experimentation can do… just keeping in mind that its like an idiot intern that tends to repeat itself and hallucinate.

And this is very different from an API model like ChatGPT because:

• It won’t refuse you.

• It’s formatted as a notebook you can continue at any arbitrary point, rather than a user/chatbot type format.

• The writing isn’t so dry, and it isn’t filled with AI slop, especially with cutting edge sampling

• All its knowledge is “internal,” with no reaching out to the web or hidden prompting under your nose

Along with all the usual reasons, namely being free, self-hosted, more ethically trained, fast and efficient with long context thanks to caching, and has nothing to do with Sam Altman’s nightmarish visions.

I’d recommend: https://huggingface.co/nbeerbower/Qwen2.5-Gutenberg-Doppel-32B

And once the story gets long: https://huggingface.co/EVA-UNIT-01/EVA-Qwen2.5-32B-v0.2

I’d recommend LanguageTool (with a local server and the browser extension) for locally hosted spelling/grammar/style as well.

I have ADD, which may be why I find this setup to be so therapeutic.

Everyone seems to be wringing hands about policy, but this is just another datapoint that propaganda won this election.

Depends how much Musks’s companies try to unionize.