naillo

Interesting to read this and think about how the future of AI might turn out if the google microsoft cartel comes into being like they're trying to.

"What about a bigger model" is basically feature creep for ML.

Life sucks and then you die, I don't need constant reminders of it from the internet though.

> Systems that don't allow people to extract value from the hard work they put into collective advancement [...] one's going to spend all day making candy and put it in the "free candy" bowl

On the other hand this is what researchers do all day every day. PhDs and professors work for the common good and get barely any pay in return. Maybe the future model in art and music is more like the academic researcher.

I wish people made unconditional predictive models for music instead of text-to-music ones. Would be so cool to give an input 'inspiration' track that it 'riffs' a continuation to. That's usually what I want, just continue this track it's too short that's what I want to hear more of. (That said this is super cool though.)

First time I've heard of `ollama` but having tried it now for a bit I'm super impressed! This is what I've always wanted playing with LLMs locally to be, just pull weights like you would packages. It all just works. Really nice work :)

Sidenote I can't imagine openai would ever be able to actually patent gpt-x given that they neither invented the model (google did) nor created the training data.

Always a bit funny when openai makes use of trademark and copyright law that they so frivolously ignore when actually training the models.

Michael Nielsen is a treasure

If I wasn't dependent on using the internet for income I would just try to find other ways of using my free time than the internet at this point if the worst case outcome in the article came to pass.

Sidenote, I love how clean and readible this site is. No popups, just clean and well written text in the middle.

Good motivation to build or support platforms where this can't even be a possibility, i.e. without phone number authentication or other identity revealing steps as part of authentication.

I like the name

> For those who don’t know, Hugging Face is an influential startup in the machine learning space that has played a significant role in popularizing the transformer architecture for machine learning

This is a crazy point of view

Adobe seems exeptionally unproductive with incredible product stagnation so this doesn't seem that infeasible to me.

You're right dang

You can tell when someone shouldn't be listened to online if they feel a need to give a "take" to this piece of news. Just wait until we get actual results if you don't have a physics phd at minimum no need for any rushed take otherwise.

They often reference this paper as the motivation for that https://arxiv.org/pdf/2203.15556.pdf I.e. training with 10x data and 10x longer can yield as good models as a gpt-3 model but with fewer weights (according to the paper) and the same principle applies in vision.

The same could have been said when photoshop or CGI tools like blender replaced hand sculpting and hand painting but I think it hasn't been a net negative across the board (I think rather the opposite).

I can't say for 1.0 but in 0.9 hands get fairly often rendered perfectly. It's not always right but it's way better than any other earlier release (where it's usually consistently wrong).

I've been playing with 0.9 and it can generate nude people so it seems not.

You can do this if you select the `pipe.enable_model_cpu_offload()` option. See this https://huggingface.co/stabilityai/stable-diffusion-xl-base-...

You get access to the weights instantly if you apply for them. It's basically not a hurdle.

(I've been having fun with this for a few days. https://huggingface.co/stabilityai/stable-diffusion-xl-base-... Not sure there's much of a difference with the 1.0 version.)

Stability AI is awesome I love them

This kind of behaviour makes me not trust security researchers (something they should desperately try to protect given that trust in the field is essential).

Nothing more like a family than the conditional love of a company.

Reading this I'm mostly thankful real brain power and the general smart programming community is seriously taking a close look at all these things. I barely feel the need to try to compete for insight gathering it feels very healtily analyzed from every perspective finally.

How is program synthesis the same as regular expression search? Honest question

It's just a random shitpost reply to a random tweet. Not sure why this is on the front page

To be fair to sama he did say that "it still seems more impressive on first use than it does after you spend more time with it"[1] but I suspect that was a kind of 4d chess move to hope to be quoted in situations like this when he knew the hype died down a few months later to cushion the blow after the hype cooled off.

[1] https://twitter.com/sama/status/1635687853324902401