I don’t mind

We have released the training corpus, models and a lot of multilingual evaluation benchmarks in Hugging Face: huggingface.co/collections/Hi…

We hope this project begins a wave of multilingual models for the medical domain!

Been preaching (nor writing) this for a while: ANY data filter is a decision to not learn about sth. Think carefully: really want that? For small runs? For large runs? What else could you be inadvertently removing? Do you really know your filter? Looked at examples it removes?

Pretty much.

Be clear about what you want up front, have a list of common questions or issues so they can’t use that as an excuse.

Can even send prerecorded loom.

Have defined terms if you can and then take meetings with folk that show excitement and can say why.

Invisible to you all but an interesting change to X is that we rebuilt the entire trends system from scratch. All of the new Grok trends are generated on just two 32 core cpu machines. It's incredibly simple and efficient.

While we are bringing you tonnes of new features, a lot…

Outpainted from the previous one

#SD3

Is your mind 🧠:

Interdimensional cable 🥒

“Star Trek” and the Strange New Worlds of Spatial Computing
#NABShow2024 🖖 #holodeck 𝚃𝚑𝚎 ❤ 𝚘𝚏 𝚂𝚝𝚊𝚛 𝚃𝚛𝚎𝚔 NVIDIA Apple Light Field Lab @otoy #RENDER #StarTrek

amplify.nabshow.com/articles/nabsh…

CosXL is an SDXL model tuned to produce full color range images. As a side effect of that, it's really fun to have it generated images from prompts starting with 'A psychedelic painting of' 🫠 I used the unofficial demo which is available at this space: huggingface.co/spaces/multimo…

Our new work in Nature Medicine

Using generative models in path, radiology and derm - we can create synthetic training data to ⬆️ AI fairness & robustness, including underrepresented groups. Work w/ a great team Google DeepMind Google AI Google Health

nature.com/articles/s4159…

Excellent work by Sanchit Gandhi and Yoach reproducing the text-description-to-speech model I developed while at Stability AI 👏❤️

Been out with a flu last few days (sorry everyone I’ll message back when my brain reboots after gunk cleared out 🤧) but at least it wasn’t Covid like a few years ago 😵

Man I was knocked out for like three months and still have PTSD over that every time I get a cough 😷

🤷🏾

Great job by cohere also super strong multilingually due to work by Cohere For AI on open language stuff.

Starling 7b remains crazy strong on small side, has anyone MoE'd it yet 👀

A Qwen 72b level model will take about 2-3 days to train on a 20k H100 cluster...

Hoy anunciamos StableLM2 12B, un modelo de lenguaje open-source en español muy potente y de coste asequible. Entrenado desde cero con datos públicos, permite el uso de RAG, funciones y tools externas. Excelente para resumir y extraer datos de cualquier texto. DM si te interesa!

COSXL looking good!

Announcing MindEye2: Shared-Subject Models Enable fMRI-To-Image With 1 Hour of Data! 🧠👁️

High-quality reconstructions of images from fMRI using 1 hour of training data vs. past work using 30-40 hrs/person!

arXiv: arxiv.org/abs/2403.11207
project page: medarc.ai/mindeye2