'They might meet for dinner if they are in the same city.' Or not.
nytimes.com/2024/04/29/tec…

Struggling to find the best meme? I've set up a Hugging Chat Assistant for that (and it's not bad, IMO!) hf.co/chat/assistant…

Already 100 people have joined our Journalists on 🤗 community! And it's just a start!

Join us here: huggingface.co/JournalistsonHF

The Latest Online Culture War Is Humans vs. Algorithms wired.com/story/latest-o…

A new model named GPT-2 Chatbot has mysteriously appeared on the LMSYS benchmark tool, and it's generating a lot of speculation due to its performance.

I've tested it using the example of Alexis, who needed a new set of business clothes.

Check it out for yourself:…

Noyb claims that ChatGPT's behavior violates the GDPR on privacy, the accuracy of information as well as the right to correct inaccurate information.

politico.eu/article/chatgp…

The Financial Times and OpenAI strike content licensing deal ft.com/content/333287…

Admittedly the CPU here is a ~$10,000 96 core AMD Ryzen Threadripper PRO 7995WX, but still, it looks like GPUs don't necessarily have the total monopoly on LLM inference performance that I thought they did

This demo by ByteDance is as strange as it is fun huggingface.co/spaces/ByteDan…

Tuesday's Practical AI for Journalism covers local models and fine-tuning!

Here's Llama 3 fine-tuned on a 1846 housekeeping book discussing whether you should pet lazy cats, 19th century prompt engineering and eating buffalo wings for breakfast.

eventbrite.com/e/practical-ai…

LLMs evals + #dataviz

« D’un côté, les internautes veulent créer leurs propres versions de leurs chansons préférées. De l’autre, des outils d’IA permettent de créer des tonnes de chansons à partir d’un bout de texte. Le résultat : des listes de lecture infinies, générées par ordinateur.
 
C’est…

When your cheating attempt ends up in a dataset:

This week-end, train your own music generation model in 10min 🔥

Dreambooth MusicGen on your favorite artist or genre, using 10GB of GPU RAM in a matter of minutes 🎸

Colab ⬇️

If you want to catchup this weekend, 4 models and 2 datasets

** Models **

Apple OpenELM huggingface.co/collections/ap… + huggingface.co/collections/ap…

Microsoft Phi-3 huggingface.co/collections/mi…

Qwen1.5-110B and 110B-Chat huggingface.co/Qwen/Qwen1.5-1…
huggingface.co/Qwen/Qwen1.5-1…

Snowflake Arctic…

Alibaba just released Qwen1.5 - 110b on Hugging Face hub🎉
Model: huggingface.co/Qwen/Qwen1.5-1…
Demo: huggingface.co/spaces/Qwen/Qw…
✨ The largest one in the Qwen1.5 series
✨ Context length 32K tokens
✨ Multilingual: Chinese, English, French, Korean, Japanese, Vietnamese, Arabic etc.