Nicola Masella
banner
nmasella.bsky.social
Nicola Masella
@nmasella.bsky.social
Engineer
Linkedin profile: http://bit.ly/3hNRDn8
Sign the Petition
United States of Europe - Now!
chng.it
March 13, 2025 at 12:45 PM
"We shift a substantial portion of LLM workloads to consumer devices by having small on-device models collaborate with frontier models in the cloud."

hazyresearch.stanford.edu/blog/2025-02...
Minions: the rise of small, on-device LMs
hazyresearch.stanford.edu
March 7, 2025 at 7:33 AM
Google Co-Scientist: extensive use of agents to enhance research.

research.google/blog/acceler...
Accelerating scientific breakthroughs with an AI co-scientist
research.google
February 21, 2025 at 9:13 AM
An interesting LLM with a fine-tuned version of Deepseek-R1-Distilled-Qwen-1.5B surpasses the performance of OpenAI’s o1-preview with just 1.5B parameters on a math evaluations

pretty-radio-b75.notion.site/DeepScaleR-S...
DeepScaleR: Surpassing O1-Preview with a 1.5B Model by Scaling RL | Notion
Michael Luo*, Sijun Tan*, Justin Wong†, Xiaoxiang Shi, William Tang, Manan Roongta, Colin Cai, Jeffrey Luo
pretty-radio-b75.notion.site
February 12, 2025 at 11:11 AM
someone said that Europe can't compete with the US and China in the AI race, fortunately, some people don't think the same way and work hard to do their part. Thank you Mistral for your contribute

mistral.ai/en/news/all-...
The all new le Chat: Your AI assistant for life and work | Mistral AI
Brand new features, iOS and Android apps, Pro, Team, and Enterprise tiers.
mistral.ai
February 9, 2025 at 6:17 PM
Understanding Model Distillation ai.gopubby.com/understandin...
Understanding Model Distillation
Learn what model distillation is and how it works by building one yourself
ai.gopubby.com
February 7, 2025 at 4:51 PM
Welcome to the era where everything can be a fake

omnihuman-lab.github.io
OmniHuman-1 Project
OmniHuman-1: Rethinking the Scaling-Up of One-Stage Conditioned Human Animation Models
omnihuman-lab.github.io
February 4, 2025 at 10:26 AM
US inaugurates a new era, the Middle Ages

darioamodei.com/on-deepseek-...
Dario Amodei — On DeepSeek and Export Controls
On DeepSeek and Export Controls
darioamodei.com
January 30, 2025 at 8:25 PM
the most useful thing i read about #deepseek topic

x.com/yishan/statu...
x.com
x.com
January 28, 2025 at 9:49 AM
in the top 3 of the modern AI papers

arxiv.org/pdf/2501.12948
arxiv.org
January 25, 2025 at 3:40 PM
DeepSeek R1 is redefining efficiency in AI. With its cutting-edge architecture, it's delivering faster insights, lower latency, and reduced energy consumption.

The future of AI is not just smarter, but greener too.

#AI #DeepSeekR1 #Efficiency #Sustainability #Tech
January 25, 2025 at 12:09 PM
SmolVLM, an interesting Vision Language Model (very small)

huggingface.co/blog/smolervlm
SmolVLM Grows Smaller – Introducing the 256M & 500M Models!
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
January 24, 2025 at 11:15 AM
Ollama on Google Vertex, a serverless LLM solution

medium.com/google-cloud...
Hey Ollama, how about running on Vertex AI?
This post walks you through deploying a Gemma 2 SQL adapter using Ollama on Vertex AI.
medium.com
January 24, 2025 at 6:59 AM
github.com
January 23, 2025 at 6:54 PM
NVIDIA released Cosmos, an open-source, open-weight Video World Model. It's trained on 20M hours of videos and weighs from 4B to 14B. NVIDIA applies Cosmos to large-scale synthetic data generation for robotics and autonomous driving, you can go deeper here:

Check it out: github.com/NVIDIA/Cosmos
GitHub - NVIDIA/Cosmos: Cosmos is a world model development platform that consists of world foundation models, tokenizers and video processing pipeline to accelerate the development of Physical AI at ...
Cosmos is a world model development platform that consists of world foundation models, tokenizers and video processing pipeline to accelerate the development of Physical AI at Robotics & AV lab...
github.com
January 7, 2025 at 9:19 AM