many futures are possible.
Joined November 2006
- Tweets2.3K
- Following690
- Followers1.2K
- Likes624
Alek Andreev retweeted
Gemma 4 outperforms models over 10x their size! (note the x-axis is log scale!)
Alek Andreev retweeted
An exciting milestone for AI in science: Our C2S-Scale 27B foundation model, built with @Yale and based on Gemma, generated a novel hypothesis about cancer cellular behavior, which scientists experimentally validated in living cells.
With more preclinical and clinical tests, this discovery may reveal a promising new pathway for developing therapies to fight cancer.
Alek Andreev retweeted
Thanks Jeff Lawson for buying the Onion so we could get this great zoom in on electoral coverage. No other news source has gone to this level of detail.
Alek Andreev retweeted
A great set of models have entered the arena!
Sadly Gemma2 9B dropped from one of their tables so I had to add it. Would be even better if all models were evaluated with same codebase as I see fluctuations in numbers.
We just released two small models, with 3B and 8B parameters. Ministral 3B is exceptionally strong, outperforming Llama 3 8B and our previous Mistral 7B on instruction following benchmarks.
mistral.ai/news/ministraux/
Alek Andreev retweeted
Gemma 2 just got even better! 🚀 New Japanese-tuned 2B model AND a $150K Kaggle competition to build Gemma models for every language. Great to have @sundarpichai here to share the excitement!
Read more: goo.gle/Gemma4Japan
#GemmaDeveloperDay
Gemma 2 is out and it’s crazy to think we can fit the best open LLM in the world on an old 32GB USB stick. 🧵
Gemma 2B (shipping very soon), will be the most capable on-device model. This is a new space and it's very exciting what app developers can do with local, private LLMs.
The whole Gemma team is irreplaceable, but I want to particularly thank the pretraining team - @pathak__shreya, @moruganribieru , Cassidy Hardin, @Mesnard_Thomas. 🙏 to @armandjoulin, who always knows when to say no.
Announcing RecurrentGemma, which removes the quadratic compute complexity of transformers!
Releasing RecurrentGemma - one of the strongest 2B-param open models designed for fast inference on long sequences and massive throughput!
Both pre-trained and IT checkpoints available + code - try them out here!
Code: github.com/google-deepmind/r…
Weights: kaggle.com/models/google/rec…
Alek Andreev retweeted
I am very happy to announce that Gemma 1.1 Instruct 2B and “7B” are out! Here are a few details about the new models:
1/11
Alek Andreev retweeted
UPDATE: Gemma 1.1 7B & 2B - Instruction Tuned! ✨
> substantial gains in quality, coding capabilities, factuality, and instruction following.
> better multi-turn conversation quality.
huggingface.co/google/gemma-…
It's a joyful journey working with the brilliant, tight-knit Gemma team! There is a bright future for open models!
Introducing Gemma - a family of lightweight, state-of-the-art open models for their class built from the same research & tech used to create the Gemini models.
Demonstrating strong performance across benchmarks for language understanding and reasoning, Gemma is available worldwide starting today in two sizes (2B and 7B), supports a wide range of tools and systems, and runs on a developer laptop, workstation or @GoogleCloud.
Excited to see what you’ll create → ai.google.dev/gemma blog.google/technology/devel…
Alek Andreev retweeted
The few people who didn't leave Irpin were mostly those too old and to frail to walk.
For them, the last month has been hell. Death, illness and frayed nerves, a few miles from Kyiv.
I found it difficult to listen to these stories & keep it together.
theguardian.com/world/2022/m…
Alek Andreev retweeted
Russia's brutal siege of Ukraine's port city of Mariupol has left residents without electricity, water, or heat for nearly 3 weeks.
“We were so happy when it snowed. We gathered the snow, cleaned the cigarette butts from it, and drank that."
foreignpolicy.com/2022/03/18…
