xiyangluo @xylmath

we launched some muse spark updates yesterday, including muse spark voice and live AI w your camera in Meta AI app + muse spark rolling out to glasses 😎 check them out!

Meta Newsroom @MetaNewsroom

4 weeks ago

Today we’re introducing Meta AI Voice Conversations powered by Muse Spark that let you talk naturally to Meta AI (interrupt, switch topics, or swap languages), and as you talk, Meta AI can generate images and pull up recommendations from Reels, maps, and more. We’re also bringing

99 158 1K 260K 305

Today we’re introducing Meta AI Voice Conversations powered by Muse Spark that let you talk naturally to Meta AI (interrupt, switch topics, or swap languages), and as you talk, Meta AI can generate images and pull up recommendations from Reels, maps, and more. We’re also bringing live AI to the app, so you can point your camera at the world and ask about what you’re seeing in real time. about.fb.com/news/2026/04/i…

After an incredible chapter at Google DeepMind working on Gemini, I’m starting something new with some exceptional people. I’ve increasingly felt that the current research paradigm — scaling models, data, and static deployment won’t get us all the way. We believe the next phase comes from something different: new learning algorithms, architectures beyond today’s stack, and systems that automate the process of building itself. If we’re right, the way work is organized changes — small teams with highly capable agents will change the unit of building. That’s Core Automation — a factory that builds factories. More soon!

Jerry Tworek @MillionInt

2 months ago

Small step for a finger on "post tweet" button. Big step for millions of future AI agents doing useful work for us.

58 37 669 166K 97

MSL *really does* run like a startup :) join us if that sounds exciting to you!

Vu Tran @vu0tran

2 months ago

Crazy to see the positive response to Muse Spark. I joined Meta in Dec and was surprised how startup-y MSL feels. Dec through now, through the holidays was literally nonstop building for me. Everyone on the team cares. People want to do great work

4 4 167 116K 13

Proud to be part of the journey. Onwards!!!

Alexandr Wang @alexandr_wang

2 months ago

1/ today we're releasing muse spark, the first model from MSL. nine months ago we rebuilt our ai stack from scratch. new infrastructure, new architecture, new data pipelines. muse spark is the result of that work, and now it powers meta ai. 🧵

742 1K 10K 4.5M 3K

Meet Muse Spark. This is one step towards personal superintelligence and a great team effort involving numerous nights of training, debugging, improving instability, and fighting reward hacking. We have just begun, and feedback is extremely welcome.

Alexandr Wang @alexandr_wang

2 months ago

1/ today we're releasing muse spark, the first model from MSL. nine months ago we rebuilt our ai stack from scratch. new infrastructure, new architecture, new data pipelines. muse spark is the result of that work, and now it powers meta ai. 🧵

742 1K 10K 4.5M 3K

Fun nine months! My first week i remember we had a long dinner in the cafeteria daydreaming about the cool research directions to pursue, then going to back to our desks to write a basic script to inference llama. Now we have a pretty complete stack and our first model is out 🥑

Alexandr Wang @alexandr_wang

2 months ago

1/ today we're releasing muse spark, the first model from MSL. nine months ago we rebuilt our ai stack from scratch. new infrastructure, new architecture, new data pipelines. muse spark is the result of that work, and now it powers meta ai. 🧵

742 1K 10K 4.5M 3K

Our first model from MSL, Muse Spark, is now available on meta.ai! This is an efficient all-rounder model. It supports fast responses, deeper thinking, visual chain of thought, a higher inference “Contemplating” mode. Plus, it’s natively multimodal. 1/

Triple Crowned! 👑👑👑🫡 "This is the first-ever sweep across text, vision, and WebDev by any model!" x.com/lmarena_ai/sta…

Arena.ai @arena

a year ago

🚨Breaking: @GoogleDeepMind’s latest Gemini-2.5-Pro is now ranked #1 across all LMArena leaderboards 🏆 Highlights: - #1 in all text arenas (Coding, Style Control, Creative Writing, etc) - #1 on the Vision leaderboard with a ~70 pts lead! - #1 on WebDev Arena, surpassing Claude

37 214 1K 531K 248

Great set of slides by my Gemini colleague @FeinbergVlad on scaling considerations in large language models, addressing the fact that the classic "scaling law" work does not take into account inference cost (!), distillation, learning rate schedules, etc. vladfeinberg.com/2025/04/24/gem…

Vlad Feinberg @FeinbergVlad

a year ago

Recently had the pleasure of lecturing back at Princeton in a grad seminar. I took the opportunity to cover how scaling laws have evolved since their inception, leaning heavily on great external content from my colleagues @borgeaud_s @jalayrac @jacobaustin132 . Content in thread

5 106 812 138K 772

Scaling laws in deep RL? Turns out that batch size, learning rate, and UTD (update-to-data) for getting the most efficient and scalable deep RL has predictable relationships. Checkout the analysis in new work by @_oleh & collaborators: arxiv.org/abs/2502.04327

.@GoogleDeepMind proposed a method that enhances LLMs with an offline coprocessor that works with the models' internal memory (kv-cache). • What's the coprocessor's role? It enhances the model's KV-cache by adding extra "latent embeddings" (compressed representations) for more accurate outputs. • What is good about it? - The coprocessor operates independently, and the base LLM remains frozen. - It operates offline and asynchronously, meaning it can improve the model’s memory in the background. - If the coprocessor isn’t available or extra computation isn’t needed, the model still functions as usual. - The model achieves lower perplexity. - This method works across various tasks without additional fine-tuning. Here are the details:

New, more interactive ways to learn about random subjects that I have no background in is one of the most fun use cases for AI that I came to enjoy.

Jeff Dean @JeffDean

a year ago

Personalized educational uses like this are one of the ways that capability advances in AI models will provide broad benefits. The idea that you can get a personalized tutor for any piece of information that knows you and knows how you learn best is going to be powerful! 🎉📚

14 73 467 80K 229

Gemini 2.0 Flash Thinking is looking super strong on Aidanbench, much more to come 😄

Here's a fun example of the model solving a problem in number theory, this was my first ever job interview question (jotted down on pen and paper)

keyboard_arrow_left Previous keyboard_arrow_right Next

A good thinker doesn't necessarily have to underperform in other tasks 😉

Arena.ai @arena

a year ago

Breaking news from Chatbot Arena⚡🤔 @GoogleDeepMind's Gemini-2.0-Flash-Thinking debuts as #1 across ALL categories! The leap from Gemini-2.0-Flash: - Overall: #3 → #1 - Overall (Style Control): #4 → #1 - Math: #2 → #1 - Creative Writing: #2 → #1 - Hard Prompts: #1 → #1

32 214 1K 331K 248

blatant racism. Simple as that.

It’s been an amazing last couple of weeks, hope you enjoyed our end of year extravaganza as much as we did! Just some of the things we shipped: state-of-the-art image, video, and interactive world models (Imagen 3, Veo 2 & Genie 2); Gemini 2.0 Flash (a highly performant and efficient foundation model); Gemini-Exp-1206 model (top of the Chatbot Arena leaderboard); 2.0 Flash Thinking (our first ‘thinking’ model, expect a lot more news on this soon - as many of you remember, we pioneered this type of model with AlphaGo, AlphaZero, AlphaProof…); upgrades to @GeminiApp with Deep Research and more, fantastic new NotebookLM features, a new image remixing tool (Whisk); and a series of agentic research prototypes that can help people get things done (Project Astra, Mariner, Jules) all built on Gemini 2.0. And then on top of all of that, we dropped the world’s most accurate weather prediction model (GenCast), the world’s most advanced quantum chip that performed a computation that would have taken 10 septillion years (Willow - could come in handy for training AI one day!), and it was the honour of a lifetime to receive the Nobel Prize for AlphaFold, which has revolutionised structural biology and is being used by 2 million researchers around the world to understand disease and accelerate drug discovery. I could not be more proud of all the exceptionally talented teams at @GoogleDeepMind & @Google who have worked so incredibly hard on all these amazing projects. It’s the greatest joy one can have professionally to get to explore the outer reaches of science and human knowledge with such wonderful colleagues. We’ve been inventing the future of AI for well over a decade now, and we’ll continue to be a relentless engine of innovation. In many ways we’re only just getting started… if you want to be at the frontier of the most exciting scientific and technological journey ever, this is the place to be, come join us!

@XueFz @m__dehghani Congrats and welcome!

We’ve been *thinking* about how to improve model reasoning and explainability Introducing Gemini 2.0 Flash Thinking, an experimental model trained to think out loud, leading to stronger reasoning performance. Excited to get this first model into the hands of developers to try out!