Updated Gemini 1.5 Pro report: MATH benchmark for specialized version now at 91.1%, SOTA 3 years ago was 6.9%, overall a lot of progress from February to May in all benchmarks

I confirm this is more what it’s like at Google, at least in research.

What counts isn’t your level, but the quality of your idea/project and your determination.

I’ve regularly led efforts which included L+1,2 contributors who did not try to take leadership or constrain me.

Education is one of the areas in which LLMs can do the most immediate good, even with their limitations, so I was excited to see that Google is fine tuning a tutor LLM.

Also, the comparison they used was Gemini 1.0 running a variation of our tutor prompt! The prompt alone did ok

We think of Google DeepMind as the engine room of @Google in the AI era. Thrilled to share our vision at #GoogleIO incl the latest Gemini model 1.5 Flash, Project Astra our universal AI agent effort, our new gen video model Veo, Imagen 3 & lots more! deepmind.google

A beautiful day for #GoogleIO Almost showtime!

Glad to support the Secure By Design pledge and reaffirm our commitment to building products with security at the core.

Thrilled to announce AlphaFold 3 which can predict the structures and interactions of nearly all of life’s molecules with state-of-the-art accuracy including proteins, DNA and RNA. Biology is a complex dynamical system so modeling interactions is crucial blog.google/technology/ai/…

1.8M+ researchers have used Google DeepMind’s AlphaFold protein predictions in their work on vaccine development, cancer treatment + more. AlphaFold 3 is the latest breakthrough, predicting the structures + interactions of all of life’s molecules with unprecedented accuracy.

Today we’re launching Google Threat Intelligence from Google Cloud to help cybersecurity professionals get better visibility of global threats. It uses Gemini’s advanced AI capabilities, plus expertise and insights from Mandiant (part of Google Cloud) + VirusTotal. cloud.google.com/blog/products/…

Results from our first ✨Med-Gemini✨models, variants of the base Gemini multimodal models that have been tuned for the medical domain.

The multimodal and long context capabilities of the Gemini models really shine in some of the capabilities.

Joint work with Vivek Natarajan

Today we’re announcing a $3B investment to expand our data center campuses in Virginia and build our newest data center in Fort Wayne, Indiana. We’re excited to support jobs and local communities across 26 US states. blog.google/inside-google/…

Proud that 50M students + teachers are using Chromebooks around the world, and that they’re the most-used laptop in K12 education. We’re launching 5 new Chromebooks built for Asia Pacific and Latin America, in partnership with local manufacturers. blog.google/outreach-initi…

Chag Sameach! Wishing everyone a wonderful celebration with family and friends.

From our 2023 US Economic Impact Report out today: our products + platforms helped generate $739 billion in economic activity for millions of businesses, creators, devs + more. We also saw more small businesses adopting AI tools to grow. #GrowWithGoogle blog.google/inside-google/…

New research from Google DeepMind brings together soccer and robotics. Using reinforcement learning, robots display agile and reactive movements similar to a soccer player, no shin guards needed:) science.org/doi/10.1126/sc…

Gemini 1.5 Pro is now available on the Vercel AI SDK Playground, in all its 1M-token context window glory.

This is a big deal. Large context windows make new classes of AI apps not just possible, but easy.
sdk.vercel.ai/s/uTHNh5mbs-Uk…

Gemma is expanding.... we just announced CodeGemma, a version of Gemma tuned for code generation. And bonus... Gemma is now bumped to v1.1, addressing lots of feedback we got.

Congrats Gemma team for one more amazing release!

developers.googleblog.com/2024/04/gemma-…

Eid Mubarak! As the month of Ramadan comes to an end, wishing health, peace and happiness to all who celebrate.

New versions of Gemma models with a bunch of improvements to instruction following, factuality, reasoning and more are now out.

⬇️

New Google developer launch today:

- Gemini 1.5 Pro is now available in 180+ countries via the Gemini API in public preview
- Supports audio (speech) understanding capability, and a new File API to make it easy to handle files
- New embedding model!

developers.googleblog.com/2024/04/gemini…