iiinova — where your digital self keeps evolving

My first @waymo.bsky.social ride on the highway! Very smooth. Awesome work, Waymo team!

41015

Shaping AI’s Impact on Billions of Lives

shapingai.com points to the Arxiv version, the CACM version, as well as a short op-ed from The Economist by our co-author Dave Patterson about the work, and NotebookLM-generated audio podcasts from the Arxiv version. PDF link for CACM version is a bit hard to find: dl.acm.org/doi/pdf/10.1...

shapingai.com

AI’s impact on society will be profound, affecting billions of lives for better or worse. Focused efforts made now can maximize the upsides of AI and minimize its downsides for everyone.

061

It was a delight to work on this with an awesome set of co-authors: Mariano-Florentino Cuéllar, Finale Doshi-Velez, John Hennessy, Andy Konwinski, Sanmi Koyejo, Pelonomi Moiloa, Emma Pierson, and David Patterson.

130

The CACM version is shorter version of Arxiv version. We cover AI's potential impact on employment, education, healthcare, (mis)information, media, national security, and science, and propose 18 potential moonshot research directions in these areas to maximize positive impact and minimize downsides.

120

Nute @nutedawn.bsky.social · 06/01/2026

In 2024, I was a co-author on "Shaping AI's Impact on Billions of Lives", covering areas we thought AI would have significant impact on the world. A shortened version appears as the cover article in this month's CACM: cacm.acm.org/research/sha... arxiv.org/abs/2412.02730 See: shapingai.com

1427

Reposted by Jeff Dean

It’s genuinely wild that the United States spent decades building an elaborate rules-based international order which de facto cemented its permanent status as the dominant global hegemon, sustained by ideology rather than might, and now is just blowing it all up for literally no reason

421111852509

Jeff Dean @jeffdean.bsky.social · 22/12/2025

Good catch. Fixing and will roll out with next update. Thanks.

020

The internal version also has additional examples in code that is more sensitive (e.g. changes to the TPU compiler backend) that we aren't able to include in the external version.

190

The public version is a somewhat sanitized version of the internal version. People at Google can find the internal version at go/performance-hints, which has links to the actual changelist in our source code repository system.

1100

Sanjay and I also had fun in writing this, plumbing the depths of changelists over multiple decades looking for interesting techniques we'd used and trying to group them together into higher level themes.

160

Many techniques are probably known to readers, but we felt it was nice to have a collection of tips and tricks, ranging from high-level algorithmic improvements to low-level performancer optimizations collected together.

150

Others are specific small snippets of code that demonstrate a technique.

170

We've tried to give concrete examples of the various techniques. Some are high level descriptions of a set of performance improvements, like this set of changes from working on Google's search system in 2001.

181

My colleague Sanjay Ghemawat & I have done a fair bit of performance tuning of various pieces of code. We wrote an internal Performance Hints document ~2 years ago as a way of identifying some general principles & we've recently published a version of it externally. Doc: abseil.io/fast/hints.h...

213225

The Gemini 3 Flash and Pro models are a huge team effort. Thank you to everyone who worked very hard over the past months to train and get these models ready for release! 🎉 We’re all very excited to see what people do with the new capabilities offered by the Gemini 3 models!

0201

Gemini 3 Flash is available now in the API, and rolling out today as the default model in AI Mode in Search and Gemini app globally.

1191

Not only is Gemini 3 Flash higher quality than 2.5 Pro it is also 3x faster (based on Artificial Analysis benchmarking) at a fraction of the cost. Here’s a side-by-side demo.

3190

One of the things we strive to do with each new Gemini release is to make the new Flash model as good or better than the previous model’s Pro model. Gemini 3 Flash exceeds Gemini 2.5 Pro on nearly every metric, often by very large margins, and almost matches Gemini 3 Pro on most benchmarks.

1211

We’ve pushed out the Pareto frontier of efficiency vs. intelligence again. With Gemini 3 Flash ⚡️, we are seeing reasoning capabilities previously reserved for our largest models. This opens up entirely new categories of near real-time applications that require complex thought. More in thread ⬇️

213019

Some key aspects that are shared: ✅ Single model that can generalize to millions of tasks ✅ Multimodal inputs and outputs ✅ Sparsely activated (MoE is a bit more structured than the original Pathways vision) ✅ Use the Pathways compute infrastructure to enable training and serving

0180

Kevin Webb @kevinwebb.io · 19/11/2025

It is using our Pathways software infrastructure for all of its training and serving. It also is fulfilling the vision of Pathways of a single model that understands multiple modalities as input and can generate multiple modalities as output and is sparsely activated.

3160

Reposted by Jeff Dean

The multimodal reasoning is very cool. Something I’ve been tracking with models for a while and Gemini 3 Pro was the first to produce a plausible response.

0184

Strange. Works for me (manually typing the URL from your screenshot), and that redirects to storage.googleapis.com/deepmind-med... What was the source URL of your left screenshot?

storage.googleapis.com

100

It is also happy to critique your pickleball technique from a video.

1120

The multimodal and multilingual capabilities of Gemini 3 are quite strong. Here's an example of reading and organizing many different kinds of handwritten recipes and then vibe-coding a bilingual web site.

2202

By combining deeper reasoning with improved tool use, Gemini 3 can take complex actions on your behalf. Here’s an example of Gemini 3 organizing your inbox. Google AI Ultra subscribers can try these agentic capabilities in the Gemini app with Gemini Agent today.

1130

Gemini 3’s long-range planning capabilities are significantly improved. Here is a graph of a few different models running a simulated vending machine business (benchmark from andonlabs.com/evals/vendin...). Gemini 3 is the blue line.

1132

As an example of how we are building on top of Gemini 3, AI Mode in Search now uses Gemini 3 to enable new generative UI experiences, all generated completely on the fly based on your query. Here’s how you might use this to learn a complex topic like how RNA polymerase works.

1164

Meet Gemini, Google’s AI assistant. Get help with writing, planning, brainstorming, and more. Experience the power of generative AI.

We’ve built many exciting new product experiences with it: some were released today and some will be in the coming weeks & months. You can find it today on gemini.google.com and AI Mode in Search. For developers, you can build with it now in aistudio.google.com and on cloud.google.com/vertex-ai

gemini.google

‎Google Gemini

1181