Google I/O 2024 Keynote live blog: Android 15, Gemini, and AI

I/O 2024 kicks off today as Google’s biggest event of the year to share what’s new for users and developers. Gemini and AI will be a big focus, while we expect to learn more about Android 15.

How to watch Google I/O 2024

I/O 2024 at the Shoreline Amphitheater in Mountain View, California starts with a two-hour keynote led by Alphabet and Google CEO Sundar Pichai at 10 a.m. PT / 1 p.m. ET / 5 p.m. GMT. There’s a live audience for this two-day conference, and you can stream the event live on YouTube.

Live Blog (Updates in reverse chronological order…)

• “AI” was mentioned 121 times during the keynote, according to Gemini Advanced
• LearnLM, based on Gemini, to make “learning experiences more personal”
  • Powering Gems “Learning Coach”
• SynthID watermarking coming to text, which will be open-soruced
• What’s new for developers at Google I/O 2024
• On-device Gemini Nano adding multimodality on Android. Coming to Pixel later this year
  • Gemini Nano-powered TalkBack coming later this year
  • “Likely scam” alerts in phone calls.
  • Gemini dynamic suggestions
• Google is on a multi-year journey to reimagine Android with AI at its core
  • Circle to Search coming to 200M Android devices by year’s end
  • Gemini becoming a foundational part of the Android experience
  • Gemini app now opens as an overlay to preserve context instead of opening a fullscreen UI
  • Ask this PDF + Ask this video capabilities

• Gemini Advanced trip planning
• Gems are custom Gemini
• Gemini Live: Talk naturally to Gemini with the ability to interrupt
  • Adding camera capabilities from Project Astra later this year

• Gemini AI Teammate in Google Workspace
  • For 2025. You tell it what to do.
  • Can have a collective memory
• AI workflow automations coming to Workspace
• Gmail on Android and iOS is getting a “Summarize this email,” Q&A (which is like side panel on desktop but for mobile), and Contextual Smart Reply. Rolling this month to Workspace Labs.

• Ask questions with video in Google Search. Using Google Lens. Understand question, breakdown the video frame-by-frame, plugged into Gemini long context window
• AI-organized search results page, starting with dining and recipes
• Multi-step reasoning in Google Search for planning use cases
• “Google does the work for you” is the company’s pitch for gen AI search. Simplifying queries that take 10+ questions
• Google Search: Real-time information + ranking and quality systems + Gemini
• Trillium is Google’s 6th-gen TPU (Tensor Processing Unit). 4.7x compute improvement. Late 2024 to Cloud customers
• Veo: Text-to-video with improved consistency, quality, and output resolution. HQ 1080p video. Can try in VideoFX, waitlist at labs.google.
• Music AI Sandbox
• Imagen 3: Can incorporate small details in a longer prompt. Best model yet for rendering text

• Demoing live at I/O for attendees
• Some capabilities will come to Gemini app this year
• “New exciting form factors like glasses”
• Astra running on prototype smart glasses. Looks the same as 2022 translation glasses
• Early prototype of Project Astra. Single-take in real-time
• Project Astra: Aim to build “universal AI agent helpful in everyday life”
  • Combines the video and speech input into a timeline of events. Caching this information for efficient recall.

• Gemini 1.5 Flash announced: lighter weight model than 1.5 Pro. For use cases where low latency and cost matters
• Sir Demis (Hassabis) of Google DeepMind takes the stage. Long-term goal of wanting to build AGI, human-level cognitive capabilities
• Google’s goal: “Making AI helpful for everyone”
• AI “Agents”: intelligent assistants the show reasoning, planning, memory. Can think multiple steps ahead. Work across software and systems. Under your supervision.
• Gemini 1.5 Pro coming to NotebookLM with “Audio overviews.” Generated audio discussions, with users able to join the conversation and steer in an impressive multimodality demo

• Gemini 1.5 Pro is now available in Gmail, Workspace side panel. Starting in Workspace labs
• Expanding Gemini 1.5 Pro context window to 2M tokens for devs in private preview

• Gemini Advanced now uses Gemini 1.5 Pro with 1M tokens across 35 languages
• Gemini 1.5 Pro improvements in translation, coding, reasoning. Updated version available globally today
• Google Photos getting “Ask Photos” with Gemini feature: Ask for your license plate + photos of your child swimming over time. Conversational search.

• AI Overview are rolling out in Google Search in the US this week. More countries soon

• 1M+ Gemini Advanced sign-ups “in just 3 months’
• All of Google’s 2 billion user products use Gemini
• 1.5M+ developers uses Gemini
• “Any input into any output”
• “We’re in the very early days of the AI platform shift.” — Pichai
• There are a few thousand developers at Shoreline today

• CEO Sundar Pichai has taken the stage
• A video is highlighting what Google has announced in the past year on the AI front
• We’re starting!
• 5-min warning
• T-shirt cannon time
• Bring back the bird, but until then we have a demo of MusicFX:

• The stream is now playing tunes created using Google’s image-to-music generation AI models
• On the screen at Shoreline: labs.google/gendino/
  • The gen AI game/experience will be available until 10 a.m. PT
• The YouTube livestream is, well, live. And we’re now in our seats!