This blog discusses new AI updates announced at Google IO event. Here are the key highlights:

Summary

• Google announced Gemini 1.5 Pro, an update to its advanced AI model.
• Gemini will be integrated into the side panel of Google Workspace apps like Gmail and Docs.
• A “lighter weight” version of Gemini called Gemini Flash was announced for developers.
• Google will use watermarking to increase transparency around AI-generated content with a tool called SynthID.
• Google plans to use a process called red-teaming to identify potential risks associated with its AI products.
• A new feature will allow users to set up a virtual AI teammate with its own Workspace account.
• Google announced a new feature that can detect and warn users about potential scam calls.
• Google Assistant’s “Circle to Search” feature will now support formulas.

Large language models (LLMs):

Google introduced a new LLM called PaLM 2 with improved multilingual capabilities, reasoning, and coding skills.
They also released a lightweight version, Gemini 1.5 Flash, designed for speed and low latency.
AI Assistants:

A new AI agent with reasoning, planning, and memory capabilities was revealed. This can perform multi-step tasks and collaborate with other software.
Project Astra, a universal AI chat assistant with near real-time response and vision abilities, was introduced.
Other updates:

A new text-to-image model, Imagine 3, was unveiled. It creates photorealistic images based on text prompts.
Music Effects allows generating music tracks from text descriptions.
Ask Photos enables searching photos based on specific details using text prompts.
Improvements to Google Search with AI overviews providing more comprehensive results.
Upgrades to Google Workspace with features like summarizing emails and meetings.
Notebook LM with audio generation capability to create summaries and answer questions based on notebook content.

Availability:

Some updates are already available (Gemini app update, Imagine 3 for testing).
Others are rolling out soon (AI overviews in Search, Ask Photos).
Summary of Google IO AI Updates (continued)
Here’s a continuation of the key updates announced at Google IO based on this transcript excerpt:

New Tools and Capabilities

Imagine 3 allows editing generated images and offers multiple outputs based on prompts.
Music AI Sandbox provides musicians with AI tools for music creation.
Veo, a new text-to-video generation tool (waitlist available).
Hardware advancements including new TPUs (Trillium), CPUs (Axion), and GPUs (Blackwell GPUs from Nvidia).
Multi-step reasoning in Google Search to find comprehensive information for complex queries (example: finding a yoga studio with specific criteria).
S Panal in Google Workspace integrates Gemini to automate tasks within workspace apps.
AI teammates – create a virtual teammate with specific tasks and access to team information (shown in Google Chat).
Data analysis with Gemini Advanced – allows code execution for data analysis within documents like spreadsheets.
Increase in context window size for Gemini Advance to 2 million tokens (available in 35 languages).
Circle Search in Android apps – identify elements and ask questions related to them.
Poly Gemma – open-source large language model with vision capabilities (available for testing).

Availability

Some updates are available now (Imagine 3 for testing, Poly Gemma).
Others have waitlists (Veo) or will be rolled out later (data analysis in Gemini Advance, 2 million token context window).