At its annual Google Cloud Next developer conference this week, Google announced several new artificial intelligence and cloud computing products and enhancements.
Duet AI Comes to Google Workspace
Google unveiled the general availability of Duet AI in Google Workspace, bringing AI assistance directly into tools like Gmail, Docs, Sheets, Slides, Meet and more. Duet AI provides features like summarizing documents, auto-generating slides, translating languages in real-time during meetings, and suggesting creative ideas.
Duet AI is powered by Google’s AI models and runs securely on Google’s cloud infrastructure. Google says none of users’ data is used to train its AI models, providing strong privacy protections.
New AI Tools for Developers
For AI developers, Google announced updates to its Vertex AI platform. Vertex AI helps developers manage the entire machine learning lifecycle including model training, deployment, monitoring and more.
New Vertex AI capabilities include support for third-party AI models like Meta’s Code Llama in the Vertex Model Garden. Google also previewed code generation and migration features using Duet AI in development environments like VS Code.
Realistic AI-Generated Images
Google unveiled a new AI system called SynthID that can embed invisible digital watermarks into AI-generated images. These imperceptible watermarks act as a form of identification that confirms an image was synthesized by AI.
SynthID watermarks are designed to be resistant to modifications and tampering. This makes it more difficult for bad actors to remove the watermarks from fake AI images before using them to spread disinformation or commit fraud.
According to Google, SynthID’s watermarking technique is an important development towards regulating the proliferation of synthetic media. The ability to definitively identify AI-generated images will assist efforts to police the spread of disinformation enabled by advanced generative AI systems.
Faster AI Hardware and Infrastructure
The company unveiled its latest Tensor Processing Unit (TPU) AI accelerator, the TPUv5. The TPUv5 delivers higher performance and efficiency for training complex machine learning models.
For its cloud infrastructure, Google announced the general availability of its network attached storage product Filestore High Scale, which provides exabyte-level scale for AI, analytics and other workloads.
AI for Structured Data
Google announced AlloyDB AI, which brings AI capabilities like natural language search directly into its fully managed PostgreSQL-compatible relational database AlloyDB.
AlloyDB AI allows developers to apply Google’s AI models to their structured data in AlloyDB to generate insights. It also helps migrate Oracle databases to AlloyDB using AI tools.