In a major announcement, Google Cloud has rolled out a series of updates to its Vertex AI platform, including new Gemini 1.5 Flash and 1.5 Pro models with expanded context windows, grounding with Google search, and pre-made Gems for Google Workspace. The tech giant also introduced a suite of AI agents designed for customer engagement and conversational experiences.
The updates underscore Google Cloud’s push to drive Gemini, its foundational large language model, across its platform and enable enterprises to tap into multiple AI capabilities. Key highlights include:
- Gemini 1.5 Flash and 1.5 Pro models with 2 million context window, double the previous capacity
- Controlled generation to dictate output format
- Prompt Optimizer, Prompt Management SDK, and GenAI Evaluation Service
- Distillation and supervised fine-tuning for Gemini models
- Grounding Gemini with Google search and dynamic retrieval
- Premade Gems in Google Workspace for tasks like brainstorming and coding
- Vids, a new Workspace feature that can generate stories from a single prompt
- 1.5 Flash model for customer engagement, along with conversational and summarization agents
- Google Cloud sees agents as a critical part of the enterprise AI transformation, enabling new applications and experiences. The company segments agents into three categories: pre-built agents in Google Workspace, Google-built agents for customer engagement, and custom agents built by enterprises.
The key drivers for agentic AI, according to Google, are the complexity of workflows and the need for agency - the ability of agents to learn, make decisions, and take action with minimal human supervision. Enterprises are expected to progress from task-specific agents to assistants and eventually multi-agent systems that can tackle complex end-to-end tasks.
Overall, the announcements highlight Google Cloud’s ambition to make Gemini and generative AI a core part of its enterprise offerings, empowering customers to build intelligent agents and transform their customer experiences.