At the Google Cloud Next 2024 event, Google announced the launch of Vertex AI Agent Builder, the public preview of Google's most advanced generative AI model, Gemini 1.5 Pro, and the addition of open-source language models to the Vertex AI platform.
Vertex AI Agent Builder is a tool designed to help developers easily build generative AI experiences. It offers a no-code console for building AI agents using natural language to open-source frameworks like LlangChain on Vertex AI. The platform also streamlines the process of grounding generative AI outputs in enterprise data. "Vertex AI Agent Builder allows people to very easily and quickly build conversational agents," said Google Cloud CEO Thomas Kurian.
One of the key features of Vertex AI Agent Builder is its ability to build production-grade AI agents using natural language. Developers can create new agents in minutes by defining the goal they want the agent to achieve, providing step-by-step instructions, and sharing conversational examples. For complex goals, developers can stitch together multiple agents, with one agent functioning as the main agent and others as subagents. Agents can call functions or connect to applications to perform tasks for the user.
Vertex AI Agent Builder can improve accuracy and user experience by grounding model outputs in enterprise data, using vector search to build custom embeddings-based RAG systems. Developers also have the option of grounding model outputs in Google Search, combining the power of Google's latest foundation models with access to fresh information to improve the timeliness of results.
Google also announced the public preview of Gemini 1.5 Pro, which can process up to 1 million tokens, around four times the amount of data that Anthropic's Claude 3 model can handle and eight times as much as OpenAI's GPT-4 Turbo. This allows for tasks like analyzing code libraries, reasoning across lengthy documents, and holding long conversations. Google is making Gemini 1.5 Pro available in a public preview in over 180 countries via the Gemini API in Google AI Studio.
Can Gemini 1.5 actually read all the Harry Potter books at once? I tried it. All the books have ~1M words (1.6M tokens). Gemini fits about 5.7 books out of 7. I used it to generate a graph of the characters and it CRUSHED it. - Deedy
Google also announced the addition of Meta's Llama 2 and Anthropic's Claude 2 open-source language models to the Vertex AI Model Garden. This expands the variety of open-source models available on the Vertex AI platform, which already includes Google's own PaLM model. Vertex AI now provides access to over 100 curated models for developers to choose from.
Developers interested in learning more can check out the material on YouTube or read the wrap-up documentation.