In the rapidly evolving world of AI, staying ahead requires a platform that not only meets the demands of today but anticipates the needs of tomorrow. Enter Google Cloud’s Vertex AI, a fully-managed, unified platform designed to streamline AI development. Leveraging over 150 first-party, open, and third-party models, Vertex AI offers unparalleled customization, monitoring, and deployment capabilities.
Leading companies like ADT, IHG Hotels & Resorts, ING Bank, and Verizon are harnessing the power of Vertex AI to accelerate their innovation cycles and maintain a competitive edge.
Key Announcements at Google I/O ‘24
At the recent Google I/O ‘24, several groundbreaking updates to Vertex AI were unveiled, promising to push the boundaries of what's possible in AI development. Here’s a closer look at the major announcements:
Available Today
- Gemini 1.5 Flash: Currently in public preview, Gemini 1.5 Flash is a lighter-weight version of the Gemini 1.5 Pro. It features an impressive context window of 1 million tokens, making it ideal for high-volume tasks where cost and latency are critical factors. This includes applications such as chatbots, image captioning, detailed video analysis, and extracting data from extensive documents.
- PaliGemma: This model marks the Gemma family’s first foray into vision-language capabilities. Available in Vertex AI Model Garden, PaliGemma excels in tasks like image captioning, visual question-answering, understanding text within images, object detection, and object segmentation. Its versatility adds significant value to the range of models available on Vertex AI, allowing developers to match the right model to their specific needs and budget.
Coming Soon
- Imagen 3: Set to be available this summer, Imagen 3 represents the pinnacle of text-to-image generation models. It is capable of producing highly detailed and photorealistic images, understanding the nuances of natural language prompts, and incorporating fine details from lengthy descriptions. Imagen 3 promises to transform how we generate visual content from textual descriptions, opening new avenues for creativity and precision.
- Gemma 2: Also debuting this summer, Gemma 2 is the latest iteration in the Gemma family of open models. It includes a powerful 27B model that delivers performance on par with much larger models. This new generation offers developers a robust tool for a wide array of AI applications, maintaining the flexibility and openness that the Gemma family is known for.
- Gemini 1.5 Pro with Expanded Context Window: Building on the success of Gemini 1.5 Pro, which already boasts a 1 million token context window, the new version will support an expanded context window of 2 million tokens. This enhancement is particularly beneficial for use cases that involve analyzing very large datasets, extensive codebases, or comprehensive document libraries. Interested users can sign up to join the waitlist for early access.
Enhancing Model Performance
Vertex AI is also rolling out several new capabilities designed to optimize model performance and reduce costs:
-
Context Caching: Set to enter public preview next month, context caching allows customers to manage and reuse cached context data effectively. This feature is crucial for applications with long-context requirements, helping to significantly cut down processing costs by leveraging already processed data.
-
Controlled Generation: Coming to public preview later this month, controlled generation enables customers to define the format and structure of model outputs. This capability is particularly useful when precise output formats like YAML, XML, or custom schemas are needed. JSON output format is already live, providing a starting point for developers to experiment with structured output generation.
-
Batch API: Available in public preview today, the batch API facilitates the efficient handling of large volumes of non-latency sensitive text prompt requests. This includes tasks such as classification, sentiment analysis, data extraction, and description generation. By allowing multiple prompts to be sent in a single request, the batch API enhances developer workflows and reduces operational costs.
Empowering AI Agent Development
Vertex AI continues to simplify the development and deployment of AI agents with new open-source integrations and tools:
Firebase Genkit: Announced at I/O, Firebase Genkit is an open-source Typescript/JavaScript framework designed to streamline the development, deployment, and monitoring of production-ready AI agents. Integrated with Vertex AI, Firebase developers can now easily utilize powerful Google models like Gemini and Imagen 2, alongside text embeddings, to create sophisticated AI solutions.
LlamaIndex: This integration simplifies the retrieval augmented generation (RAG) process, covering everything from data ingestion and transformation to embedding, indexing, retrieval, and generation. Vertex AI customers can now leverage Google’s advanced models and AI-optimized infrastructure in conjunction with LlamaIndex’s flexible data framework to connect custom data sources to generative models.
Grounding with Google Search: Now generally available, this feature allows AI outputs to be grounded in proprietary databases or designated sources of "enterprise truth." This integration combines Google’s latest foundation models with access to up-to-date, high-quality information from Google Search, significantly enhancing the accuracy and completeness of AI-generated responses.
The Future of AI with Vertex AI
With these new features and ongoing support for tools like LangChain on Vertex AI, Google Cloud reaffirms its commitment to providing developers with the cutting-edge tools needed to create intelligent, responsive AI agents. Vertex AI’s continued expansion and innovation ensure that organizations can maximize the performance of their generative AI models at scale, accelerating the journey from experimentation to production.
Get Started Today
Ready to experience the next wave of AI innovation? Start using Gemini 1.5 Flash on Vertex AI today. Discover how generative AI is transforming businesses by exploring our latest resources, including the ebook "Crossing the Generative AI Tipping Point: From Quick Wins to Sustained Growth," and delve into "101 Real-World Gen AI Use Cases from the World’s Leading Organizations" to see how others are leveraging AI to drive success.
