AI agents are the new hot craze in generative AI. Unlike the previous generation of chatbots, these agents can do more than simply answer questions. They can take actions based on the conversation, and even interact with back-end transactional systems to take actions in an automated manner.
On Tuesday at Google Cloud Next, the company introduced a new tool to help companies build AI agents.
“Vertex AI Agent Builder allows people to very easily and quickly build conversational agents,” Google Cloud CEO Thomas Kurian said. “You can build and deploy production-ready, generative AI-powered conversational agents and instruct and guide them the same way that you do humans to improve the quality and correctness of answers from models.”
The no-code product builds upon Google’s Vertex AI Search and Conversation product released previously. It’s also built on top of the company’s latest Gemini large language models and relies both on RAG APIs and vector search, two popular methods used industry-wide to reduce hallucinations, where models make up incorrect answers when they can’t find an accurate response.
Are AI models doomed to always hallucinate?
Part of the way the company is improving the quality of the answers is through a process called “grounding,” where the answers are tied to something considered to be a reliable source. In this case, it’s relying on Google Search (which in reality could or could not be accurate).
“We’re now bringing you grounding in Google Search, bringing the power of the world’s knowledge that Google Search offers through our grounding service to models. In addition, we also support the ability to ground against enterprise data sources,” Kurian said. The latter might be more suitable for enterprise customers.
In a demo, the company used this capability to create an agent that analyzes previous marketing campaigns to understand a company’s brand style, and then apply that knowledge to help generate new ideas that are consistent with that style. The demo analyzed over 3,000 brand images, descriptions, videos and documents related to this fictional company’s products stored on Google Drive. It then helped generate pictures, captions and other content based on its understanding of the fictional company’s style.
Although you can build any type of agent, this particular example would put Google directly in competition with Adobe, which released its creative generative AI tool Firefly last year and GenStudio last month to help build content that doesn’t stray from the company’s style. The flexibility is there to build anything, but the question is whether you want to buy something off the shelf instead if it exists.
The new capabilities are already available, according to Google. It supports multiple languages and offers country-based API endpoints in the U.S. and EU.
https://techcrunch.com/2024/04/09/google-cloud-next-2024-everything-you-need-to-know/
