Google has given its Vertex AI Search functionality new generative AI features. This allows businesses to develop even better and more powerful applications for search and chat functionality as well as personalized recommendations.
Google Vertex AI Search, which became generally available in August this year, is intended to help companies, employees, and other stakeholders quickly discover information, extract hidden insights from data and improve productivity.
One seeks to accomplish this by being able to quickly build applications and integrate them with proprietary integrations or LLMs. Ultimately, this should produce a set of applications that combine generative AI and semantic search. Companies can then use these for a variety of applications, such as Web site search or digital assistants.
New features
With the new generative AI properties in Google Vertex AI Search, developers can enable even more AI-supported search functionality. These new properties also provide better grounding and compliance capabilities.
The new generative AI properties for improving search functionality include customizable answers, refining search queries and answers, and introducing “proprietary” search engines with vector search as well as Google Vertex AI integrations for using different AI models.
Fending off hallucinations
For the new ‘grounding’ functionality, which should counter potential hallucinations of generative AI, several options have now emerged. The feature will ground data in information with which enterprises are most comfortable. These options are grounding in enterprise data or grounding in selected public datasets.
Further enhancements have been made to compliance functionality. Among other things, this now includes support for access transparency, so customers know what access Google employees have to their data.
In addition, Virtual Private Cloud Service Controls (VPC-SC) ensure that customer employees cannot pull in or filter out data. Finally, Customer-managed Encryption Keys (CMEK) ensure that customers can encrypt their core content with their own encryption keys.
Also read: Google Cloud expands Vertex AI, more than 100 LLMs available
2 days ago
