MADRID, Nov. 2 (Portaltic/EP) –
Google Cloud has announced new features for Vertex AI Search, which will allow you to develop more powerful search and chat applications, as well as generate even more personalized and safer recommendations thanks to Generative Artificial Intelligence (AI).
Vertex AI is a platform that enables the building, deployment and scaling of machine learning models and provides access to more than one hundred foundational models through the Model Garden service.
The company has recognized that, after launching Vertex AI Search in August of this year, it has aimed to capitalize on its experience in information retrieval and generative AI to put it at the service of companies seeking improve your productivity.
After developing new functionalities for Vertex AI, Google has announced new search tools assisted by artificial intelligence “much more personalized“, according to a press release.
Firstly, the company has commented that the idea is that developers can design the entries they use to generate summaries or responses with greater control over aspects such as style, tone, format or the length in which the information is presented. information.
On the other hand, he has commented that he has worked on refining the searches, so that organizations use their own data to improve the priority of results more precise.
Likewise, he has clarified that from now on Vertex AI Search will be able to use the REACT framework to divide a complex query into smaller queries and thus provide “results that help satisfy the needs of more sophisticated users”, as he has clarified in the writing.
Furthermore, Google has noted that Vector Search (formerly Vertex Marching Engine) indexes data as vector embeddings and finds those that are most relevant and at scale. With this, he has indicated that This platform has been redesigned with a more accessible interface, which allows developers to create and deploy their indexes without coding. Likewise, Google has worked to reduce indexing time, going from hours to minutes, as well as to improve filtering and documentation capabilities.
On the other hand, the company recalled that Vertex AI offers incorporation models for different data modalities and has presented a new multimodal incorporation model that supports text, image and video.
LINK RESULTS TO DATA
Google is aware that one of the things about generative AI that worries companies the most is the tendency toward hallucinations in basic models. For this reason, it has proposed two ways to link the expected results with the companies’ data.
On the one hand, it has announced that it is testing linking with data sets open to the general public, an option with which developers can put a broader set of information sources at the service of employees.
On the other hand, it has released a “basic element of Vertex AI Search”, generally available, which proposes that the results be supported by data from the organization, incorporating summaries and citations. “In this way, organizations can help users verify and validate the results obtained using various data sources,” she said.