Data Lakehouse Provider Databricks is presenting four new updates to their portfolio to help companies have more control over the development of their agents and other generative applications based on AI.
One of the new features launched as part of the updates is centralized governance, which is designed to help govern large language models, both open and closed source, within Mosaic AI Gateway. The function is currently in a public preview.
“Our research shows that governance is one of the main concerns that companies have about their AI initiatives, since it is complicated by the fact that there are multiple components in the process,” said David Menninger, executive director of the ISG advice firm.
The centralized governance capacity, according to the director of Technology and Experience of the West Monroe consulting firm, Doug Macwilliams, is a “quite large simplifier.”
“Ensures constant security, access controls and compliance, while reducing costs by eliminating duplicates and optimizing license rates. In addition, it facilitates monitoring and solving problems such as drift or bias, ”Macwilliams explained.
“In all this, you should also simplify the approval process for legal, compliance and security equipment, allowing them to review and approve models through a unique interface,” Macwilliams added.
SQL unique consultation to execute batch inference
To help companies execute an AI consultation without the need to configure the infrastructure, Databricks is adding a new capacity called Lotes without disposition.
The new skill, which is in public prior view, is a novel way to execute a lot inference through Mosaic AI with a single SQL query and companies pay the infrastructure they use, said Lakehouse’s supplier.
“The inference for lots without provisions is a great step forward for the implementation of AI, since it makes it easier to climb the AI and save costs only using resources when necessary,” said Macwilliams.
ISG Menninger sees the new skill as a functionality without server that eliminates the need to configure things in advance.
“Without this capacity, developers must do additional job: they need to supply or establish some resources to process inference applications,” Menninger explained.
In addition, Macwilliams believes that the SQL -based interface makes lot inference accessible to data analysts who have no experience in MLOPS.
“This opens new possibilities, such as the processing of millions of customer service tickets during the night to detect trends, enrich the data of the product catalog with descriptions generated by AI, execute regular compliance controls and rate client databases for the risk weekly, all without the need for special infrastructure,” Macwilliams explained.
Databricks has also updated its previously published agents evaluation review application that now allows domain experts to provide evaluations, send traces for labeling and define custom evaluation criteria, without the need for spreadsheets or custom applications.
“By facilitating the collection of structured comments, the teams (business) can continually refine the performance of the AI agent and boost systematic precision improvements,” the company explained.
In December, Databricks had updated its Mosaic IA agents evaluation module with a new synthetic data generation API that would be expected to help companies evaluate agents faster.
Genie API to extend data analysis to personalized and productivity applications
As part of the update, Data Lakehouse provider has introduced the API API suite AI/Bi Genie in a public prior view that is expected to help developers to embed the natural language chatbots directly in custom productivity applications or tools, such as Microsoft, SharePoint and Slack equipment.
Genie is a tool without code with an interface that allows users to analyze data by asking questions in natural language. The tool is capable of producing visualizations to explain the data.
“With the Genie API, users can send programming measures and receive ideas as they would in the genie user interface. The API is with the State, which allows you to retain the context in multiple monitoring questions within a conversation thread, ”the company wrote in a blog post.
The API, according to IDC’s vice president, Arnal Dayaratna, not only increases the extensibility of conversation attendees who take advantage of Databricks data, but also join the gap between data availability and accessibility, which allows faster derivation of processable ideas.
Another advantage of the API is that it democratizes access to data by allowing business users to interact with data using natural language, eliminating technical barriers such as SQL experience.
Alternatively, for developers, API reduces work by offering preconstructed conversation characteristics, so that they can focus on other important tasks instead of building these interfaces from scratch, Macwilliams said of West Monroe.
Comparing Genie’s API with the API of the Salesforce Agent of Salesforce recently launched, he said that Databricks’s version is more integrated with his data lake and BI Tools, make the analysis a little more conversational versus Salesforce’s approach to create independent agents.
This approach, according to Moor Insights and the main strategy analyst Jason Andersen, is very similar to the AWS approach with Amazon Bedrock.
Databricks’ Strategy and the agent landscape
Analysts also see updates such as Databricks strategy to approach business users and increase the stickiness of their offers.
“When unifying the data pipe Ai, Databricks is creating a platform that manages everything, from unprocessed data to operational, eliminating the need for other products,” said Macwilliams of West Monroe, and added that this strategy makes its platform more vigilant, reducing customer rotation and increasing income by expanding the user base within companies.
In the agent space, Menninger de ISG believes that Databricks has an advantage over others, since its approach is more technical, “allowing the creation of more complex agents potentially activities” in any domain with data.
But Menninger believes that this advantage occurs at the expense of who can create these agents, less likely to be commercial users.
“All suppliers are trying to win the advantage in agents wars. However, much of what is happening today is only ‘agents’ washing’, calling chatbots agents. The true agent capacity remains complicated and technical. It requires programming, ”Menninger said. “Salesforce and Servicenow seem to be very focused on conversation capabilities, which facilitates the creation of agents, but perhaps at the expense of what types of tasks the agents can perform.”
#Databricks #updates #aim #facilitate #development #GEN #application #agent
