Google unveils Gemini AI for robots struggling with dexterity

Oracle is set to incorporate Google's Gemini models into its AI Agent Studio for Fusion Applications, enhancing its collaboration with Google Cloud and offering customers an alternative for constructing AI agents. The company aims to utilize Gemini models for embedded AI applications within Fusion Applications and Oracle NetSuite.

According to a statement from Google Cloud VP Satish Thomas, the partnership aims to simplify the use of Gemini in applications and agentic workflows, thereby enabling organizations to automate tasks, expedite decision-making, and achieve desired outcomes. Gemini 3.1 Flash Lite and Gemini 3.5 Flash models will be accessible through AI Agent Studio, complementing Oracle's existing model options.

Chris Leone, Oracle's VP of Applications Development, emphasized that this integration provides customers and partners with enhanced flexibility when creating and enhancing agents and agentic applications capable of resolving intricate, real-world business issues. When Oracle initially unveiled its platform for integrating LLM-powered agents into Fusion applications, the intended migration route for numerous corporations worldwide using Oracle applications, it affirmed that customers could employ models from Cohere and Meta, while connecting other supported models via the platform.

The Register has reached out to Big Red to ascertain whether Cohere and Meta models will continue to be available; an update will be provided once a response is received. Following Oracle's introduction of Fusion Agentic Applications earlier this year, Gartner raised concerns regarding potential implementation challenges. Balaji Abbabatulla, Gartner's VP and vendor lead analyst for Oracle, remarked cautiously, stating, "On the surface, this appears promising, yet be wary.

The complexities underneath may not be fully addressed yet. However, improvements might occur over time." He also highlighted that Oracle and other vendors need to address the issue of accountability for AI decision-making should the technology fail. Should an AI agent make an erroneous decision at scale and speed, errors could propagate before detection becomes feasible.

So far, Oracle's approach to this matter is centered around monitoring and audit tooling. However, Abbabatulla remains unconvinced, asserting, "I do not observe a clear response from any vendor regarding liability."