By Google Cloud CEO Thomas Kurian revealed that reinforcement learning is enhancing Gemini Pro’s capabilities, a new API introduced by Google for developers and organizations. The rollout includes an array of AI tools, models, and infrastructure, accessible through Google Studio and the Vertex AI platform for enterprises.
“We’re rapidly integrating Gemini’s cutting-edge abilities across our services, extending it to every facet within Google,” Kurian expressed during an interview with indianexpress.com.
In addition to Gemini Pro, Google introduced various models within Vertex AI, empowering developers and enterprises. Among these offerings are an upgraded Imagen 2 text-to-image diffusion tool, specialized healthcare-focused models like MedLM (available to US Google Cloud customers), and the general availability of Duet AI for developers and security operations.
Kurian emphasized Google’s structured approach to model introduction, starting with preview phases to gather feedback from developers and users. This collaborative process aids in refining and evolving the models based on valuable insights.
Acknowledging the importance of user feedback, especially concerning Gemini’s applications in information search, brainstorming, and coding, Kurian highlighted Google’s engagement in close collaborations. These collaborations span diverse domains, recognizing the unique language nuances and requirements across industries like legal, medical, and creative writing.
Josh Woodward, VP of Google Labs, added that active participation in hackathons enables them to stay connected to user needs, influencing the design and development of their tools.
While Google is making Gemini Pro available for developers and enterprises, they announced plans to fine-tune the model in the upcoming weeks and months. The initial version of Gemini Pro is now accessible via the Gemini API, outperforming similar-sized models on research benchmarks. Future iterations are expected to offer larger text context windows. Currently free with limitations, Google intends to competitively price it in the future.
