Google is making efforts to address the concerns of enterprise customers by ensuring that its AI platform minimizes inaccuracies and hallucinations. This is crucial for organizations whose executives are already skeptical about AI technology. To build trust, Google is focusing on model grounding and is partnering with reputable third-party services such as Moody’s, MSCI, Thomson Reuters, and Zoominfo. These services will be available within Google’s Vertex AI starting next quarter, providing developers with qualified data to verify the accuracy of their model outputs.
For enterprise developers, this means they can leverage high-quality data and expertise from Subject Matter Experts to meet their standards. Moody’s Chief Product Officer, Nick Reed, explains that Google’s service offering allows other people to use their data to ground their own responses in assistant applications. By combining Google’s models, data, and Google Search, customers can ensure that their interactions are based on accurate information.
Previously, Google offered Google Search as part of its grounding service, allowing businesses to supplement Gemini outputs with reliable information from the web. However, with the addition of trusted partners, companies now have more options for accessing source-of-truth information without the risk of inaccurate data from blogs that may have manipulated SEO rankings.
In addition to the grounding partnerships, Google is introducing high-fidelity grounding through an experimental preview. This feature enables AI systems to work better with specific information, particularly for tasks like summarizing multiple documents or extracting important data from financial reports. High-fidelity grounding is powered by Google’s Gemini 1.5 Flash.
When asked if these new partnerships and high-fidelity grounding make Google’s generative AI the most reliable factual model compared to competitors, Google Cloud Chief Executive Thomas Kurian highlights three key differentiators. Firstly, Google utilizes its reputation as the most trusted source of web data and real-world understanding through Google Search to ground AI against web data. Secondly, Google provides options for grounding data in specific sub-verticals, giving users control over what they want to ground against. Lastly, high-fidelity grounding directs AI to focus on the prompt, improving the quality of the response by considering relevant information rather than all the data the model was trained on.
These efforts to improve model grounding and reduce inaccuracies aim to enhance trust in AI models. This announcement aligns with other Google Cloud news, including the public release of Gemini 1.5 Flash and 1.5 Pro, Gemma 2, and other updates to Vertex AI. By prioritizing accuracy and providing reliable data sources, Google is seeking to win over enterprise customers and position itself as a leader in the AI market.