Google partners with Moody’s to counter its AI’s hallucination problems

Google is ramping up efforts to bolster the reliability of its AI systems by integrating real-world factual data, with a notable new partnership with Moody’s to leverage financial information.

In a bid to tackle the issue of AI generating inaccurate or misleading information, Google Cloud has expanded its initiative to ground Vertex AI results in verifiable data sources, including web search and internal enterprise data. Now, the tech giant is introducing a novel approach: incorporating third-party data from partners such as Moody’s, Thomson Reuters, and ZoomInfo to further enhance the accuracy of AI-generated outputs.

According to Google Cloud CEO Thomas Kurian, this partnership aims to build trust in AI models by ensuring they are anchored in credible, factual information. “You can actually trust the model to do a task on your behalf because you have a basis for trusting it,” Kurian emphasized in an interview with Axios.

The initiative underscores a broader industry effort among leading AI providers to validate the safety and dependability of generative AI systems for business applications. Google’s introduction of a “confidence score” feature further enhances transparency, providing users with a numeric assessment of the model’s confidence in its responses.

Moreover, Google is rolling out capabilities that allow customers to instruct AI models to prioritize information from specific documents or sources, reducing reliance on generalized training data. “We’ve taught the model how to guarantee that when it responds, it takes what’s in the input prompt as the primary information it needs to pay attention to,” Kurian explained, highlighting efforts to mitigate distractions from extensive training datasets.

In addition to these enhancements, Google has announced the general availability of its low-latency Gemini 1.5 Flash model and Gemini 1.5 Pro, capable of processing extensive contexts equivalent to up to two hours of video content.

These developments mark a significant step forward in fortifying AI systems against inaccuracies and reinforcing their utility and reliability across various business sectors. As Google continues to innovate and collaborate with industry leaders like Moody’s, the trajectory towards more trustworthy AI applications in enterprise settings appears promising.