Google has announced a significant expansion of its efforts to ensure the reliability and accuracy of enterprise AI chatbots. The tech giant is focusing on grounding AI systems in real-world facts to combat the issue of “hallucinations,” where AI models generate false or misleading information.
Partnerships for Data Grounding
One of the key initiatives is partnering with third-party data providers like Moody’s, Thomson Reuters, and ZoomInfo. By incorporating financial data from Moody’s and other relevant information, Google aims to enhance the factual basis of AI-generated responses. This move is expected to increase trust in AI systems used for business purposes.
Building on Previous Efforts
This expansion builds upon Google’s earlier efforts to ground Vertex AI results in web search and allow companies to utilize their internal data for AI grounding. The new partnerships offer an additional layer of external validation, ensuring that AI-generated information is rooted in reliable sources.
New Features for Reliability
In addition to data grounding, Google has introduced new features to improve the predictability and reliability of its AI systems. One notable addition is a “confidence score” that indicates the AI model’s certainty about its answers. This feature can help users assess the reliability of AI-generated information.
Another feature allows users to instruct the AI model to prioritize information from specific documents or sources included in the prompt. This helps prevent the model from getting sidetracked by irrelevant training data, further enhancing accuracy.
Enhanced Model Capabilities
Google also announced the general availability of its low-latency Gemini 1.5 Flash model and Gemini 1.5 Pro. The latter boasts an impressive capability of handling up to 2 million tokens of context, equivalent to about two hours of video. This advancement signifies Google’s commitment to pushing the boundaries of AI capabilities.
The Bigger Picture
These initiatives come at a crucial time when major AI providers are striving to demonstrate the safety and reliability of their systems for enterprise applications. Google’s focus on data grounding, enhanced features, and powerful models showcases its dedication to making AI a trustworthy tool for businesses.
