For instance, OpenAI's ChatGPT solution was able to answer low-order thinking questions used in the Canadian Royal College and American Board of Radiology exams, but struggled with higher-order ones involving image descriptions. When upgraded to a higher version, the software was able to answer these questions more easily, but made mistakes with some of the lower-order ones it originally answered correctly, raising questions about its reliability.

Additionally, using generative AI to build search engines often requires breaking down large data files and documents, and scanning information in segments, a process that takes considerable time, is typically not scalable, and does not generally provide quality insights based on multiple sources of information. Also, some technologies may not be able to process certain data formats, increasing the risk for errors.


To further refine its Gen App Builder, Google Cloud will incorporate feedback from Mayo Clinic, which plans to test the technology over the next several months in different use cases. In preparation for this, it has created what it calls “safe sandboxes” for workers to test applications, and identify where the technology holds the most promise, reported CNBC.

Back to HCB News

Artificial Intelligence Homepage