• Analysis: It's striking that AI is already optimizing its own prompts. I suspect that, in a few years, most users won't need to bother with "prompt engineering" at all, since AI will be able to translate even clunky requests into well-specified prompts.

How to get the best AI-written summaries? Just. Keep. Asking. This intriguing working paper proposes a method called "chain of destiny" prompting to help AI create information-packed content summaries. It asks the AI to work through an iterative series of steps:

(1)    Draft a summary of a document.

(2)    Identify several elements from the source document that are missing in the draft summary.

(3)    Integrate the missing elements without adding any words to the summary.

(4)    Repeat steps 2 and 3 again … and again … and again.

The resulting summaries are quite rich and information-dense (perhaps too information-dense for comfortable reading). But what mostly interests me here is the approach: Rather than treating AI's first output as a finished product, the researchers treat it as an early draft that the AI can independently critique and improve upon. It’s a method that holds promise for many, many use cases.

While AI is not a new technology in healthcare, the remarkable advances in large-language models and generative AI across the past year have made 2023 the year every healthcare organization has had to contemplate the AI future of their business.  Join us to learn about the challenges to consider as your organization invests in AI  and key steps your organization should be taking to prepare for future developments