Check out [[Daily Reading]] for context. ## Today's Picks of the Internet Today, we have a couple of articles around **Generative AI**. #### [Prompt Structure in Conversations with Generative AI](https://www.nngroup.com/articles/ai-prompt-structure/) A deep dive into prompt structures. The authors use a *iceberg model* to drive the point home - In very prompt, there are components that explicitly ask the LLM to do something and implicit components that assume the LLM's knowledge. Our task as users of generative ai is to reduce the uncertainty associated with this implicit component by providing as much *context* as one can. There is also a short, but useful section on *recommendations for designing the user experience (UX) for generative ai apps).* Also, if you were interested in ChatGPT's prompt engineering guide, do check out [[ChatGPT Prompt Engineering Guide Notes]] #### [How real is the threat of data poisoning to generative AI](https://techmonitor.ai/technology/ai-and-automation/how-real-is-the-threat-of-generative-ai-data-poisoning) What if you could *poison* training data to present an incorrect view of training data to your LLM? Would this be an effective way to protect artists from having AI train itself on their works? How big of a concern would this be for corporates? --- #### Like what you see? Would you like to support me? Easy! Head over to [this link](https://refind.com/?invite=7b7e76f6e0) and subscribe to receive **Refind** newsletters. Every day Refind picks the most relevant links from around the web for you. Loved by 400k curious minds.