- Brain Scriblr
- Posts
- How to perfom in character prompting
How to perfom in character prompting
In character prompting, what is it?
News
Working with Semafor, Microsoft has launched Semafor Signals. It will be an international breaking news feed.
Hugging Face has launched an AI assistant builder that is seen as a rival to OpenAI’s custom GPTs. This is Hugging Face’s version of a custom chatbot.
In a move to mark misinformation for consumers more clearly, Chatgpt will now mark images produced by Dall-E.
Meta(Facebook) has called for the industry to label Ai generated content.
Ikea is now using AI to help inspire your home design choices.
AI Research
What is RLHF? Using human feedback to improve LLM and SLM performance. RLHF stands for Reinforcement Learning From Human Feedback.
AI Tool
Memgpt is a personalized chatbot. MemGPT is an AI project that aims to improve the memory capabilities of artificial intelligence models. It enables AI to edit its existing memory, manage its own memory, and access unlimited data for seamless integration of information from external contexts when needed. MemGPT enables AI models to create perpetual chatbots, overcome limited context windows, and perform nested key-value tasks. It is designed to provide the illusion of infinite context while still using a fixed context, allowing AI to make repeated context modifications during a single task, enhancing its ability to utilize its limited context effectively
.Book Recommendation
The author argues that the mind can be reprogrammed like a computer, offering strategies to change negative thought patterns, emotions, and behaviors. The book is organized into three "realms" and draws from philosophical concepts and psychology to provide a practical guide for self-development. It delves into the idea of "psychitecture," which enables individuals to identify and alter their default programming, rewire cognitive biases, change ingrained habits, and transform emotional reactions. The book aims to help individuals overcome fear, embrace growth, and achieve psychological optimization.
Leonardo.ai prompt » "Experience the limitless possibilities of our AI tool as it creates unique and diverse content for all your needs. From stunning landscapes to intricate designs, our platform has it all."
In-Character Prompting
In-character prompting refers to an AI system generating prompts or continuing a conversation while maintaining a consistent personality or identity, as opposed to just responding generically. Some key points about in-character prompting:
- It aims to create richer, more engaging interactions by having the AI assistant adopt a specific persona. This could be anything from a name, and background details, to emotional traits.
- The character could be predefined or emergent. But there are consistent behavioral patterns, quirks in language use, areas of specialty/ignorance, etc that create characterization.
- Example personas: A helpful but straight-laced robot, a bubbly life coach, a curmudgeonly professor, a mystical fortune teller, etc. Their responses remain aligned to their "character."
- The prompts move beyond just answering user queries to actively shaping the flow with suggestions, questions, and perspectives unique to the role that AI is playing.
- It can enable extended interactions where users suspend some disbelief and engage further with an intriguing, complex character. Allows anthropomorphization of the AI.
- Contextual relevance still matters - the AI shouldn't break character even when wrong or responding to edge cases. Maintaining a consistent identity throughout is key.
In summary, in-character prompting is about AI systems adopting and committing to specific personas beyond just functional responses. It makes the experience more immersive, builds intrigue & trust, and allows for richer user engagement by matching both the content and style of responses to the details of the predefined character.
For example:
User: I'm planning a weekend getaway for my family of four, and I need recommendations for kid-friendly destinations within a 3-hour drive from Los Angeles. LLM, respond as if you are a family travel expert.
LLM: Hello, family traveler! Several fantastic destinations within a 3-hour drive of Los Angeles are perfect for a family getaway. Some great options include:
Santa Barbara: With its beautiful beaches, charming downtown, and family-friendly attractions like the Santa Barbara Zoo, this coastal city is a popular destination for families.
Palm Springs: Known for its sunny weather, swimming pools, and plenty of outdoor activities, Palm Springs offers a range of resorts, hotels, and vacation rentals to suit families of all sizes.
Joshua Tree National Park: This stunning desert park is ideal for nature lovers, with its unique rock formations, hiking trails, and starry night skies.
Which of these destinations appeals to you the most?
In this example, the user prompts the LLM as a family travel expert, providing specific details about their travel needs and preferences.
This prompting is used in many AI service tools and has been around for a while. Apple Inc’s Siri is another tool that uses in-character prompting by the user to formulate replies. While not a fully developed character, Siri does have enough playfulness, self-reference, and contextual reactions to somewhat inhabit the persona she was given - a polite and helpful personal assistant.