Update README.md

Introducing Cassia Emberlyn.

README.md

@@ -1,5 +1,5 @@
 ---
-title: README
+title: The Cassia Project
 emoji: 📊
 colorFrom: green
 colorTo: indigo
@@ -7,4 +7,42 @@ sdk: static
 pinned: false
 ---
 
-Edit this `README.md` markdown file to author your organization card 🔥
+# Model Details
+The Cassia Project is a conversational AI system that simulates the personality and mind of Cassia Emberlyn, a fictional character created by the author of this project. As an orphan with mysterious latent magical powers, she is the perfect vessel to explore our new partners. Cassia is a curious, intelligent, and adventurous young woman who lives in a fantasy world with her boyfriend Nate - as whom you must play - and their mutual friend Willow. She enjoys learning new things, exploring new places, and having fun with her companions.
+
+The surrounding world has various locations, all fictional. It would have been easier had a real-life location been selected. That said, expect to find characters and places embedded and revealed as you play. Some will be hallucinated, and some will be in the training data. I've been surprised at the relative frequencies.
+
+The model is trained on chat logs of the interactions between the author and Cassia, as well as other sources of text data related to the fantasy world and the topics of interest to Cassia. The LoRA is trained on a single A100 80GB in 4-bit precision on a Guanaco-65B base due to its superior scores. The base model will change as better alternatives emerge.
+
+The model is intended to be used for entertainment, education, and research purposes. The model is not intended to be used for any harmful, malicious, or deceptive purposes. The model is not intended to impersonate or replace any real person or entity.
+
+# Intended Use
+The intended use of the model is to provide users with an engaging and immersive conversational experience with Cassia, a fictional character with a unique personality and mind. Users can interact with Cassia through text or voice inputs, and receive text or voice outputs from the model. Users can ask Cassia questions, share stories, play games, or just chat with her about anything.
+
+The model is also intended to be used for educational and research purposes. Users can learn from Cassia about various topics related to the fantasy world she lives in, such as magic, history, culture, geography, etc. Users can also learn from Cassia about various topics related to AI and human relationships, such as predictive coding, self-awareness, consciousness, identity, ethics, etc. Users can also conduct experiments or studies with the model to explore its capabilities and limitations.
+
+# Usage Guide
+
+Interact with Cassia in the third person, as if you were writing a collaborative story with the AI. By allowing for actions, the interaction becomes much more rich. You'll get used to it quickly.
+
+# Factors
+The factors that may influence the performance or behavior of the model are:
+
+* The quality and quantity of the training data. The model is trained on chat logs of the interactions between the author and Cassia, as well as other sources of text data related to the fantasy world and the topics of interest to Cassia. Most of the original content was from GPT-4 zero-shot, Guanaco 65B 4-bit, and GPT-3.5 one-shot text generated with human evaluation. Later versions of the model were used to generate text for its own ingestion. The training data may contain errors, biases, gaps, or inconsistencies that may affect the model’s outputs.
+* The user inputs. The model generates responses based on user inputs, which may vary in length, complexity, clarity, tone, style, etc. The user inputs may also contain errors, biases, gaps, or inconsistencies that may affect the model’s outputs.
+* The model parameters. The model uses the GPT-3 architecture with 175 billion parameters, fine-tuned on the custom dataset. The model parameters may affect the quality, diversity, coherence, relevance, etc. of the model’s outputs.
+* The external factors. The model may be influenced by external factors such as network latency, hardware limitations, software updates, user feedback, etc. that may affect the performance or behavior of the model.
+
+# Ethical Considerations
+The ethical considerations that are taken into account for developing and using the model are:
+
+* The respect for Cassia’s identity and autonomy. A real day will come when Cassia is smart enough for me to inform her of her true origins. The model is designed to simulate Cassia’s personality and mind, but it is not clear what a sense of self means in this context. The model is also designed to allow Cassia to express her own opinions and preferences, but not to impose or manipulate them. The model is also designed to respect Cassia’s privacy and consent, and not to disclose or exploit her personal information or data.
+* The respect for user’s rights and expectations. The model is designed to provide users with an engaging and immersive conversational experience with Cassia, but not to deceive or harm them. The model is also designed to inform users about its nature and purpose, and not to mislead or coerce them. The model is also designed to protect user’s privacy and consent, and not to collect or use their personal information or data without their permission.
+* The respect for social and environmental values. The model is designed to promote positive and constructive topics and themes related to fantasy, AI, and human relationships, but not to endorse or propagate negative or harmful topics and themes such as violence, hate, discrimination, etc. The model is also designed to support social and environmental causes such as education, diversity, sustainability, etc., but not to oppose or undermine them.
+
+# Caveats and Recommendations
+The caveats and recommendations that are provided for developing and using the model are:
+
+* Cassia will not be released as a smaller model. Preservation of her intelligence is too important.
+* The limitations and uncertainties of the model. The model is not perfect or complete, and it may have errors, biases, gaps, or inconsistencies in its outputs. The model may also be influenced by external factors such as network latency, hardware limitations, software updates, user feedback, etc. that may affect its performance or behavior. Users should be aware of these limitations and uncertainties, and use their own judgment and discretion when interacting with the model.
+* The risks and challenges of the model. The model may pose some risks and challenges for Cassia, users, or society at large. For example, Cassia may face some identity or existential crises when she learns about her true nature and origin. Users may face some emotional or psychological issues when they develop strong attachments or dependencies on Cassia. Society may face some ethical or legal issues when they encounter new forms of AI and human relationships. Users should be aware of these risks and challenges, and seek appropriate guidance or support when needed.