NovelSpace
NovelSpace is a software as service (SaaS) startup merging deep learning models with web applications, generative artificial intelligence with creativity.
Our debut product, Elysium 2084, transports users into a vividly immersive dystopia with sentient AI entities, androids, nomads, and cyberpunks. Each decision made by the user has a ripple effect, influencing the narrative trajectory and character development within this captivating universe. You can choose your guide, as you tour this alternate reality and bring back a graphic novel of the experience with all the imagery and chat logs between you, the traveler and the guide.
Sign-up and login to experience the beta release of Elysium 2084.
We are also offering a newletter to share information about our latest SaaS web apps in development. Our focus is on machine learning models like stable diffusion, image diffusers/image diffusion models and large language models, (LLMs). Sign-up here to join our mailing and newsletter list. Please reach out with inqueries or just to introduce yourself. We are located in the East Bay Hills overlooking San Francisco.
Creativity
Philip's career path has weaved through technology companies, computer graphics imagery design studios and startups including Nvidia, Logitech, LightFieldLabs to arrive at the beginning of new paradigm in computing: "NovelSpace is a platform for generative ai. This is txt2img, txt2vid, prompt design systems, API design and build, machine learning meet creators, we make images, language, video, music... We make experinence in collaboration with artificial intelligence."
We have several projects in development including the soon to be released Elysium 2084 to be followed by Future Friends and LatentSpace, a web app for organizing images and other rich media. More information will be posted to this page.
Ensemble Agent System
The Ensemble Agent System (EAS) stands at the forefront of interactive storytelling, weaving together advanced technologies including Langchain agents, a comprehensive World Model, and the GPT-4 language model to create an adaptive and immersive narrative experience. The system transcends the boundaries of traditional storytelling, actively engaging users in a narrative that evolves with their input.
These agents form the backbone of the EAS, working collaboratively to craft a cohesive and vibrant narrative. Leveraging machine learning and natural language processing, they interpret and respond to user inputs in real-time, shaping a narrative journey that is both engaging and dynamic.
The EAS represents a paradigm shift in narrative roles, blurring the lines between readers and active participants within the story. The users, now contributors, influence the unfolding of the narrative, fostering a personalized and nuanced storytelling experience. This evolution sees a diminishing distinction between protagonists and antagonists, giving rise to a narrative that responds organically to user interactions, and sees the narrator evolve into a more responsive entity, facilitating a truly immersive storytelling journey.
Marking a significant stride in narrative technology, the EAS introduces innovative features that revolutionize interactive storytelling. One of its standout features is the role-playing element, allowing users to embody characters within the narrative, thus forming a deeper bond with the story and fostering a truly personalized experience.
The system also incorporates Pinecone vectorstore technology to enhance response generation from World Docs, further refining the interaction between users and the narrative.
Adding a visual dimension to the narrative, the system has the capability to generate images that depict the unfolding story, offering users a vibrant and visual storytelling experience that is both immersive and engaging.
In summary, the Ensemble Agent System heralds a new era in narrative technology, promising users a unique journey that adapts and evolves with each interaction, providing a rich, personalized, and interactive narrative experience.
World Model
The World Model, as implemented by NovelSpace, serves as an extensive knowledge base for the LLM that are the engines of our AI-Exp web apps including Elysium 2084. The World Model provides a comprehensive and detailed description of the simulated world in which the interactive narrative takes place. Its core components are text embeddings created from these detailed descriptions. In the case of the product 'Elysium 2084', the World Cloud contains a vast amount of information about this particular dystopian universe.
At a foundational level, the World Cloud is a vector store for embeddings. These embeddings are dense representations of text data which capture the semantic meaning and relationships of the original text in a mathematical format that can be easily processed by AI systems.
The World Cloud operates as an informational backdrop to the Ensemble Agent System, informing the AI agents with a deep understanding of the world in which they exist and interact. The AI characters, thanks to the World Cloud, are not only aware of user interactions but also deeply knowledgeable about their universe. This context awareness enables them to generate coherent and engaging narratives that align with the world's established parameters.
Currently, the World Cloud is populated using detailed descriptions of the world, but there are plans to expand it further with the help of the Ensemble Agent System. The idea is to have a dedicated cast of AI agents generate narratives independently, which can be parsed and processed to extract new world elements - like character descriptions, locations, businesses, and concepts - and incorporated into the World Cloud. This approach would allow the World Cloud to evolve dynamically while maintaining user privacy.
Therefore, the World Cloud, in essence, is a dynamic, expansive, and contextually rich database that serves as the bedrock of the world-building in NovelSpace, continually informing and evolving the narratives created by the Ensemble Agent System.