Our AI writing assistant, WriteUp, can assist you in easily writing any text. Click here to experience its capabilities.

The First Month of NovelAI

View Original View Raw

Summary

NovelAI is a text-generation AI powered by GPT-Neo model Calliope, trained on quality writing and novels. It has 2048 token memory, allowing it to remember events from further ago than AI Dungeon's 2800 character limit. NovelAI allows users to easily export and import stories, as well as adjust settings and customize themes. The closed alpha had 100 users and processed 40000+ actions in 3 days. The open beta will be released soon with pricing and feature roadmaps. Join NovelAI's Discord, Twitter, and Reddit for updates.

Q&As

What features does NovelAI offer?
NovelAI offers AI powered by the GPT-Neo model, full capacity of the AI’s memory, access to settings, easy to start and edit, easy export and import, memory and author’s note, customizable themes, encryption and storage.

How does NovelAI store and encrypt stories?
NovelAI stores stories in the browser by default and users may choose to store their story in encrypted form on the servers. Stories are locally encrypted with AES-256 before being sent to the servers for storage.

How many users participated in the closed alpha?
100 users participated in the closed alpha.

What datasets and training hyperparameters were used to create the AI model?
Different datasets and different training hyperparameters were used to create the AI model.

What will be released this week in terms of pricing and release date for the open beta?
Pricing and release date of the open beta will be released this week.

AI Comments

👍 This article provides an in-depth look into the features and technical aspects of NovelAI. It is great to see the team has put so much effort into providing users with options for customizing the AI model and encrypting their stories.

👎 The article does not provide much information on the pricing of the open beta and how to access it. This information would be useful for potential users.

AI Discussion

Me: It's about the first month of a new AI technology, NovelAI. They discuss the features of the technology, how it works, how it can be used, and what's coming next.

Friend: That sounds really interesting! What implications does the article have?

Me: The article implies that AI technologies are becoming more advanced and can help us in creating quality stories. It also implies that AI technology is becoming more accessible and easy to use, as it can be tailored to users' specific needs and stories can be exported and shared easily. Finally, the article implies that encryption and storage are becoming more secure, as stories are stored in the browser by default and can be encrypted before being sent to servers for storage.

Action items

Join the NovelAI Discord, Twitter, and Reddit to stay up to date on the open beta launch.
Try out the features of NovelAI and provide feedback to the team.
Research the GPT-Neo model and other AI models to better understand the capabilities of NovelAI.

Technical terms

GPT-Neo: Generative Pre-trained Transformer-Neo is a natural language processing model developed by OpenAI. It is a large transformer-based language model trained on a large corpus of text data.
Finetuned: Finetuning is the process of fine-tuning a pre-trained model on a new dataset. This is done by adjusting the weights of the model to better fit the new data.
Hyperparameters: Hyperparameters are the parameters of a machine learning model that are set before training. They are used to control the behavior of the model and can have a significant impact on the performance of the model.
AES-256: Advanced Encryption Standard (AES) is a symmetric encryption algorithm used to protect data. AES-256 is a variant of AES with a 256-bit key size.
GPU: A Graphics Processing Unit (GPU) is a specialized electronic circuit designed to rapidly manipulate and alter memory to accelerate the creation of images in a frame buffer intended for output to a display.
Perplexity: Perplexity is a measure of how well a probability model predicts a sample. It is calculated by taking the exponential of the average log-likelihood of the sample.