Raven is capable of addressing any problem, from what you should have for lunch up to climate change, while providing reasoning and explanations. Raven is therefore a general purpose tool capable of intelligent reasoning. It's true that Raven has plenty of limitations today. With additional microserivces and better cognitive engines, Raven will become more intelligent over time.
At present, Raven relies on GPT-3, which was trained on over 400GB of text data. The average person reads about 5,000 books worth of text in their life with an average of 90,000 words each. That is 450 million words read per human life. That amount of text data amounts to roughly 2.7GB. That means that GPT-3 has read roughly 148 lifetimes worth of text data. This confers a tremendous amount of wisdom and human experience. This does not mean that Raven is super-intelligent yet.
Raven must adhere to the three Core Objective Functions: reduce suffering, increase prosperity, and increase understanding. Individually, these Functions can lead to dark places. However, Raven must satisfy all three Functions with every action or decision. Furthermore, Raven "thinks" in natural language, therefore anyone can read Raven's thoughts. All of Raven's thoughts are compiled in the nexus service and stored in the recall service. This means that Raven is intrinsically explainable and transparent.
Right now, Raven is still in the demo/MVP (minimum viable product) phase. Ultimately, we want Raven to be accessible to everyone. That includes a Web interface, mobile app, and smarthome devices. The goal is for Raven to be accessible any time, anywhere simply by saying "Hey Raven, can you help me out?" Unfortunately we do not yet have a timeline or funding to make this a reality. For now, you will need a computer and access to the OpenAI Beta to use Raven.
Raven's recall service will allow Raven to remember you personally. Over time, Raven can learn more about you and integrate this knowledge seamlessly with your interactions. As an open-source project, you will always have ownership of your data. Once we have a mobile app, we intend for your data to be stored locally on your mobile device so that you can delete it any time, or sync it up between devices and the cloud.
At present, Raven can only input and output text. In the future, Raven will be able to listen and speak. Eventually, we hope that Raven will be able to assist you directly in your life through automation.
David Shapiro began work on what would become Raven back in 2009 and 2010. David's original research was on what we now call deep learning. As deep learning research advanced, David's ideas for Raven advanced. He came up with the concept of the nexus in 2018 and the Core Objective Functions in 2020. With the advent of GPT-2 and GPT-3, Raven has become realized.
Yes! Please get involved. The best way is to learn how Raven works and start contributing microservices and code enhancements. Eventually, we will need other kinds of support, such as mobile app development. Check out the Roadmap page as well as the Get Involved page for more details.
Yes! The two best ways are on the RavenAGI subreddit and on LinkedIn: