PartyRock University

Inspiration

I wanted to learn about neurobiology and did not have a clear plan. I thought that AI could help create a plan for me in a similar way to how universities provide course plans.

What it does

Creates a course based on a chat with the user. Helps the user learn more about the topic through a detailed 12-week plan.

How we built it

Organized different types of PartyRock widgets that depend on each other for input.

Challenges we ran into

Figuring out the optimal hyperparameters. Learning how the "input" fields affected the text generation and optimizing them. Optimizing the prompts.

Accomplishments that we're proud of

The fact that it creates a good course outline and guide for the student.

What we learned

Learned how to optimize prompts. For hyperparameters, I learned that CFG=10 was optimal for the pictures, and I learned that a higher P-value and temperature are optimal for chats (which need to be more creative), while lower P-value/temperature is optimal for components that depend on factual information (such as textbook recommendations), and a medium P-value/temperature is optimal for text generation which depends on the chat.

What's next for PartyRock University

Tuning the hyperparameters further, making sure that the course outline and weekly plans are aligned with each other. Labs and assignments should return appropriate text.

Built With

• partyrock