2 AIs & A Human

• 

  A flowchart we made for the frontend side of development!

• 

  A whiteboard summary of the project:

Inspiration

This is essentially the software version of a game one of us used to play manually with AI chatbots. In a different sense, it is interesting to, in an age where AIs begin to proliferate more and more throughout society, to examine how our language and by extension the way we interact in the world is impacted by AI. Also, it's a fun game.

What it does

2 AIs & A Human is a game where players attempt to trick each other by mimicking generative AI responses to questions. Each player gets 3 questions, 2 of which are already answered by AI, and one of which that they must answer, trying their best to mimic the 2 AI responses already generated. These questions and answers are then sent to their opponent, who has to guess which answer was written by a human. At the same time, the player has to guess which of their opponent's answers was written by them.

How we built it

We built the frontend by interfacing with the backend API using python's requests module and rendered the game itself using the tkinter GUI library.

Challenges we ran into

One of the first issues was the fact that the two of us building the frontend side of the project had never built anything remotely related. We picked up the tkinter library fairly quickly and for the most part learned the intricacies of how it worked as we developed the frontend, winding up with a passable frontend that, while not wowing anybody besides our grandmas, worked, despite challenges.

An issue we had later in the project was integrating the two sides of the project. We ran into bugs as well as some miscommunication between our two subteams. After reconciling these initial misunderstandings we had some issues with waiting for player input which we resolved by introducing Python's threading library.

Accomplishments that we're proud of

For all of us being at our first software hackathon, to get a project to the point where we were happy with it was in itself an achievement. Was the code messy, hard to read, annoying to debug, overall finnicky? Yes. Does it work? Mostly!

On another note, essentially none of our code was written by AI (we had an AI generate most of the questions in the JSON file)! Ironic for a program that depends on generative AI to work, but nonetheless very cool!

Finally, for James & Wyatt, this was their first time using Tkinter, so successfully making a frontend is no small thing to scoff at. For Connor, this was his first time coding in Golang, making API calls, and using Postgres database, so this was valuable experience for him.

What we learned

First and foremost, we realized the importance of documentation proactively rather than reactively. Humans are, generally speaking, unable to run multiple threads at the same time which resulted in significant blocking in API (Anthropomorphic Person of Interest) requests. With proper documentation we could have cut down how long it took for us to integrate the frontend and backend.

More concretely, we learned how to make GUIs with tkinter. We also learned how to work with APIs using Go and gin.

What's next for 2 AIs & A Human

We all hope to get much needed rest and eventually forget this code was ever written.

Built With

• genai
• gin
• go
• google
• json
• postgresql
• python
• random
• requests
• sql
• threading
• tkinter