Coming into this hackathon, we wanted to explore a project that used machine learning technology. We first thought of a Vaporwave art aestheticizer that would take photos/videos/audio and apply styling to make them more "ＡＥＳＴＨＥＴＩＣ". We instead decided on something more lowkey -- a game that is based around the Turing test.
What it does
The concept of the game involves multiple players online who connect one-on-one with one another, or one-on-bot.
How we built it
We used Wolfram's API to manage machine learning, React to write the front end of the site, and Google's App Engine to host/deploy the application.
Challenges we ran into
Learning a the new Wolfram language was especially challenging, given its many divergences from the usual programming constructs (fully symbolic expressions, multiple operators for mapping and application, etc). Another challenge was trying to figure out how to harness the power of the machine learning frameworks we used. Without any experience, it was hard to figure out why our models didn't predict data as well as they should have (e.g. given training set data after training produced different results than those during training).
Accomplishments that we're proud of
The application is nearly functional! Though in some cases the bot will reply in stuttering nonsense, we all had a few laughs at what the bot can produce (it turns out that it is very sassy).
We want to continue expanding the capabilities of the bot and this project. Given more experience in ML and more time/money to spend on keeping cloud servers running, we can definitely see this as a future project to follow through on together.