m4m4_b0Iz

Inspiration

We were inspired by extrapolating the trolley problem to self-driving cars. As self-driving cars are the future, what are the moral dilemmas of the deaths that surround them?

What it does

It is a web app that generates characters, their traits and avatars, on the fly. Players must make the difficult decision of saving a limited number of these characters. Furthermore, a neural network is training on the data of those who are saved, learning to playing the game like a human.

How we built it

We used python for the backend and for the neural network. We made the app itself in typescript. This is all ran on google cloud.

Challenges we ran into

This is the first time developing a dynamic website for most of us. Furthermore, we are all not used to working as a team. Consequently, coordinating our efforts and code was more difficult than anticipated.
We had to generate data and then parse it for the web app, then we'd collect the game data and parse it for a neural network.

As an example of our struggles, we had to rewrite code meant for the backend and move it to the frontend at three in the morning.

Accomplishments that we're proud of

The game looks polished and the neural network can learn from users playing. These were key goals we had set for ourselves. We wanted an AI to learn human morality, but also have a game that looks nice.

What we learned

We learned how to work together as a team and utilize multiple languages to patch our app together.

What's next for m4m4_b0Iz

Could the neural network use what it learns to generate harder moral dilemmas? How could we visualize the weights of the feature contributions? These are questions we would like to explore further.

Built With

• domain.com
• google-cloud
• javascript
• python
• svelte
• typescript