We were inspired by our interest in NLP and by carnival games that tell your fate.
What it does
Our application tells you a prediction about the future generated by a natural language-producing AI that was trained on sample predictions.
How we built it
We used GPT-2 to build our model, trained on fortunes we wrote. We used Flask to adapt it to a web program with an original drawing.
Challenges we ran into
We at first had trouble setting up the environment for GPT-2. We also ran into trouble with manipulating our dataset of predictions into the right format to feed the model.
Accomplishments that we're proud of
We are really proud that we started with no knowledge of GPT-2 and are leaving with a finished, working program!
What we learned
We learned about natural language processing using GPT-2, working with trained models, and web development using Flask.
What's next for Magical Predictions
Adding more styling to the interface and decreasing run time.