Wolfram Beta: like Wolfram Alpha, but worse.
A neural network attempts to do math.
It does not go well, but it's wildly entertaining.
Running on Tensorflow 2.0, deployed with Flask.
The current model is an LSTM encoder-decoder architecture.
We're currently in the middle of training transformers to do this!
See our GitHub repository for the most up-to-date progress.