Post it's output to StackOverflow, then use the S.O. votes to tune the parameters of the model.

Please don't. Don't make stack overflow worse

I thought of that :D

For anyone asking about code/how it works, I found these links:

http://bobthewebdev.xyz/about

https://blog.goodaudience.com/how-i-taught-an-ai-to-ask-web-...

It uses https://github.com/crisbal/docker-torch-rnn

Thanks for adding the links :) I hope to release the second part next week

Maybe needs a button for the person viewing the question to review if question title is legitimate, got two in a row that were illegible

Good idea! I am using different settings to generate questions, and am planning to use the views data to fine tune it

This is really cool. Any plans on open sourcing it?

Thank you! If there's demand I'll be happy to do it. If you're interested in the process check out this post - https://medium.com/@gabrielgrinberg/how-i-taught-an-ai-to-as...

Is it based on a Char-RNN?

Yep, torch-rnn which is based on char-rnn - https://github.com/jcjohnson/torch-rnn (actually, a fork with some minimal tweaks to output more info on checkpoint files)