“
I received access to this pretty entertaining ML tool and wanted to share it here. It’s still in the early stages so any data you want to input is good data!
Username: BSTester
Password: WhatsThatSmell2020
It’s all for fun so take the results with a grain of salt. Enjoy and share any findings you think are interesting!
“The Bullshit Detector’s architecture is a Recurrent Neural Network with Gated Recurrence Units (GRUs). Additionally, in order to mitigate bias and boost generalizability, we employed a semi-supervised training approach. Model A employs straightforward supervised learning on ~100k texts, and model B employs semi-supervised learning on about ~40k texts. We advise experimenting with both models A and B. This will help us literally A/B test the two models. Due to the black-box nature of neural networks, we are unable to perfectly digest the model’s classification into humanly understandable rules or criteria. But when we have developed an AI that can interview the AI behind Bullshit Detector, we will let you know and post the conversation.”
”
– Foreign-Smoke
