The Mad Libs Report & Robo Mueller

[knowtheory]

I HEREBY INVOKE MY RIGHTS...

As described in section 11.11.B of the BIFFUD Corporate Bylaws I hereby invoke
my rights as a sentient being who has not uploaded their mind to the cloud for consideration of this project application by the BIFFUD Hive Mind. With this
application I submit my interest in becoming a Member of BIFFUD and having this
project supported and adored by all who can 🤔.

Project Information

• Project Name: The Mad Libs Report & Robo Mueller
• Project Haiku:

What has Barr blacked out?
Either fill mad libs in, or
Ask Robo Mueller

• Project Analogy: ok ok, it's like machine learning, but unredacting the mueller report.

Project Description

Imagine if you could take a redacted page from any document, and turn the redactions into Mad Libs.

NOW IMAGINE if you could make a computer guess what to fill into those Mad Libs!

Okay, now do that to the whole Mueller Report.

Bylaw Questions

How is this project a bad idea?

Don't take my word for it. Here's a professor of linguistics at University of Washington's take:

I've seen several different #NLProc folks suggesting today that it would fun/interesting/worthwhile to use BERT or GPT-2 to fill in the redacted bits of the Mueller report. A short thread on why this is a terrible idea /1

— Emily M. Bender (@emilymbender) April 19, 2019

(Note for Professor Bender: I'm really sorry i accidentally name checked you here, i didn't think through the fact that pasting the tweet embed in here might happen to line up with your Github profile! Also, thank you for writing your tweet thread so i could point other people at it).

If this project were a D&D Character, what alignment would it be and why?

Oh this is 💯% a Chaotic Neutral project. There's a lot of public interest in the Mueller Report and its contents. This project does absolutely nothing to help further the public's understanding of that report. In fact, it obfuscates understanding by page layout analysis and machine learning.

Where are the lulz?

This project is 40% lulz, 80% machine learning, and 64% commentary on politics, language, and technology.

How does this project make people thinking face emoji?

Machine Learning isn't magic. But we can make it magical. And make some points about the fact that it can't read minds.

Who is involved?

Name: Ted Han
Twitter: @knowtheory
Github ID: @knowtheory
Skillz: trolling, machain lurning, PDF sorcery
Project Role / expectations: Instigator, abandoner of project
Project stake: 🥩

Name: Jeremy Merrill
Twitter: @jeremybmerrill
Github ID: @jeremybmerrill
Skillz: Machine Learning Journalism
Project role / expectations: co-conspirator until bored
Project stake: 🥩

Name: Mike Tigas
Twitter: @mtigas
Github ID: @mtigas
Skillz: See this tweet. And then this one.
Project role / expectations: More computers
Project stake: 🥩

Who will be the project's Comptroller?

It's ted.

Is this realistic to implement via BIFFUD?

QAnon says yes.

Next Steps

1. Attend the next scheduled BIFFUD plotting session to plead your case.
2. ok:
   1. Get the Mueller Report and all of the documents that the Special Council's Office has filed
   2. Learn some machines with the the SCO's documents.
   3. ???†
   4. PROFIT

†: (use PDF.js to rasterize pages & and ... something else ... to find the black boxes in the image. Ted will ask @mtigas for advice)

How (often) will you be providing updates to the organization?

This idea has a shelf life and should probably happen sooner rather than later. I guess Ted will promise weekly updates?

[slifty]

This is great! We discussed during today's plotting session and one point of feedback was that doubling down on the more generic concept of turning redacted documents into mad libs is where the lulz truly lie (also takes pressure off the timing of Mueller specifically)

So in summary:

1. The product is the "redacted document -> mad libs" (specific documents are more of a demo)
2. Focusing on the silly aspect of mad libs rather than training models using "real" looking data is more likely to be funny AND not accidentally create misinformation!