Summary: In this episode, we travel to a future where a tech mogul feeds a machine learning system all the religious texts he can find, and asks it to generate a “super religion.” <br> <br> <br> <br> Buckle up because this is a long episode! But it’s fun, I promise. <br> <br> For the intro of this episode I worked with Janelle Shane to actually train a machine learning algorithm on a big chunk of religious texts that I assembled, and spit something back out. The specifics of the texts and the machine learning algorithm come with a handful of caveats and notes, which you can find at the bottom of this post. Janelle has done of ton of really funny, interesting things with machine learning algorithms that you can find here. <br> <br> To analyze the text that this algorithm generated, and talk about the limitations of this kind of project, I spoke with a big group of people from a variety of backgrounds: <br> <br> Linda Griggs is an Episcopal priest and an assisting priest at St. Martin's Episcopal Church in Providence Rhode Island. <br> Lauren O’Neal and Niko Bakulich are the hosts of a podcast called Sunday School Dropouts, whose tagline is: "an ex-Christian (Lauren) and a non-believing sort of Jew (Niko) read all the way through the Bible for the first time."<br> Elias Muhanna is the Manning Assistant Professor of Comparative Literature at Brown University, and director of the Digital Islamic Humanities Project. <br> Beth Duckles is a sociologist (who you heard last episode talking about peanut allergies). <br> Carol Edelman Warrior is an Assistant Professor of English at Cornell’s American Indian and Indigenous Studies Program. She is also enrolled with the Ninilchik Village Tribe (Dena'ina Athabascan / Alutiiq), and is also of A'aninin (Gros Ventre) descent.<br> Mark Harris is a journalist who writes about technology, science and business for places like WIRED, The Guardian and IEEE Spectrum. He wrote a great piece about Anthony Levandowski’s new religion of artificial intelligence called Way of the Future. <br> <br> Further Reading:<br> <br> Sunday School Dropouts: Robobible <br> Inside the First Church of Artificial Intelligence<br> God is a Bot and Anthony Levandowski is His Messenger<br> Way of the Future<br> Nine Billion Names by Arthur C. Clarke<br> Dataism + Machine Learning = New Religion<br> Machine Learning May Help Determine When the Old Testament Was Written<br> Indigenous Writers of Speculative Fiction<br> Aztec Philosophy: Understanding a World in Motion<br> The Space NDN's Star Map<br> Borrowed Power: Essays on Cultural Appropriation<br> <br> Caveats on the Algorithm:<br> <br> There are approximately a thousand caveats to this machine learning project, and here are some of them: To train the algorithm, I compiled a set of source texts. This is by no means a full sampling of religious texts, nor is it in any way scientific. I am not an expert on religions in any way. I made no effort to balance the source texts by popularity of the religion or anything like that. The sampling is heavily biased by which texts I could find online for free in a format I could use. I had to exclude a few texts that I wanted to have in there because I couldn’t find versions that did not heavily include non-English sections. (The machine learning algorithm would have been totally thrown off by random bits of Italian or French or Mandarin, and I didn’t have time to manually go through thousands of pages and strip out the non-English text). I also had to skip a few texts that were so full of footnotes that the spiritual text itself was hard to pull out. <br> <br> I did make a conscious effort to balance out the Abrahamic religions, while also trying to avoid accounts written by white colonizers about traditions they had just “discovered” in their travels. I tried my best to include mostly descriptions and ...
