We chat with Emily Bender and Alex Hanna — authors of AI Con: How to Fight Big Tech’s Hype and Create the Future We Want — and pierce the veil of hype by getting into how these systems actually work and, importantly, the work they cannot do despite claims by boosters and doomers alike. Think of datasets like ImageNet or LAION-5B as big vats of pink slime and LLMs like ChatGPT as “synthetic text extruding machines” that turn pink slime into nuggets of text. It’s easy to forget that these magical mystery machines are direct descendants of very unexciting things like “T9 word.” We end the episode by chatting about why we shouldn’t trust the hype about how AI is going to destroy (or revolutionize) the education sector. ••• The AI Con | Emily Bender and Alex Hanna https://thecon.ai/ ••• On the genealogy of machine learning datasets: A critical history of ImageNet https://journals.sagepub.com/doi/full/10.1177/20539517211035955 ••• Mystery AI Hype Theater 3000 https://www.dair-institute.org/maiht3k/ Standing Plugs: ••• Order Jathan’s new book: https://www.ucpress.edu/book/9780520398078/the-mechanic-and-the-luddite ••• Subscribe to Ed’s substack: https://substack.com/@thetechbubble ••• Subscribe to TMK on patreon for premium episodes: https://www.patreon.com/thismachinekills Hosted by Jathan Sadowski (bsky.app/profile/jathansadowski.com) and Edward Ongweso Jr. (www.x.com/bigblackjacobin). Production / Music by Jereme Brown (bsky.app/profile/jebr.bsky.social)
