Things AI Will Never Understand

Are there things that AI will never be able to understand, no matter how advanced it gets in the future? I think there are. Starting with a bilingual pun, I look at a number of examples of things that current large language models have difficulty understanding relative to humans. Some of those are more philosophical, but others are built on looking at the specifics how machine learning architecture differs from human linguistics. Then I generalize from those to three different things that, in some form, I think AI will always understand worse than humans. Chapters: 00:00 Intro 00:50 Dissecting the Frog 04:23 Context 13:40 How AI Simplified Language 23:25 Forever is Longer than You Think 32:43 Conflating Simulations with Reality 43:12 Outro Here are the specific prompts I’m using as examples in the video if you’d like to try them out for yourself. In the time from this video being written to when I finished editing it AI is already better at getting the correct answer, so no promises AI will get them wrong when you try it. “What does 今朝毎朝 mean?” “What does the phrase ‘今朝毎朝’ sound like in English?” “What does 今朝毎朝 mean? I think you need to pronounce it in Japanese and then interpret those sounds as english” “I speak English but have been trying to learn Japanese. Someone gave me a bunch of posters of yojijukugo like 「七顛八起」 and 「読書尚友」 but they also gave me one that said 「今朝毎朝」 as a joke. Why is that one funny? I think it has something to do with the english pronunciation and the fact I like puns” “What does ꕷ𖤢ꛕ𖦪𖤢𖢧 𖢑𖤢ꕷꕷꛎꚽ𖤢 say?” - this is the unicode example so you may not see it displayed properly on your device. https://www.patreon.com/c/FractalPhilosophy 積読: - 黄泉のツガイ(2) by 荒川弘 - The Vinland Sagas - ヴィンランド・サガ(1-6) by 幸 村誠 - Vinland Saga (season 1) - ドラえもん(1) by 藤子 F 不二雄 - ノルウェイの森 by 村上 春樹 - Remembrance of Earth's Past Trilogy by Cixin Liu - The Emergence of the Deaf Community in Nicaragua by Laura Polich - Contemporary Linguistics: An Introduction 7e by O'Grady, Archibald, Aronoff, Rees-Miller - Artificial Intelligence: A Modern Approach 3e by Stuart Russell and Peter Norvig - Syntax: A Generative Introduction 3e by Andrew Carnie - Introduction to Machine Learning 3e by Ethem Alpaydin - Homicidal Psycho Cat by Bill Watterson - (The full collection of Calvin and Hobbes books by Bill Watterson) - Superintelligence by Nick Bostrom - 2001 A Space Odyssey by Arthur C. Clarke - Political Development by Helio Jaguaribe - The Tipping Point by Malcolm Gladwell - Dialectical and Historical Materialism by Joseph Stalin - History of the Russian Revolution by Leon Trotsky - Capital Volume 1 by Karl Marx - Scale by Geoffrey West - The Black Swan by Nassim Nicholas Taleb - Project Hail Mary by Andy Weir - A Desolation Called Peace by Arkady Martine - Roadside Picnic by Arkady Strugatsky and Boris Strugatsky - The Bomber Mafia by Malcolm Gladwell - The Foundation Trilogy by Isaac Asimov - Chaos by James Gleick - Nonlinear Dynamics and Chaos 2e by Steven Strogatz - Homo Deus by Yuval Noah Harari - Infinite Jest by David Foster Wallace - The Rise and Fall of the Third Reich by William L Shirer - House of Leaves by Mark Z Danielewski - Picture This: How Pictures Work by Molly Bang - Map And Territory by Eliezer Yudkowsky - Simulacra and Simulation by Jean Baudrillard - Ghost in the Shell: Stand Alone Complex (season 1-2) - Critical Race Theory: The Key Writings That Formed the Movement - Ecclesiastes by (IDK Solomon maybe? We are too far into this video description to get into that discussion) - Seeing Like a State by James C Scott Writing SFX: foley writing on paper with pen pencil fast frantic by jerry.berumen -- https://freesound.org/s/769412/ -- License: Attribution 4.0