Yin & Young Podcast EP 76: Matt Maybeno - AI, Machine Learning, Existential Risks & Benefits

Hope everyone had an amazing Thanksgiving! What a better way to enter the holiday season than to discuss with my ol’ college friend Matthew Maybeno how artificial intelligence will take away all our jobs (or not). It was great to reconnect with Matt and to see how he’s moved from biology into the tech space and thrived. His conversation with Dan and I is a great primer in understanding these new technologies and how to make better use of them. Dan also raises good questions about hallucinations and privacy concerns. Matt and I also reminisce about our dorm days and using high speed internet for educational purposes : ). Enjoy! - James Highlights 00:00:06 Intro. Disclaimer: Large shifts in the field since recording earlier this year, but key principles remain the same. 00:02:08 Matt’s use cases for artificial intelligence (AI) and his preferred Large Language Models (LLMs). 00:08:40 AI hallucinations. 00:11:23 Zero-shot prompting (get it right the first try), LLMs: subjective experience and training. 00:16:34 Pluses and minuses to different types of prompting. 00:21:49 Using AI as a tool to be more productive. 00:26:39 Matt’s early interests, working in biology, overcoming frustrations with math, now in the tech space. 00:45:10 Early internet memories: DC++, dorm networks, IRC, and the suite mate with a terabyte of adult videos. 00:48:28 What is intelligence? Consciousness? 00:55:46 Human-like AIs, Sci-fi parallels: Ex Machina, Cylons. 01:03:25 Defining AI, machine learning (ML), natural language processing (NLP) in simple terms. 01:07:32 Explaining LLMs 01:15:02 How to survive in an AI-driven world. 01:33:26 What are weights in AI. Current risks in AI: the weights used by humans to create malicious AI, jailbreaks, security threats. Potential existential risks (à la Terminator, The Matrix, etc…). 01:43:44 Language Corner: Japanese: 進化（shinka）- evolution English (GenAlpha slang): glazed - favored, skibidi - weird, odd Taunting phrases: Japanese - そーそーみーみー (sō sō mī mī), にゃー (nyah) cat sound. English - nananabooboo (trademarked by Matt’s daughter) Spanish - Lero lero, candelero (sucks to be you) 01:48:28 Staying balanced: marathon experiences, running, indoor soccer, Kung Fu forms (horse bench). Closing remarks. Article/Movies/Show referenced: - Ted Chiang’s “Why A.I. Isn’t Going to Make Art”: https://www.newyorker.com/culture/the-weekend-essay/why-ai-isnt-going-to-make-art - Ex Machina: https://www.imdb.com/title/tt0470752/ - Her: https://www.imdb.com/title/tt1798709/ - The Matrix: https://www.imdb.com/title/tt0133093/ - Terminator: https://www.imdb.com/title/tt0088247/ - Battlestar Galactica: https://www.imdb.com/title/tt0407362/ Note about the transcript/subtitles: AI transcription (Adobe Premiere) was used, so there may be errors. Follow and connect with us on: YouTube: https://www.youtube.com/@yinyoungpodcast Instagram: https://www.instagram.com/yinyoungpodcast Facebook: https://www.facebook.com/yinyoungpodcast If you enjoyed this episode, please like and subscribe. It helps us continue doing this podcast and reach more people. Thank you! Inquiries: yinyoungpodcast[at]gmail.com