Podcast Club ONLINE: AI Fine-tuning, Anthropomorphizing & the Black Box

This week's podcast is Offline with Jon Favreau and guest Jonathan Zittrain, the George Bemis Professor of International Law at Harvard Law School, a Professor of Public Policy at Harvard's Kennedy School of Government, a professor of computer science at Harvard's School of Engineering and Applied Sciences, director of Harvard's Law School Library, and co-founder and director of Harvard’s Berkman Klein Center for Internet & Society. Zittrain is a member of the American Academy of Arts and Sciences and on the Board of Directors of the Electronic Frontier Foundation. He was on the Board of Advisors for Scientific American, a Trustee of the Internet Society, a Forum Fellow of the World Economic Forum (which named him a Young Global Leader), and the Distinguished Scholar-in-Residence at the FCC, where he chaired the Open Internet Advisory Committee.

PLEASE NOTE: We require that attendees listen to the podcast below before attending the meeting to optimize discussion.

https://crooked.com/podcast/ai-zittrain-chatgpt-llms-openai/
[Note: If you use this link, please play Episode 192 (E192), NOT the episode that's readily cued up]

or

https://podcasts.apple.com/us/podcast/offline-with-jon-favreau/id1610392666?i=1000714588578

(1 hr., 21 min.)

Optional reading: Here's a link to The Atlantic article, by Zittrain, mentioned in the podcast: https://www.theatlantic.com/technology/archive/2025/05/inside-the-ai-black-box/682853/

-------------

What's this podcast about?
This week's podcast delves into questions about what's behind the AI that we use today, including LLMs and chatbots. Now that so many of us are using this technology daily, often in casual chatbot conversations where we reveal things about ourselves without much thought other than to get the information or help we seek, what are we surrendering to the AI? What is it gleaning from us, and how does that affect what we get back? This podcast explores the fine-tuning of AI and sycophancy, and it queries what we want from AI and what we actually should want to want from it (at times, a wholly different thing). Other topics covered include the anthropomorphizing of AI, the black box nature of AI, existential-risk concerns, whether humans are getting more stupid with AI use, the use of AI by Reddit co-founderAlexis Ohanian to create a video of his mother (who died when he was a boy) hugging him, and much more.
Just for interest, the Ohanian video is here: https://x.com/alexisohanian/status/1936746275120328931

Zoom link: https://us02web.zoom.us/j/3682869318?pwd=SC80SzI1S3JkeW1pbGhzcTNVQWVmdz09. Password for event might be W5hDy7

What's a podcast club?
It's like a book club for podcasts. Each week we discuss an interesting and currently-relevant episode from podcasters like Ezra Klein, Michael Shermer, or Russ Roberts. We start off as a large group and spend most of the time talking in small break-out groups, which we remix 3 times over the course of the evening. Our conversations are casual and open. We ask that everyone speak respectfully at all times, and we encourage free and concise discussion relevant to the week's podcast topic. To do this, we try to adhere to Grice's Maxims: https://bit.ly/2p4uSQm

Ultimately, the idea is to help each other think a little differently, learn a thing or two and have some fun.