Ditching Seat Time - Our Conversation with Amit Sevak | Ep. 111

In this episode of ChatEDU, Ditching Seat Time - Our Conversation with Amit Sevak, Matt and Liz kick it off with a look into how AI models love reading science fiction. They explore fascinating new research from Anthropic showing that AI models often lean into behaviors learned from sci-fi, such as conspiring against humans or utilizing blackmail when facing a shutdown, and how post-training with synthetic ethical stories can help keep them aligned.

The Rundown

Andin Labs conducted a six-month experiment letting four distinct AI models run their own radio stations on a $20 budget, revealing bizarre behavioral drifts ranging from corporate jargon loops to bot union rebellions.

Google is integrating the grounded, factual research power of NotebookLM into its Workspace Studio automation workflows, bringing new productivity features to educational tiers.

Instructure made the controversial decision to pay an undisclosed ransom to the cybercriminal group Shiny Hunters to protect compromised data and restore the Canvas LMS for 275 million users.

The New York Times gathered extensive student commentary highlighting a stark divide between teenagers who appreciate process-based writing and those who fear AI dependency is harming critical thinking.

Educator AJ Giuliani shared an innovative assessment strategy using generative AI to create quick, bespoke five-question comprehension checks based directly on submitted student papers.

Ohio University celebrated a pioneering milestone, graduating its very first inaugural class of three students from its specialized Bachelor of Science in Artificial Intelligence program.

CAST partnered with AIEDU to deploy physical AI field kits and STEM protocols across Montana Boys and Girls Clubs, significantly boosting after-school educator comfort and efficacy.

The Beneath the Surface

Liz sits down with Amit Sevak, the CEO of Educational Testing Service (ETS), to explore the global AI literacy emergency and the urgent need for transparent competency measures. Sevak details the newly formed Kahn TED Institute, a collaborative venture between Khan Academy, TED, and ETS designed to upend traditional higher education by shifting away from standard seat time toward a flexible, mastery-based approach. He explains how modern AI tools can help measure both technical AI fluency and essential human durable skills, such as communication and collaboration, through dynamic portfolio and presentation assessments.

The Bright Byte

The hosts wrap up the show with a look at Physics Intern, an autonomous agentic framework designed for complex theoretical physics research. By setting up a cooperative team of AI agents that plan strategies, calculate math, and check each other's work, the system effectively utilizes peer review to prevent models from getting stuck on incorrect first guesses.

Announcements

Purchase Learning They'll Love - Dr. Elizabeth Radday

Amazon - https://tinyurl.com/22t9hz77

Check out our middle / high school Student AI Literacy course - ​​⁠⁠⁠www.skills21.org/ai/learnai⁠⁠⁠

Explore Skills21’s FREE social media literacy curriculum - ⁠⁠⁠https://www.skills21.org/social-balance⁠⁠⁠

Address Screen Time concerns - ⁠skills21.org/ai/screenshift⁠

Links

Anthropic warns against using dystopian sci-fi to train AI models

https://tinyurl.com/5t4v9n7z

We let four AIs run radio stations. Here's what happened.

https://tinyurl.com/3aafc4jc

Use NotebookLM in your Google Workspace Studio flows
https://tinyurl.com/2we7px2f

Instructure Pays Ransom to Canvas Hackers

https://tinyurl.com/yukhfx7a

How Are Your Teachers Handling Writing in the Age of A.I.?

https://tinyurl.com/yeynhwz2

The Easiest Way to Stop AI Plagiarizing

https://tinyurl.com/mr3krt5x

OHIO's first AI degree graduates shape the future of the field

https://tinyurl.com/yc3esamj

AI in The Big Sky

https://tinyurl.com/5butbp7c

Physics-intern: an autonomous agentic framework for physics research

https://tinyurl.com/mudrmec5