Limits of agentic AI and error multiplication

In this episode of ChatEDU (Beyond the Vision: Building a Policy Infrastructure to Meet the AI Moment), Matt and Jonathan demo ChatGPT’s new voice mode, dig into AI labor-market news, tackle school chatbot safety, outline a four-point AI policy plan, and share a geothermal bright byte.

Story #1: The Iceberg Index and the Clickbait Problem

A new MIT study maps AI’s impact across 151M workers and 32K skills, but headlines distort the findings. Matt and Jonathan unpack what the research really shows, why the hype misleads, and what Microsoft’s agentic AI struggles and Google’s hard-drive wipe say about how far autonomous agents still have to go.

Story #2: Will My Job Survive AI? I Asked Gemini and ChatGPT

Matt uploads his job context to ChatGPT and Gemini to see how each predicts his future. ChatGPT delivers sharp, personalized coaching, while Gemini stays vague under stricter privacy rules. Jonathan argues that “evolving, weird” jobs may actually be the safest and that both tools can drive real insight when used strategically.

Story #3: AI Chatbots in Schools: A Practical Guide to Safety, Liability, and Mandated Reporting

A new EDSAFE AI Alliance guide asks whether school chatbots should act as mandated reporters. As tools like Character.ai and MagicSchool spur more student disclosures, the guide lays out a four-part framework to flag, notify, assess, and act. Matt and Jonathan break it down and argue that child-protection protocols must extend into virtual spaces.

Beneath the Surface: Jonathan’s Four Policy Shifts to Meet the Moment

Jonathan highlights four policy shifts for Connecticut:

1. A Shared Vision of the Graduate: Replace 169 local skill frameworks with one statewide model.

2. Lean Out the Curriculum: Use AI to refocus on power standards and durable skills.

3. Accountability Beyond the Basics: Measure the competencies that matter in an AI-driven world, not just literacy and math.

4. Policy Alignment with Practice: Bring state and local policies in reporting, AI-use rubrics, and more into sync with these instructional shifts.

Matt says this may be the most consequential chapter of Jonathan’s career and possibly for education as a whole.

Bright Byte: AI Uncovers Hidden Geothermal Energy

Zanskar Geothermal used AI modeling to uncover a viable energy system in western Nevada, a site long thought tapped out. It’s the first breakthrough in 30 years and a clear reminder of AI’s power to tackle tough environmental challenges when used well.

Links and References

ChatGPT’s voice mode no longer separate interface

https://tinyurl.com/4m3dtmcu

‘6-7’ viral phrase may have a meaning

https://tinyurl.com/4m3dtmcu

Measuring Skills-Centered Exposure in AI Economy

https://tinyurl.com/2w8p6es7

AI Agents Disaster

https://tinyurl.com/2s3u3tb5

Google’s AI Deletes User’s Hard Drive

https://tinyurl.com/3n284eht

AI on Jobs in Five Years

https://tinyurl.com/42vtyvus

AI Chatbots in Schools

https://tinyurl.com/yvntwfr6

Professors using AI in course design

https://tinyurl.com/yevrmbet

Dropout Hired at OpenAI After Learning PhD-Level AI with ChatGPT

https://tinyurl.com/ym7tt62t

Poll: Americans Doubt Degree Value

https://tinyurl.com/3cvccufr

NewCollege Major: A.I.

https://tinyurl.com/35d32zhu

Geothermal company makes big discovery using AI

https://tinyurl.com/3fenuej7

Announcements

The Winter Micro-Credential cohort launches this January. Join educators from around the world for a six-week dive into AI literacy, pedagogy, assessment, and ethics.

EdAdvance is now offering a Middle and High School Student AI Literacy course — email chatedu@edadvance.org to bring it to your district.

Interested in Skills21’s free social media literacy course? Check it out here - https://www.skills21.org/social-balance

Sponsor

This episode is brought to you by the National Center for Next Generation Manufacturing, helping educators and students build workforce-ready skills through innovation. Learn more at nextgenmfg.org.