AI Commons Bulletin 2/19/2025

🧠 AI Tools Soon to Decide How Much They Need to “Think”
Expect the answers from AI tools to generally improve over the next few months, as more of them incorporate “reasoning” into their process. These are models that can discern when a prompt is more complex and would require a multi-step reasoning process. OpenAI is starting this with ChatGPT soon.

Learn More: https://www.youtube.com/watch?v=KtwK3hBAjDY

📗 Five Generations of Intelligent Textbooks
Sosnovsky & Brusilovsky compile the literature on intelligent textbooks and organize five generations:
Engineered: AI-powered adaptive reading.
Integrated: Linked with external smart content.
Extracted: AI analyzes and structures knowledge.
Datamined: Tracks student engagement for insights.
Generated: AI creates content, questions, & chatbots
Learn More: Sosnovsky, S., Brusilovsky, P. & Lan, A. Intelligent Textbooks. Int J Artif Intell Educ (2025).

🚫 Guidance for Uses of AI Banned by EU’s AI Act
The EU regulates AI much more than the US does. When it adopted the AI Act, it banned “unacceptable risk” uses, but didn’t provide much explanation. A new report lays out examples, including manipulative, deceptive, and exploitative practices.

Learn More: https://ec.europa.eu/newsroom/dae/redirection/document/112367

⏳ Waiting 5-10 Minutes for an AI to Answer?! What?!
Deep Research is a newer function of Google’s AI, Gemini. You can ask it an extended question and it will break it down into parts, research each part (including multiple web searches), and write up a report you can download. It’s available both on the web and on Android. Additional $ required.

Learn More: https://youtu.be/IBKRyI5m_Rk

Bulletin items compiled by MJ Jackson and Sarah Freye with production assistance from Lisa Batchelder. Get the AI-Commons Bulletin on our Microsoft Teams channel, at aicommons.commons.msu.edu, or by email (send an email to aicommons@msu.edu with the word “subscribe”).