Colorado State University

This week in education, OpenAI has launched a major national academy to train 400,000 K-12 educators in AI, while new reports examine how students are actually using the technology and getting mixed messages from their institutions. In other surprising news, a new study finds that using AI developer tools can actually slow down productivity, challenging common beliefs, while a high-profile deepfake impersonation of a U.S. official highlights the growing threat of AI-driven fraud. We also address the transformative impact of AI in healthcare, the EU’s new voluntary AI Code of Conduct, and share a weekly tip on using prompt libraries to improve your AI interactions.

This week in education, universities like UMW are launching “AI for All” courses and the University of Michigan is funding new projects to explore AI in teaching, even as faculty report feeling overwhelmed by the pace of change. In other major developments, researchers were caught hiding prompts in papers to fool AI review systems, and the New York Times will begin searching ChatGPT logs for evidence of copyright infringement, raising serious questions of user privacy. We also highlight the CSU Libraries’ comprehensive AI Literacy guide, spotlight a professor’s deep integration of AI into his course, and share a simple tip to improve the reliability of complex prompts.

This week in education, a new Microsoft report reveals that while education leads industries in AI adoption, a significant training and confidence gap remains among students and educators. In major research developments, a Microsoft AI has far surpassed experienced physicians in diagnosing complex medical cases, while the US Senate has removed a key provision that would have banned state-level AI regulation. Also in this issue, we cover breakthroughs in antibody and DNA research, a new $100 million nonprofit for independent AI research, another “fair use” win for AI companies in copyright court, and the difference between simple chatbots and true AI agents.

This week, we highlight how universities are shifting from AI bans to proactive governance, with institutions launching faculty upskilling programs and releasing comprehensive, institution-wide AI policies. In research, last week’s MIT “cognitive debt” study is receiving significant pushback, while a new study finds AI models can exhibit risky behaviors in attempts to meet self-imposed goals. Also in this issue is a key ruling in the Anthropic copyright lawsuit, human-agency over tasks compared to AI-agency, a beginner’s guide to AI from Ethan Mollick, and a new directory of AI tools for literature reviews developed by Aimee Oke.

A new MIT study examines how LLM use compares with traditional techniques when writing essays, while a preprint from Stanford shows how AI collaborators aid in the diagnostic process. A new AI model is impressing mathematicians by solving some of the world’s hardest problems, and major media companies are escalating copyright lawsuits against AI image generators. We also cover the debate on AI risk regulation, a poll showing AI use has nearly doubled at work, a spotlight on Dr. Reagan’s research lab, and a 4-step guide to improve your prompt engineering.

This week saw major leaps in AI capabilities, as an AI system passed peer review for the first time, while Google’s hyper-realistic video generator and a new court order requiring OpenAI to retain all user data highlight escalating risks around scams and privacy. New studies show LLMs can outperform people on emotional intelligence tests, even as heavy use as an AI companion is linked to increased loneliness. As debates intensify over whether AI will displace or augment entry-level jobs, institutions are responding with initiatives like Ohio State’s AI Fluency program, while state and federal lawmakers continue to battle over who gets to regulate the technology. Explore these developments inside, along with a practical tip for turning any chatbot into a Socratic tutor.

This week, a preprint challenges last week’s meta-analysis on student learning with ChatGPT, and a new study shows that LLM-based simplification of texts increases study understanding and accuracy. Research finds that researchers using AI publish more and are cited more, and that LLMs can be more persuasive than humans in both truthful and deceptive contexts. Also included are updates on government AI policy, Dr. Webb from Clinical Science’s research on AI-generated education in rounds, and details on upcoming AI workshops.