GenAI Studio: News, Tools, and Teaching & Learning FAQs

This week

News of the week

Tool Showcase

FAQs

This week in AI highlighted how students are learning with AI, as well as advances in chatbot privacy, research, and reliability. Students at the University of Leeds continued building Learning with AI, an open collection of reflections on how generative tools are reshaping study habits. Anthropic pushed a new Economic Index showing how its Claude model is being used across countries and industries, revealing differences in reliance on automation-heavy tasks versus collaboration-driven applications. In the research space, Google introduced VaultGemma, a privacy-preserving language model trained with differential privacy, and Math Inc. unveiled Gauss, an AI system capable of formalizing complex mathematical proofs at unprecedented speed. Meanwhile, Simon Willison examined Claude’s transparent memory system, and Thinking Machines outlined methods to make AI outputs more reliable by tackling non-determinism in large language models.

Here’s this week’s news:

[This week’s recording will be placed here when it becomes available.]

Learning with AI: A Student-Led Project Exploring Student’s Use of Generative AI

With support from the University of Leeds, the Learning with AI project brings together student reflections on how generative AI is shaping learning. The collection is open and continuously updated, featuring first-hand accounts of how students use AI tools in their studies, what challenges they face, and the ethical questions that arise. It serves as a living record of how AI is changing student’s experience in education.

Read the Full Page Here

---

Learning with AI: Using Generative AI to Understand Concepts

In this student reflection, undergraduate Oliver Fletcher shares how he uses AI to better grasp challenging topics in economics and law. Rather than turning to AI to learn new materials from scratch, he uses it when he needs clarification on difficult concepts or faces roadblocks in his learning. His story highlights the value of using AI as a supportive tool to deepen understanding, not as a replacement for traditional learning methods.

Read the Full Page Here

---

Anthropic Economic Index: AI’s Growing Economic Footprint

Anthropic has released a new index showing how people use its Claude AI around the world. The data shows that countries with higher average income are more likely to use Claude, tend to use it for collaboration than automation, and apply it to a wider range of tasks beyond coding. Across all users, businesses are more likely to use Claude to automate processes, while individuals users typical use it for guided collaboration.

Read the Full Article Here

---

Simson Willison’s Weblog: Claude’s Memory Explained

Simon Willison examines how Anthropic’s Claude AI handles memory differently from other chatbots. Instead of automatically storing information, Claude starts each conversation fresh and only saves details when memory is explicitly enabled. Users can also review what the AI remembers, making the process more transparent and easier to control.

Read the Full Blog Post Here

---

VaultGemma: Google’s New Privacy-Protecting AI

Google has introduced VaultGemma a large language model (LLM) designed to protect user privacy. It is trained using a method called “differential privacy,” which adds protective noise during training to safeguard personal data. While this increases training cost and complexity, it offers stronger privacy protection than traditional models. On academic benchmarks, VaultGemma performs at a similar level to non-private models from about 5 years ago, showing both progress and the gap still left to close.

Read the Full Article Here

---

Gauss: A New AI Model Helping Mathematicians Tackle Complex Proofs

Math Inc. has unveiled Gauss, an AI model that helps mathematicians tackle complex proofs. It recently formalized a major mathematical challenge in record time, producing thousands of lines of verified code. By accelerating work that might otherwise take years, Gauss demonstrates how AI could transform the pace of mathematical discovery.

Read the Full Article Here

---

Making Large Language Models (LLMs) More Reliable

Researchers at Thinking Machine studied why LLMs give slightly different answers even when asked the same question with identical settings — a problem known as non-determinism. They propose strategies to reduce this issue, making AI outputs more consistent and trustworthy. These improvements could make AI more dependable for critical tasks in science, healthcare, and other fields where consistency is essential.

Read the Full Article Here

---

---

Tool of The Week: The CREATE (Content, Retention, Engagement, Accessible, Teaching Presence, Evaluation) App

What is the CREATE App?

CREATE is an app currently being developed by UBC students, faculty and staff that uses generative AI to help instructors design interactive learning content. By analyzing uploaded course materials, it helps instructors create and refine learning objectives and transform them into flashcards, quizzes, or gamified questions.

How is it used?

Instructors simply upload their course materials to the app, then use CREATE’s AI features to generate and refine learning objectives, along with a variety of interactive formats aligned with those objectives. The outputs, such as flashcards, quizzes, or-game-style questions, can be exported and shared with students.

What is it used for?

CREATE supports instructors in developing engaging, interactive content for their courses that enhances student learning while reducing the time and effort required to create them.

Click Here to Learn About the Project Proposal

---