GenAI Studio: News, Tools, and Teaching & Learning FAQs
These sixty minute, weekly sessions – facilitated by Technologists and Pedagogy Experts from the CTLT – are designed for faculty and staff at UBC who are using, or thinking about using, Generative AI tools as part of their teaching, researching, or daily work. Each week we discuss the news of the week, highlight a specific tool for use within teaching and learning, and then hold a question and answer session for attendees.
They run on Zoom every Wednesday from 1pm – 2pm and you can register for upcoming events on the CTLT Events Website.
News of the Week
Each week we discuss several new items that happened in the Generative AI space over the past 7 days. There’s usually a flood of new AI-adjacent news every week – as this industry is moving so fast – so we highlight news articles which are relevant to the UBC community.
This week in AI, we saw notable developments in open-source datasets, model design, and critical discussions on the cognitive boundaries of language models. Hugging Face expanded its commitment to transparent training data with the Common Pile and released two massive COMMA models trained on 1T and 2T tokens respectively. Mistral introduced Magistral, a high-context model designed for scalable reasoning, while Apple’s research challenged assumptions about AI understanding in The Illusion of Thinking. OpenAI launched O3-Pro, enhancing performance and latency for advanced reasoning tasks. Meanwhile, Sam Altman outlined a vision for a gradual, humane AI transformation in The Gentle Singularity, and The American Vandal warned of AI’s encroachment on educational autonomy, urging a return to human-centered pedagogy. Finally, we explored the tool of the week Clara, an open-source local AI tool suite.
Here’s this week’s news:
The Common Pile and COMMA Models
The Common Pile is a curated dataset initiative developed by the Hugging Face team to offer a transparent, diverse, and openly accessible foundation for training large language models. It aims to improve data quality through meticulous documentation and ethical filtering, distinguishing itself from proprietary datasets. COMMA, the “Common Model of the Modern Age,” builds on this work with two large-scale models—COMMA v0.1-1T and COMMA v0.1-2T—trained on one and two trillion tokens respectively. These models are designed to be more representative, interpretable, and socially aware by virtue of their deliberate dataset composition.
Explore the goals and datasets behind the Common Pile and COMMA models.
Check out the Comma v0.1-1t model.
Check out the Comma v0.1-2t model.
Mistral Debuts ‘Magistral’: A High-Performance Model with Long Context
Mistral AI has introduced Magistral, a new model architecture tailored for complex reasoning tasks over extended contexts. The release reflects Mistral’s continued focus on producing open, competitive alternatives to closed-source frontier models. Magistral stands out for its ability to maintain coherence over longer inputs while improving latency and throughput. It represents Mistral’s next step toward scalable, efficient, and transparent language model design.
Read about Magistral’s capabilities and design.
Apple Research Questions LLM Understanding in ‘The Illusion of Thinking’
A new paper from Apple’s Machine Learning Research division, The Illusion of Thinking, challenges assumptions about reasoning in large language models. The authors argue that many LLMs exhibit superficial fluency and coherence without genuine understanding or deductive capability. Through a series of cognitive and logic tasks, the study reveals fundamental gaps in current AI systems’ interpretative depth. This work invites a reevaluation of what constitutes true machine “thinking” and how it should be measured.
Examine Apple’s analysis on LLM cognition (PDF).
OpenAI Introduces O3-Pro: Enhanced Reasoning with Faster Latency
OpenAI has released O3-Pro, an upgraded version of its O3 series, optimized for improved reasoning and reduced response time. The model introduces refined internal routing techniques and architectural changes that significantly boost performance on multi-step and long-context tasks. O3-Pro is designed to compete in both commercial deployment and benchmark evaluation, offering a more cost-efficient alternative to heavier models. It continues OpenAI’s efforts to scale capability without sacrificing latency or alignment.
Learn more about O3-Pro’s improvements and use cases.
Read Techcrunch’s O3-Pro overview.
The Gentle Singularity: Sam Altman’s Reflections on AI Futures
In The Gentle Singularity, Sam Altman explores the philosophical and societal implications of approaching superintelligent AI. He envisions a scenario where careful stewardship and open collaboration lead to a gradual, non-catastrophic intelligence explosion. Altman emphasizes alignment, transparency, and humanity-centric values as central pillars in this transition. His essay serves both as a personal vision and a call for shared responsibility among developers and policymakers.
Read more on Sam Altman’s humane AI future
Against Technofeudal Education: Reclaiming Critical Pedagogy in the AI Era
This essay from The American Vandal critiques how AI tools are being co-opted by institutional forces to erode teacher autonomy and standardize student output. It warns against the commodification of education into a platform-driven, metrics-obsessed system that mimics corporate logic. The piece calls for resisting this trend by reasserting critical pedagogy, emphasizing the role of educators in fostering genuine intellectual inquiry. It positions AI not as inherently harmful, but as a battleground for control over the purpose of education.
Read the critique of AI’s role in reshaping education.
Tool of The Week: Clara – Open-Sourced AI Tool Suite
What is Clara?
Clara is a local-first, privacy-focused AI assistant designed to operate entirely on your machine without requiring an internet connection. Built with a user-centric philosophy, Clara integrates various AI models into a unified interface, allowing users to manage tasks, retrieve information, and interact with tools through natural language. It distinguishes itself by being lightweight and modular, with an emphasis on transparency and user control. Clara is open-source, giving users the ability to inspect, modify, or extend its capabilities.
How is it used?
Clara runs locally through a user interface, where users can interact with it via natural language prompts. It supports plug-ins for various AI models, file systems, and APIs, enabling actions like summarizing documents, querying local files, or interfacing with code. Users typically configure Clara to fit their workflows, whether for research, development, or daily organization. Because it operates offline, it provides a responsive, secure experience suited for both personal and professional environments.
What is it used for?
Clara is designed for individuals who want a powerful AI assistant without sacrificing data privacy or relying on cloud services. It’s well-suited for tasks like drafting content, managing information, automating repetitive workflows, or serving as a companion for technical problem-solving. Clara’s architecture supports modular growth, making it adaptable across domains from creative writing to local data analysis. Ultimately, Clara offers a vision of AI that is both empowering and private by default.
Questions and Answers
Each studio ends with a question and answer session whereby attendees can ask questions of the pedagogy experts and technologists who facilitate the sessions. We have published a full FAQ section on this site. If you have other questions about GenAI usage, please get in touch.
-
Assessment Design using Generative AI
Generative AI is reshaping assessment design, requiring faculty to adapt assignments to maintain academic integrity. The GENAI Assessment Scale guides AI use in coursework, from study aids to full collaboration, helping educators create assessments that balance AI integration with skill development, fostering critical thinking and fairness in learning.
-
How can I use GenAI in my course?
In education, the integration of GenAI offers a multitude of applications within your courses. Presented is a detailed table categorizing various use cases, outlining the specific roles they play, their pedagogical benefits, and potential risks associated with their implementation. A Complete Breakdown of each use case and the original image can be found here. At […]