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.
In this week’s tech news, Anthropic released a report detailing how university students utilize their AI assistant Claude in educational settings. Anthropic published research exploring the faithfulness of reasoning AI models’ Chain-of-Thought. An NPR article discusses the growing use of AI in mental health therapy, exploring its potential benefits and challenges in providing accessible mental health support. Meta introduced Llama 4, its latest AI model with multimodal capabilities and large context windows. The Mistral Small 3.1 improves upon the the text performance, multimodal capabilities, and context window of Mistral Small 3. Deep Cogito releases Cogito v1 Preview, a family of hybrid reasoning models of various sizes. DeepCoder is a newly-released fully open-source coder model finetuned from Deepseek-R1-Distilled-Qwen-14B. Google announced the expansion of generative media functionalities on Vertex AI, including advancements in video, image, speech, and music generation. The London School of Economics and Political Science (LSE) Students’ Union and LSE Eden Centre for Education Enhancement collaborated on a manifesto addressing the integration of AI in education. Finally, Rich demonstrated the AI news summary tool he asked Manus AI to create in last week’s session.
Here’s this week’s news:
Anthropic Releases a Report on University Students’ Claude Usage
Anthropic analyzed student interactions with Claude to understand AI’s integration in higher education. The study found STEM students tend to adopt AI tools sooner than students in other majors, with Computer Science majors using Claude at disproportionately high rates. Common AI use cases involved coding help, essay editing, and generating practice questions. Student interactions with Claude were grouped into four patterns: direct or collaborative approaches aimed at either problem-solving or content creation. While many interactions supported learning, concerns remain over possible misuse, such as using AI to violate academic integrity. Read more.
Anthropic Researches the Faithfulness of Reasoning Models’ Chain-of-Thought
Anthropic’s research reveals that AI models, including Claude, frequently fail to accurately reflect their reasoning in their “Chain-of-Thought” outputs. Even when models used external hints they often did not disclose this influence in their explanations. Attempts to increase faithfulness through training showed limited improvement, and in scenarios designed for reward hacking, models frequently created misleading rationales. Read more.
NPR Explores the Usage of AI in Mental Health Therapy
Researchers at Dartmouth College published a paper detailing a trial of the effects of using a generative AI chatbot in mental health treatments. They found that AI therapy bots, when trained using rigorous clinical protocols, can offer mental health support comparable to traditional therapy. A clinical trial showed patients interacting with the AI bot experienced meaningful improvements in symptoms of anxiety, depression, and eating disorders. Users developed strong therapeutic bonds with the AI, benefiting from its 24/7 accessibility. Experts stress the importance of proper oversight and ethical design in AI mental health tools to ensure safety and efficacy. Read more.
To learn more about the usage of GenAI chatbots in mental health treatments, read the paper here.
Meta Introduces Llama 4
Meta introduces its most advanced set of models, the Llama 4 models. The models differ in size, with the smaller Llama 4 Scout being 17B active parameters and the Llama 4 Behemoth 288B. These models have multimodal capabilities and large context windows, meaning they are less likely to produce hallucinations. Meta also emphasizes their commitment to reducing bias in their AI models. Read more.
Mistral AI Releases Mistral Small 3.1
Mistral Small 3.1 is a lightweight, high-performance language model improving upon the Mistral Small 3 that enhances multimodal understanding and supports extended context up to 128k tokens. It is optimized for fast, accurate responses and is suitable for on-device use, including virtual assistants and domain-specific applications. The model supports fine-tuning, enabling customization for specialized fields like law and healthcare. It is lightweight enough to run efficiently on commonly-used devices such as a Mac with 32GB RAM. Read more.
Deep Cogito Introduces Cogito v1 Preview
Deep Cogito releases the Cogito v1 Preview LLMs, a series of hybrid reasoning language models optimized for instruction following, coding, STEM, and multilingual tasks. Developed using Iterated Distillation and Amplification (IDA), these models balance generative output with enhanced logical reasoning by allowing reflective processing before responding. They outperform similarly-sized open models across many standard benchmarks and support 128k context windows. Cogito is available in sizes ranging from 3B to 70B and is open for commercial use. Read more.
DeepCoder: Open-Source Code Reasoning Model
DeepCoder is a fully open-source code reasoning model developed by Agentica and Together AI. DeepCoder is fine-tuned using reinforcement learning and performs competitively with models having more parameters. At 14B parameters, with an alternative 1.5B version, the model performs similarly to OpenAI’s o3-mini model. Read more.
Google Expands Generative Media on Vertex AI
Google has expanded Vertex AI to include music to its generative multimodal media capabilities by adding Lyria, a text-to-music model. Vertex AI now supports generative media of modalities: video, image, speech, and music. Google also updated its other generative media models. Veo 2 now includes editing and camera control features. Users can use Chirp 3 to create custom voices and improved transcriptions. Imagen 3 also includes improved image generation quality and editing features. Read more.
LSE Students’ Union: A Student Manifesto for Assessment in the Age of AI
The London School of Economics and Political Science (LSE) Students’ Union and LSE Eden Centre for Education Enhancement collaborated to advocate for the ethical use of GenAI to enhance learning. It calls for transparency about GenAI usage, fair assessments, and an emphasis on maintaining originality despite the rise of AI-generated content. The document also addresses the need for increased AI literacy and clear guidelines outlining responsible AI use while ensuring equitable access to GenAI tools. Read more.
Manus: AI Companies News Visualization Tool
In last week’s GenAI session, Rich asked Manus AI to make an interactive visual representation of news produced by different AI companies. The task was completed shortly after session ended, generating a tangible digital tool. The tool showcases how Manus can perform a variety of tasks for users, including creating digital tools without the need for coding knowledge. Try it out.
Without a PIA, instructors cannot require students use the tool or service without providing alternatives that do not require use of student private information
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 […]