Derek Hunt

Sep 16

“We must prepare students for a world where the question is not, ‘Can you use AI?’ But rather, ‘Can I trust your thinking without using AI?’” https://www.theglobeandmail.com/business/article-professors-experiment-as-ai-becomes-part-of-student-life/

Sep 12

✅ AI use is increasing and becoming a new mode for cheating but cheating numbers overall have been high well before Gen AI 🙊Students are sharing incidents of being incorrectly accused of using AI, leading to unexpected behaviors like using AI so it looks like they didn’t use AI (!) ❓There is stil...Read more

✅ AI use is increasing and becoming a new mode for cheating but cheating numbers overall have been high well before Gen AI 🙊Students are sharing incidents of being incorrectly accused of using AI, leading to unexpected behaviors like using AI so it looks like they didn’t use AI (!) ❓There is still a lot of confusion for students about when AI is permissible and it is varying from teacher to teacher https://www.vox.com/technology/458875/ai-cheating-data-education-panic

Sep 9

"Sycophancy" - a new term I learned today that describes one of the main things that bug me about ChatGPT / GenAI chatbots - "Sycophancy, in which chatbots agree with and excessively praise users" https://www.nytimes.com/2025/08/08/technology/ai-chatbots-delusions-chatgpt.html "Sycophantic Improv ...Read more

"Sycophancy" - a new term I learned today that describes one of the main things that bug me about ChatGPT / GenAI chatbots - "Sycophancy, in which chatbots agree with and excessively praise users" https://www.nytimes.com/2025/08/08/technology/ai-chatbots-delusions-chatgpt.html "Sycophantic Improv Machine ... they also use the history of a particular conversation to decide what should come next, like improvisational actors adding to a scene. The story line is building all the time”

Sep 4

This Linkedin post was very helpful for me to understand why GPTs using custom resources do not always give the greatest results, despite having a clear system prompt asking it to use the included resources. It also points out that well structured documents as custom resources perform the best - ...Read more

This Linkedin post was very helpful for me to understand why GPTs using custom resources do not always give the greatest results, despite having a clear system prompt asking it to use the included resources. It also points out that well structured documents as custom resources perform the best - "The search and retrieval isn't done by the LLM itself ... is a hack using Azure AI Search with naive, vector-based semantic similarity scoring and ranking. The problems come with unstructured documents and the inherent ambiguities and assumptions in language that an expert human can pick out, but non-experts (LLMs) can't."

Sep 2

"To the class of 2025: Will you let AI think for you, or learn how to think for yourself?: The challenges and frustrations that can come with thinking deeply - a skill that’s more valuable than ever - are a feature, not a bug. So you might reasonably wonder: If artificial intelligence can do these ...Read more

"To the class of 2025: Will you let AI think for you, or learn how to think for yourself?: The challenges and frustrations that can come with thinking deeply - a skill that’s more valuable than ever - are a feature, not a bug. So you might reasonably wonder: If artificial intelligence can do these things, why should you spend four years learning to do them yourself? Here’s the truth: You need to learn these things precisely because machines can now do them. Your ability to think for yourself has become more precious than ever. " https://www.theglobeandmail.com/opinion/article-artificial-intelligence-university-humanities-critical-thinking/