2024 Week 14

⚙️ Work

  • A vendor came to campus to offer sessions for instructors and staff on how to use their product. The product has changed pretty significantly from when we adopted it several years ago, and we’ll need to figure out if those changes are things that we as a community care about. Otherwise, they’ll become the de facto platform for that other stuff, short-circuiting a lot of strategic planning that’s already in progress.
  • Someone decided to replace the trusted UCalgary contacts database for quickly finding staff info with a new Drupal-powered thing that seems to find .doc attachments of CVs more often than anything else, after a several second delay. Awesome. With like 48 hours notice. (Ironically, when Contacts originally launched in 2009 to replace the even-older Seek cgi-bin tool, it was a lightweight Drupal site providing a front-end for UNITIS. It’s the UNITIS back-end that’s being retired, partially because the Drupal front-end is so old it can’t be patched.) With all of the effort we spend trying to manage and communicate and support changes for teachy-learny tools, an admin tool can just go “heyyyy - we decided to do something different. here you go.”
  • Leanne Wu started a fundraising campaign in memory of Fairooz Shafin, a student who passed away last year.

📚 Reading

  • Lindstrom, G. (2024). Rethinking critical thinking, diversity and Indigenous awareness from a Blackfoot perspective. Identities, 1–17. https://doi.org/10.1080/1070289X.2024.2335813 - Gabrielle Weasel Head is a Blackfoot scholar, was an educational developer in the TI, and is now an assistant professor at MRU. In this article, she describes the role of “relationality” as opposed to “critical thinking” in engaging with EDI concepts. (I have a PDF of the article if needed)

    Balance requires being an active participant in relationships as opposed to simply learning how to be more aware of others who we may see as different than us.


    Thus, a real person knows who they are. A real person is authentic and operates from a space of personal autonomy and is aware that others also require their own autonomy and individual freedom to develop their own unique identity. Being real is about the embodiment of being in respectful relations with both human and non-human entities.



A I A I O!

  • Tang, R., Chuang, Y.N., & Hu, X. (2024). The Science of Detecting LLM-Generated Text. Communications of the ACM. Posted March 6, 2024.

    Turns out, it’s really complicated to try to detect LLM-generated text, attempts take a LOT of jiggery-pokery, and aren’t very reliable. It works best when the LLM is provided as a “white box” service that also tries to detect the LLM-generated text (say, MS Word’s Copilot thing, and MS Word could add an “AI detector” that’s tuned to detect Copilot or something)

  • Andrew Nikiforuk @ The Tyee: No, AI Won’t Outsmart Our Climate Calamity. Hey - have we tried boiling the oceans with AI-generated waste heat to try to find a solution to this pesky “climate problem” yet?

  • Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2023). GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models. arXiv preprint. (via Scott Leslie)

    Not sure why it’s still a preprint a year later (originally released as a preprint on arXiv on March 17, 2023 - there’s the arXiv copy, and one hosted by OpenAI, but still nothing in a peer-reviewed journal?). 3 AI researchers and a Wharton economist (walk into a bar) try to figure out how to calculate which jobs are most at risk of being replaced by generative AI. They create a new “exposure” measurement, based on job duty descriptions from an O*NET database national database.

    We define exposure as a proxy for potential economic impact without distinguishing between labor-augmenting or labor-displacing effects. We employ human annotators and GPT-4 itself as a classifier to apply this rubric to occupational data in the U.S. economy, primarily sourced from the O*NET database.

    Which sounds awesome and useful and also like it would involve sweeping generalizations and assumptions. They had ChatGPT classify the job duty descriptions, which, I mean, I have questions about the value of that. But you have to start somewhere (or do you?). In their conclusion, they state that:

    Our analysis indicates that approximately 19% of jobs have at least 50% of their tasks exposed to LLMs when considering both current model capabilities and anticipated LLM-powered software.

    Their analysis of aggregated and ChatGPT-classified job duty descriptions, compared with estimated genAI capabilities at some point in the future. This is maybe a calmer reaction than the “OMG ALL JOBS ARE OBSOLETE AND ALL HUMANS ARE BEING REPLACED BY ROBOTS!” hype that makes up genAI articles, but still…

  • What AI Can Do Today - a list of 5,667 (and counting) AI-powered tools to do stuff. (via Scott Leslie)

  • Ian Bogost @ The Atlantic: AI Has Lost Its Magic - how much of the energy being poured into (and consumed by) genAI tools has been about the sheer novelty of the experience? And now that the novelty is fading, will the fascination start to fade as well?

  • Jon Stewart on the False Promises of AI - The Daily Show:

🍿 Watching

  • ★★★★☆ Loot (AppleTV) - We started watching season 2. Only the first 2 episodes have been released so far though.
  • ★★★★☆ The Reluctant Traveler (AppleTV) - another episode of season 2 dropped, with Eugene Levy checking out Milos, Greece.
  • ★★★★☆ The Gentlemen (Netflix) - The series by Madonna’s ex. A decent start. I’ll probably keep watching, at least until Fallout drops next week. For the record, I named the AI section of links long before watching the first episode of this…

🧺 Other

  • I spent a surprising amount of time cleaning up my RSS subscriptions, because they’re visible on the Blogroll page now.

🗓️ Focus for next week

  • Copyright committee meeting
  • Learning Technoloy Forum session
  • TI staff meeting
  • The Boy™ gets his wisdom teeth yanked
  • Retiremement hootenanny for 2 long-long-term Com/Media staff.
comments powered by Disqus