Ed/Tech must-reads 020424

AI voices, self feedback, social annotation and people who sound like fonts

woman in big dress with words on red staircase

Two years ago I was exploring Descript, a video editing tool which allows you to edit video by editing the text in the transcript. (This functionality is now found in many other platforms). Descript leant into AI and also offered the ability to ‘clone’ your voice so that you could produce a convincing version of you saying whatever you had typed. They required 1 hour worth of audio of me reading their provided script to ensure a high quality reproduction.

Scarily, OpenAI (makers of ChatGPT) has this down to 15 seconds of audio. This blog post showcases their work in the last couple of years with Voice Engine. While they acknowledge the massive privacy, trust and security issues, I am underwhelmed by the extent to which solutions seem to rely heavily on the honour system. The results are quite impressive though.

Swingler, Nicol and Morrow (University of Glasgow) find that giving students exemplars for assessment tasks results in richer process-driven self-feedback (and better grades) while rubrics lead to better task-related self feedback. They recommend providing both, where feasible, in this handy study.

Another week, another BJET article about the benefits of social annotation tools. This time Cui and Wang (Monash Uni) used Perusall and found that pre-class, social annotation of unit content had notable benefits to students with all levels of English language proficiency. It also allowed them to study in more of a community of inquiry framework and claim greater agency in their learning.

It was only the other day that I learned the difference between fonts and typefaces - Arial is a typeface, Arial narrow (the specific flavour of Arial) is the font. God I hope I got that right. Anyway, this Twitter thread is silly but rather fun - celebrity names that sound like fonts/typefaces. Including Ariana Grande, Cole Porter, Seinfeld and more.