I am returning to Clozemaster after a hiatus because I need to learn Russian. I used the platform with satisfaction to learn German, Mandarin Chinese, and Italian, for well over 5 years.
I’ve noticed that Clozemaster likes to experiment with AI in some new features. Don’t get me wrong: TTS is AI, and it works overall really great. But I don’t see the added value of the AI-generated pictures in the “Sentences list” - they are mostly distracting in my opinion, and they look really bad to me. I mean, we are on the wrong side of the uncanny valley there. How about we let users imagine the situation behind the cloze sentences themselves? Or else, give these AI a few more years to improve?
I am aware this doesn’t affect my experience with the regular fast track currently. My main wish would be that this doesn’t change – I would certainly completely stop using Clozemaster if the AI-generated images were to be implemented in the fast track. To me, the low quality of the pictures gives me a feeling of low quality of the platform, which I am sure isn’t what we want. Quality human-generated content is what makes my learning experience pleasant and effective.
I suggested https://thenounproject.com/ as one option that has a lot of human-generated icons (at least, back when I used it in my classroom almost a decade ago), and in general to prioritize human-generated content unless there’s a really good reason to use ML/LLM/AI stuff (accessibility, equitable access to education and other resources; this category is the big one for me).
I find these AI-generated images disturbing, or at least distracting. There’s a reason the documentary about Capitalism in the 21st Century, based on Thomas Piketty’s book and narrated by George Monbiot (who wrote a shorter history of capitalism), used AI-generated images of obscenely wealthy individuals; they’re disturbing.
Making up a funny link to someone’s name can help remember their name, and sometimes these (private) funny mental links can be absurd! I’d rather that that the emotions I feel when I see this off-putting slop.
So far I avoid using images in Clozemaster.
They definitely have that uncanny valley feel to them, but I’ve found them quite helpful when doing listening practice. First listen with my eyes closed, then check out the image, and only then view the text if I need it.
Ideally I always prefer quick black and white sketches (done by an actual person) like the ones you see in the Nature Method books from the 1960s. But for listening I prefer any image at all, no matter how it was made.
I must admit that when the AI-generated images first showed up in Clozemaster, I liked them, just because they were weird and sometimes grotesque, with people having six-finger hands and three legs, etc. But after some time my interest waned and finally I took the option to only show them on demand, as I found them distracting.
