I like your use case for discussion boards. Narrowing the purpose and scope is key to making currently technologies work well. Just telling it to take on a role is very limited.
I do think it might be possible to build a "historical" LLM that is more useful, but it would take a lot of work curating and structuring the data used by a customized model ... more work than almost anyone would would want to do ... particularly for this purpose.
I think that's right. And it would mean creating some context or frame for the encounter with an LLM. That lack of context and the fact that getting an LLM to do what you want it to do reliably is hard are real challenges.
I read this a few days ago and canβt stop thinking about it. (Not only because I have a Phaedrus-inspired tattoo.) i think there is so much room to broaden our traditional canon of literature that is often taught in classrooms as just the bare minimum to start engaging with a wider array of readers.
Adding discursive features, as enrichment and not replacement, has so much cool potential. Anyway, youβre brilliant. Thank you for giving me something to ponder.
Hi Lily, First off, a Phaedrus-inspired tattoo? That's awesome! Thanks so much for your comment. I'm so glad the essay gave you pondering material.
If my students go for it, I'll be writing about using an LLM as a learning tool in a history class. I'm hoping both my experiments with an LLM and my radical Deweyan pedagogy will be of interest to readers. Rather than essays, I'm going to try out a dedicated page that I update through the semester. Here is the link: https://ailogblog.substack.com/i/147973450/welcome-to-π¨π°-π³ππ-teaches Hope you find some things worth pondering there. I know I will!
LLMs are tools not people!
I like your use case for discussion boards. Narrowing the purpose and scope is key to making currently technologies work well. Just telling it to take on a role is very limited.
I do think it might be possible to build a "historical" LLM that is more useful, but it would take a lot of work curating and structuring the data used by a customized model ... more work than almost anyone would would want to do ... particularly for this purpose.
I think that's right. And it would mean creating some context or frame for the encounter with an LLM. That lack of context and the fact that getting an LLM to do what you want it to do reliably is hard are real challenges.
I really donβt have anything thoughtful to add.
I read this a few days ago and canβt stop thinking about it. (Not only because I have a Phaedrus-inspired tattoo.) i think there is so much room to broaden our traditional canon of literature that is often taught in classrooms as just the bare minimum to start engaging with a wider array of readers.
Adding discursive features, as enrichment and not replacement, has so much cool potential. Anyway, youβre brilliant. Thank you for giving me something to ponder.
Hi Lily, First off, a Phaedrus-inspired tattoo? That's awesome! Thanks so much for your comment. I'm so glad the essay gave you pondering material.
If my students go for it, I'll be writing about using an LLM as a learning tool in a history class. I'm hoping both my experiments with an LLM and my radical Deweyan pedagogy will be of interest to readers. Rather than essays, I'm going to try out a dedicated page that I update through the semester. Here is the link: https://ailogblog.substack.com/i/147973450/welcome-to-π¨π°-π³ππ-teaches Hope you find some things worth pondering there. I know I will!