Greetings from the Australian National University in Canberra, where Dr. Katy Gero, is speaking at AI & Friends, on "How do we get the benefits of LLMs without causing long-term harm?".They conducted an experiment where student written and AI assisted essays were mixed together and graded by a human marker. The marker couldn't tell which were from AI, and graded the AI submissions slightly higher than the human written ones.
One insight from this research was the question of the role of writing in learning. Students may be willing to outsource essay writing as they don't see it as a core skill and something taking up time they could use for more important study. The student might use AI to write their idea better. This is a view I have sympathy with, as someone who barles passed English at school and failed French. I require a human editor, or tools for spelling and grammar, to be able to write well enough to be employed.
This work also raises questions about how people cooperate to produce written works. I suggested Dr. Gero look to the history and politics of open source software for inspiration.
ps: Will we have an accelerated Peter Principle, with people promoted far beyond their ability, due to AI assistance? ;-)
No comments:
Post a Comment