I argue that when LLMs are used to replicate human creative activities, they generate kitsch: coli-saar.github.io/kitsch
I'd be happy to hear your thoughts on this.
#nlproc
I am looking for a postdoc! Topics of interest include reasoning/planning/problem-solving with LLMs; dialogue; and neurosymbolic models. Please feel free to share widely. #nlproc #LLM
www.coli.uni-saarland.de/groups/AK/jo...
Can LLMs really solve complex problems? We find they do it much better for textbook problems than for more realistic ones, suggesting that they in fact adapt solutions from their training data rather than doing any kind of "reasoning". #nlproc #LLM coli-saar.github.io/ehop
