That ##LLM paper is quite worrying, not least for the fact that people were influenced even when they knew they were being led by the chatbot. But, anyway, I do love the fact that they used two Murakami books for the experiment. #村上春樹
Reposted from
Stephanie Wang
#eeac Day 13: @tfwerner.com JMP on algorithmic vs. human collusion is of course terrific; his paper on LLMs steering consumers is also worth a look arxiv.org/pdf/2409.12143 "Using a behavioral experiment, we show that [LLMs] can indeed significantly shift consumer preferences." #econsky #econjmp
Comments