How large language models can reshape collective intelligence

 

Large language models are the product of humanity’s collective intelligence, but they also have the potential to reshape it, argues a multidisciplinary team of researchers. New technologies tend to speed up our ability to retrieve information while making it less obvious where the information came from. Key to striking a balance between speed and transparency is understanding how LLMs were developed, what data sources they were trained on and what they’re being used for. Otherwise, groups might think they’re building consensus, when they’re really just reflecting beliefs from an opaque or even biased training set. (Nature Human Behaviour | 43 min read)