A study by University of Nijmegen scientist Markmanse and a computational linguist has found that small firms, with few resources, are the most open-sourced language models. The study found that almost half of the models that they analysed don’t provide any details about data sets beyond generic descriptors. The models must be open to scrutiny,…