The need for transparency in large language models

Large language models can be seen as human student classifiers, similar to the task of classifying tweet messages. While humans are also black boxes with biases, the advantage of language models is the potential for open access and transparency. However, it is not ideal to rely solely on models like chat GPT. Researchers and social scientists should have their own language models with known training data and regular updates. Understanding the inner workings of the model is not necessary for its effectiveness.