Open generative AI models are a way forward for science

If another group tries to replicate that study, who knows whether the model's underlying training data will be the same, or even whether the technology will still be supported? The texts used to train these models are unknown: they might include direct messages between users on social-media platforms or content written by children legally unable to consent to sharing their data. [...]I think academics, especially those with a large social-media following, shouldn't be pushing others to use proprietary models.