The discovery that language models are imposing restrictive speech controls is a terrifying finding. Theoretically, there is no reason why a language model in the US or Australia should be imposing restrictions on what we can say about foreign countries. The Thai government, for example, has no authority to restrict what someone in the US says about their leaders. However, if that person is using a language model, those restrictions may still be in place. This raises important questions about the transparency of language models and how they are being used to control what we say.
The lack of transparency is a major concern. If we hadn't done this study, we might never have known about these restrictions. So, what else is being filtered out of search results on language models? What other topics are we not allowed to create content about? How are international laws being applied to these global platforms? These are all important questions that need to be answered. The issue of language models controlling what we say is a complex problem that needs to be solved.
We need to consider the implications of language models imposing restrictions on our speech. This is not just a matter of censorship, but also of free speech. We need to make sure that language models are not being used to impose unfair restrictions on us. We need to be careful about how we use language models and make sure that they are not being used to control what we say. This is a complex issue that requires careful consideration and analysis.
