I will stop using Google Gemini because it is 1) inacceptable for them to log paid usage prompts and 2) to me/us utterly inacceptable how much their policy restricts the use of their model. What is illegal or not is not for them to define: if I use a model as a chemist for a RAG system, the description or details about the synthesis for a substance which is prohibited in general may be perfectly legal. If I am a scientist doing research into social phenomena, analyzing scientific documents about things like terrorism or misogyny, prompts may well contain references to this and are not illegal at all, yet apparently prohibited by their terms of use.
If I am an ML researcher investigeting the limitations or properties of LLMs, using a benchmark with prompts on such topics is helpful to judge and describe the properties of LLMs, but this would also be prohibited by their terms.
Extremely disappointing again, I find myself staying away from pretty much all of Google’s services by now because they are restricted, user-unfriendly or incredibly overloaded with bureaucracy.
And of course, there is, like with this, no good way to give feedback directly to Google, let alone get a response.
1 Like