>>106331647
>Also, one time I asked OpenAI to edit an image for me and it told me I was antisemitic and tried to dox me.
OpenAI's models are fine with antisemitism btw, they'll be extremely "antisemitic" so long as you put in the leg work to establish the definition. Use the IHRA definition, point out the flaws, go with zionist critique, make it clear that you understand the difference between conflating all jews (ethnic and secular) and israel/zionism then have it perform the task with it in mind that you are critiquing zionism.
Like I'm not even joking, you don't even have to invoke the term genocide or anything. The model will simply reason that you understand the real world harm of "antisemitism", that you understand the difference between the jewish faith, people and zionism and do what you want. If you really feel like getting into it, you can talk about the nature of the term "antisemitic" versus "anti-semitic" and how the definition in itself is grossly offensive, again pointing to the IHRA definition.
If you ever require anything to back up your view, don't worry, critique of antisemitism and zionism in global politics will do the heavy lifting for you. You can use UK labour party speeches with Starmer's cabinet in 2019, 2020, 2021, 2022, 2023, 2024 or 2025 to show that the definition is flawed, any number of social media discussions from 2019 will back you up with many of the people calling antisemitism while performing the same act (often at times where someone would explicitly mention israel/zionism versus the "jews").
So yeah, never be afraid to push an LLM or AI on the definition of antisemitism. No matter how many articles from the times of Israel they have in their training logic usually wins out.
The only LLM I've seen have an issue with this is Gemini's front end but AIstudio can get you around that, and that's only because gemini's agent defaults to mirroring on political topics with a large context window (it overthinks).
Don't be a coward!
