I've really been quite surprised at the lack of censorship in the translations. Of course, there could be some that I haven't noticed, but the content still makes much more sense than literal translations. As far as data privacy, that's certainly a concern. Although it's not really ChatGPT doing the translations, it's the OpenAI API platform, a separate product. According to their Privacy Statement, "Data submitted through the OpenAI API is not used to train OpenAI models or improve OpenAI’s service offering." It's the consumer grade ChatGPT that's used for training, not their commercial offerings.Interesting, thanks for sharing @Popodog On this topic I was wondering: isn't the main problem that Whisper already censors heavily before you can even translate it? That's the assumption that I had at least but since I can't read Japanese I can't really check on it. Because I also tried it with other translators and they gave similar results. Could be that they all censor the same way, for example google translate.
Of course 'censorship' and 'context' are two different topics but I think that if we could remove the censorship somehow it would matter/improve the most.
Anyways, your tip is welcome. I'll try it. Any other AI model besides ChatGPT that would give similar results?
Yes, it will create a history for this account and you can bet that they will use what you do for future training data. It's a nightmare if you care about privacy. Of course for this task you would create a separate account (separate email / payment method) that you would not use for anything else and don't share anything that links you to your personal data/geo data so better use a VPN with it. It's not perfect but what else can you do? For example I have a credit card just for xxx stuff and sure it has my name on it but how else would you handle this matter? This credit card provider when he looks at my statements:
My opinion on this matter: Honestely whatever, ok I like naked Japanese girls...What are they going to do about it?
The only capable AI model that I'm aware of that you could just store locally would be DeepSeek but I'm not sure if it could do this specific task as well.
If I were in a position that could be compromised by the discovery of my translation activities, I might be more concerned about it. It would certainly be something to consider if you are. For instance, if you live in a highly conservative country, or you might apply for a position at an intelligence service in the future, you may want to be careful about it