Using Generative AI
Q: My question is regarding the recent advancements in the computing world in “Artificial Intelligence”, specifically around Generative AI. Generative AI algorithms have advanced the field significantly where we can now generate text and media that is virtually indistinguishable from one created by humans. The AI is trained on large datasets of existing text and media. There are a growing number of software services offering these capabilities to businesses and consumers. While many of the use-cases are likely permissible when it comes to text generation, a growing number of impermissible applications will likely become available to the masses soon.
I will list some of the applications:
- generating audio: will allow generating any type of audio using any voice. Could be Quran-like verses from a renowned Qari, a fake conversation/dialog between public figures, or songs and music. Since these are generated by computer, the voice could be the emulation of a person who was not involved at all. Audio generation has quite a lot of harmful applications which could mislead people, given that it is human nature to trust voiceprints.
- generating text: enables consumers to generate prose in any style desired. Innocent applications are where we can use it to author content, summarize content, explain content, etc. The prose generated could be the emulation of anyone we desire. Therefore, harmful content could be something like generating fake verses of Quran, fake hadith, and worse.
- generating images/videos: enables users to generate images/videos just by providing a text prompt describing the scene/image desired in the output image/video. It may generate fake images, sow distrust, and increase immodesty significantly.
Note that all these generations are possible by a variety of input to the application (text-based prompts, image-/video-based prompts, sound-based, etc.).
I have some questions regarding the generation of content that would have been deemed impermissible if generated by a human:
1) Do the rulings of digital photography extend to photos generated by a computer through Artificial Intelligence? i.e. if one believes that capturing/drawing photos by hand or through camera is haram, should that belief carryover to computer-generated media? It seems like there’s a nuance here in who’s generating the content — human vs computer. Does that change the Sharee’e rulings of digital photography?
2) There’s a hukm regarding the purdah of a woman’s voice when speaking to a non-mahram, even when separated by a door/wall. Women are supposed to use harsher tones in those situations. Regarding generated audio, would the Sharee’e ruling of purdah apply here too? For example, if a female-emulated computer-generated voice is used in a maps navigation program, would that violate the principle of Purdah?
3) Until now, if anyone wanted to photograph me, they would require my presence. With the new technology, it may be possible to generate a person’s face that looks very close to mine. Would it be my responsibility to make sure nobody generates my image?