Generative AI is still far from perfect in its current state based on the numerous issues that are linked to it. That is especially true in the case of Meta's AI, considering that it has been experiencing issues with its AI sticker generator in WhatsApp, previously with Messenger as well.
Inappropriate WhatsApp Stickers
Meta has made it possible for users to create their own stickers through text prompts. You can generate a sticker of a baby dressed as Superman or a llama doing skateboard tricks. However, the company did not anticipate that the feature would generate stickers that could be seen as offensive.
A similar problem happened with Messenger's AI generator for stickers, wherein it used the likeness of government officials like former US President Donald Trump and Canadian Prime Minister Justin Trudeau in unflattering scenarios.
It's been found that WhatsApp, another app under Meta, sometimes generates stickers of children holding guns when the text prompt includes the word "Palestine." According to The Verge, adding the word "Israel" does not have the same effect.
This flaw is undeniably controversial, especially with all that has been going on between Israel and Palestine. It's not just a problem ethically, but Meta appearing like it is biased against Palestinians may cause bigger issues.
Meta spokesperson Kevin McAlister stated that the company is already addressing the problem and that Meta will "continue to improve these features as they evolve and more people share their feedback."
It doesn't bode well for Meta that even before the conflict between the two Middle Eastern countries flooded social media, it had already been experiencing issues with its AI generator. This contributes to the growing concerns people have over AI.
Apps Under Meta with AI Sticker Generators
Messenger was the first to have the feature, and users immediately started to test the boundaries of the feature only to find that it might not have one. One user, namely Pier-Olivier Desbiens, managed to create content that violates Meta's own rules.
For instance, certain text prompts managed to create stickers of the beloved character Elmo from Sesame Street holding a knife. Upon typing "child soldier," the AI tool generated photos of children in military uniform holding assault rifles, as reported by Ars Technica.
Seeing as Meta is expanding the feature to other apps, the best decision might be to fix the tool first and implement policies that prevent users from creating insensitive stickers. It also should not be limited to moderating text prompts.
Meta should start looking into the kind of images that can be used to generate stickers as well, especially since it is already testing an Instagram feature where users are able to create stickers from uploaded photos.
While the potential feature does not use AI, being able to use photos also gives users the opportunity to create stickers out of anything they want, which can be dangerous. It all boils down to Meta's ability to prevent users from using their tools to create offensive or insensitive content.
