Last week, Meta announced that it was rolling out AI stickers across its apps, including Facebook Messenger. However, the new AI sticker feature, which is meant to allow for fun and creative expression, is already running into trouble, with some users creating disturbing or offensive images.
The tool uses Meta's Emu image synthesis model to turn text descriptions into simulated pictures that can be shared as stickers. However, within days of the rollout, artist Pier-Olivier Desbiens posted a viral thread on X (Twitter) showing problematic images he and others had created using the feature.
These included copyrighted characters like Mickey Mouse in violent or sexualized scenarios. Desbiens commented that "no one involved has thought anything through" with this launch.
I'm still trying to figure out if I love this or hate this pic.twitter.com/rrm6oKyVQb— Kazzer✘ (@kazzerx) October 4, 2023
Other AI-generated stickers quickly spread on social media, pushing boundaries, such as depictions of the 9/11 attacks, toy figures with weapons, and political figures in humiliating poses. This follows a trend seen previously with uncensored image generators, where some experimenters actively try to break content filters.
While Meta has screening methods to catch the most inappropriate content, the near-infinite possibilities of these AI tools make thorough moderation extremely difficult. The company acknowledged this, with spokesman Andy Stone pointing to its commitment to improving responsible development. He noted filters will evolve as more feedback is received on unwanted outputs.
found out that facebook messenger has ai generated stickers now and I don't think anyone involved has thought anything through pic.twitter.com/co987cRhyu— defend trans rights🏳️⚧️ - podesbiens.bsky.social (@Pioldes) October 3, 2023
Users argue that Meta moved too fast, introducing the consumer-facing sticker feature without stringent enough safeguards. Before stricter guidelines were implemented, earlier beta tests of generative AI from companies like OpenAI produced similarly problematic images.
It remains to be seen if Meta will pull back the new sticker tool or bolster its keyword blocking in response to this backlash. However, as AI's capabilities advance, the challenge of moderating unpredictable creative systems will only intensify.
Source: X via VentureBeat