With recent advances in Artificial Intelligence and increasing emphasis on human augmentation and collaboration, time is ripe for AI-enhanced support tools which empower the public to formulate and visualize a collective vision of societal issues such as climate change. Here, we report on crea.visions, a platform for human-AI co-creation within Sustainable Development Goals centered community engagement. We present in-the-wild experiments with four versions of crea.visions involving 1,000+ participants and 25,000+ generated images over three years: Versions 1 and 2 focused on developing the novel tool empowering citizens to artistically communicate their favorite abstract societal issues. In versions 3 and 4, the generic image generation GAN was replaced by custom-trained versions for Venice and Paris respectively. Refining the platform towards communityspecific action, users of version 4 can geotag their identified problems, submit solutions candidates, and are actively linked up with relevant NGOs. Finally, version 4 includes the first workflow todate which combines AI image-generating modalities of sliders and text-to-image.
|Udgivet - 2023