Ethical and Responsible AI in the Age of Adversarial Diffusion Models: Challenges, Risks, and Mitigation Strategies
International Journal Science and Technology · 2022
DOI: 10.56127/ijst.v1i3.1963 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal Science and Technology, 2022. Verfasst von Tejaskumar Pujari und 2 weiteren Autor:innen. Die Arbeit wurde bisher 4-mal zitiert (Crossref).
Abstract (Originalsprache)
The rapid pace of diffusion models in generative AI has completely restructured many fields, particularly with respect to image synthesis, video generation, and creative data enhancement. However, promising developments remain tinged with ethical questions in view of diffusion-based model dual-use. By misusing these models, purveyors could think up deepfaked videos, unpredictable forms of misinformation, instead outing cyber warfare-related attacks over the Internet, therefore aggravating societal vulnerabilities. This paper explores and analyzes these potential ethical risks and adversarial threats of diffusion-based artificial intelligence technologies. We lay out the basis for good AI-the notion of fair, accountable, transparent, and robust (FATR) systems-discussing efforts underway to mitigate these ethical risks through watermarking, model alignment, and regulatory mechanisms. Thus, from the dialogue with ethical viewpoints, also touching upon cybersecurity, military policy, or governance, we present a conceptual model to encapsulate probable ethical considerations in the development and deployment of diffusion models. Human-centered values need to be advanced by a proactive convergent bonding among researchers, decision-makers, and civil society players during the strengthening of a tributary of generative AI's power.
Zitieren
APA
Pujari, T., Goel, A., & Kejriwal, D. (2022). Ethical and Responsible AI in the Age of Adversarial Diffusion Models: Challenges, Risks, and Mitigation Strategies. International Journal Science and Technology. https://doi.org/10.56127/ijst.v1i3.1963
Vancouver
Pujari T, Goel A, Kejriwal D. Ethical and Responsible AI in the Age of Adversarial Diffusion Models: Challenges, Risks, and Mitigation Strategies. International Journal Science and Technology. 2022. doi:10.56127/ijst.v1i3.1963
BibTeX
@article{Pujari2022,
author = {Pujari, Tejaskumar and Goel, Anshul and Kejriwal, Deepak},
title = {Ethical and Responsible AI in the Age of Adversarial Diffusion Models: Challenges, Risks, and Mitigation Strategies},
journal = {International Journal Science and Technology},
year = {2022},
doi = {10.56127/ijst.v1i3.1963}
}
EUVIMED