Home / Technology / AI Slop: Humans Losing War on Deepfakes?
AI Slop: Humans Losing War on Deepfakes?
9 Sep
Summary
- Generative AI is rapidly evolving and being weaponized.
- AI detection tools constantly play catch-up with generative models.
- Watermarking is imperfect but a necessary security measure.
The digital landscape is now flooded with AI-generated content, blurring the lines between authentic and synthetic media. Dr. Hany Farid, a computer science professor, highlights the rapid evolution of generative AI, noting its increasing weaponization and the growing difficulty in discerning real from fake. His company, GetReal, specializes in digital forensics to identify AI's telltale signs.
Companies like Pangram and Originality.ai are developing tools to detect AI-generated text and media, but they face a constant challenge. These detection systems are in a perpetual arms race with generative models, which are continuously improving. As Max Spero of Pangram states, detection tools must "accumulate more certainty using more data" to remain effective.
Watermarking is emerging as a strategy to label AI-generated content, with the EU's AI Act mandating clear labeling and invisible watermarks for public interest information. However, watermarks are not a foolproof solution, as they can be bypassed and do not always account for AI-assisted content.
Ultimately, navigating this complex digital environment requires a combination of technological solutions and critical thinking. Dr. Farid advises developing a strong understanding of trusted sources, as reputable news organizations employ editorial standards and fact-checking processes that individuals and social media platforms often lack.
