Etzioni outlines three key areas that need to be quantified to understand the extent of the problem: the volume of deepfakes, the number of people who see them, and the impact they have. He predicts that over the next four years, we will become much more adept at measuring these factors. He also warns that industry attempts to make generated media more obvious, such as watermarking images and text, are insufficient to combat malicious actors who want to avoid detection.
Key takeaways:
- Oren Etzioni, an AI researcher, warns that the threat of AI-generated disinformation is real and pervasive, but often not targeted at the general public.
- TrueMedia, a nonprofit, offers a free service for identifying fake or real media, and is building a foundation of ground truth material to improve detection.
- Etzioni outlines three key ways to quantify the problem of disinformation: determining how much is out there, how many people see it, and how much impact it has.
- Despite the threat, the most recent election took place with less AI-generated disinformation than feared, not because it isn't common, but because its creators didn't feel the need to participate.