AI can now generate entire videos, songs, and deepfakes, posing a significant threat to public discourse, as shown by the real-world Zelensky deepfake.
With major global elections underway, AI tools can create psychographic profiles and spread false stories six times faster than true ones, echoing the Cambridge Analytica scandal.
Creating a real-time deepfake involves a multi-step process including face detection with MediaPipe, pixel segmentation, background inpainting, and using a dual autoencoder to swap faces.
A live demonstration and real-world examples show how deepfakes can be used for social engineering, leading to significant financial loss and tricking people into executing malicious files.
A GPT-2 model was fine-tuned on data scraped from Reddit and Twitter to automatically generate convincing but inflammatory social media comments, demonstrating bot capabilities.
Automated AI detection software often fails to identify novel generation methods, while simple "shallow fakes" like slowed-down videos or out-of-context images are highly effective at misleading the public.
The most reliable way to combat misinformation is through critical thinking, which involves investigating the source, checking the date, considering satire, and being aware of personal biases.
46:42Thomas Endres & Martin Förtsch & Jonas Mayer
30:38Timo Zander
24:23Balázs Kiss
56:55Ekaterina Sirazitdinova
28:05Toju Duke
27:26Cassie Kozyrkov
22:08Zachary Powell
33:24Laura Moritz, Nina Levchuk, Laura Möller & Nica Huestegge
Jobs that call for the skills explored in this talk.
DTS Systeme GmbH
Deutsches Medizinrechenzentrum GmbH
RWTH Aachen