That Isn't Me - How to Recognize Deepfakes and AI Generated Videos

That Isn’t Me examines how deepfakes and AI-generated videos are made, and demonstrates both the technical process and the practical difficulty of spotting convincing fabrications. The video begins by acknowledging sponsor Bitdefender and then dives into a live demonstration of creating a deepfake, highlighting how a target actor is matched for body shape, the training of a face model on thousands of images, and the resulting increase in realism over five years. Linus explains that syncing lips with audio remains challenging but feasible for short clips, and that multiple takes with different angles can produce a surprisingly long sequence that feels authentic to many viewers. The host then contrasts the ease of creating deepfakes with the real world impact, emphasizing the threat to everyday people and families, and sharing alarming statistics about rising scam losses. The discussion moves from the technical hurdles of generating convincing footage to practical advice on spotting fakes, including cues like inconsistent shadows, vanishing points, lighting anomalies, and non convergent perspective in AI imagery. The piece emphasizes skepticism on social media, urges digging deeper, and advises verifying through trusted channels before reacting to urgent requests. Finally, the video covers the evolving nature of AI video generation, the tradeoffs of different models, and encourages audiences to protect themselves and loved ones with critical thinking and cyber security tools. The closing segments reiterate the sponsor’s security solutions and invite viewers to explore further resources on how to recognize AI-generated content, while acknowledging the rapid pace of AI improvement. The overall message is clear: as AI video quality improves, vigilant evaluation and corroboration remain essential for distinguishing real events from fabricated media.

Topics · science & technology · cybersecurity · ai ethics · media literacy