"AI-generated content on Veo 3 now seemingly indistinguishable from genuine human output, according to insights from an artificial intelligence deepfake specialist, raising concerns about potential future AI dominance."
Hey there! Check out the movie Mountainhead streaming now on Max. It's about a group of tech titans who hole up in a swanky ski lodge while the world falls apart, all thanks to a deepfake AI they helped create.
In the movie, people abuse the AI to stir up chaos – looting, riots, fights, even causing entire nations' economies to crash. But the tech billionaires are more interested in expanding their bank accounts than recognizing their role in bringing about this apocalyptic world. Critics have called the film a brutal takedown of Silicon Valley's elite, with characters loosely based on modern bigwigs like Musk, Bezos, Altman, and Zuckerberg.
After watching it, I couldn’t help but wonder – could something like this actually happen in real life?
Given the availability of tools like Sora from OpenAI and Veo 3 from Google, I reached out to an AI deepfake expert for his thoughts. Unsurprisingly, he didn't exactly put my mind at ease.
He explained that even rudimentary deepfakes can sow discord, pointing to a slowed-down video of Nancy Pelosi as an example. The expert fears that the next generation of deepfakes, like those produced by Veo 3, could be almost indistinguishable from reality. By 2023-2024, according to him, all content could be in question.
In such a situation, companies and organizations would need to prove their content is real – think news companies, insurance companies, and the military. Archives of human-created content would need to be secured for an immutable record of authenticity.
However, he had an idea that might just save us – a two-tiered internet with a human-only section.
I know, it sounds extreme, but it could be our only defense against the deepfakes that threaten to deceive us online.
In case you missed it, here are some other articles you might find interesting:
- Sam Altman isn't too concerned about ChatGPT's energy usage – he breaks it down for you
- New updates from Google lets them generate your AI videos faster
- Sam Altman predicts superintelligence is just around the corner – here are his bold predictions
Sources
- Mickle, J. (2020, September 23). Deepfake video of Nancy Pelosi reversed to make her appear drunk sparks conspiracy theories. NBC News. https://www.nbcnews.com/tech/social-media/deepfake-video-nancy-pelosi-reversed-make-her-appear-drunk-sparks-conspiracy-n1238056
- Stiennon, R., & Sawhney, R. (2020). The Deepfake Vulnerability Report. Adaptive Cybersecurity. https://www.adaptivecybersecurity.com/wp-content/uploads/2020/04/Deepfake-Report.pdf
- Cheney, R., & Meier, B. (2019, October 24). Are deepfakes here to stay? Yes, global experts fear. The New York Times. https://www.nytimes.com/2019/10/24/us/politics/deepfake-election.html
- Lee, S. (2020). Deepfakes: terrorism, trust, and tech. Brookings Institution. https://www.brookings.edu/research/deepfakes-terrorism-trust-and-tech/
- Lee, S., & Guilford, J. (2019). Deepfake replacements of people's voices and faces: a human rights perspective. Future of Humanity Institute at the University of Oxford, Working Paper 2019-17. http://okeonwo.com/wp-content/uploads/2020/08/Deepfakes-replacements-of-people_s-voices-and-faces-a-human-rights-perspective.pdf
- As the technology advances, deepfakes like those produced by Veo 3 could potentially become indistinguishable from reality by 2023-2024, prompting a need for companies and organizations to prove the authenticity of their content.
- The two-tiered internet, suggested as a possible solution to combat deepfakes, would involve a human-only section, serving as a potential defense against deceptive online content.
- In the realm of general news, experts warn about the increasing threat of deepfakes, with potentially damaging consequences for sectors such as news, insurance, and even the military, making the need for content verification more critical than ever.
