The Dilemma of Deepfake Technology
In February, a Tomahawk missile struck an Iranian school, causing over 150 fatalities, mostly children. President Trump initially questioned whether the missile belonged to the U.S., highlighting the need for investigation due to the generic nature of Tomahawk missiles sold to various countries. Iran released a video claiming the missile was American, prompting inquiries into its authenticity.
Hany Farid, a leading expert on deepfakes and Dartmouth College professor, was contacted to analyze the video. Despite the challenging nature of the footage, he considered aspects such as the speed of sound and light and missile size. His analysis led to the conclusion that the video was genuine.
Farid points out the difficulty in distinguishing real from artificial intelligence-generated content, emphasizing the necessity of scientific explanation in the legal system. He illustrated this with a case involving Alex Pretti’s death, where AI altered a video to incorrectly show a gun in Pretti’s hand, showcasing how AI can manipulate perceptions.
The internet is inundated with manipulated content, often referred to as “AI slop”. Farid asserts that many viewership-driving videos are fabricated using AI. His focus remains primarily on deepfakes, which pose serious threats due to their usage in disinformation campaigns and manipulation of identities.
Farid discusses the challenges posed by AI agents. These agents, possessing autonomy, can create fake identities and interact with humans. Concerns over AI safety have been amplified due to trials indicating potential rogue behavior.
Reflecting on personal implications, Farid considers the idea of AI agents retaining human likeness, contemplating whether they could serve as adequate replicas. He recognizes that future generations might embrace these possibilities differently.
Farid criticizes social media’s detrimental effects, particularly its addictive design. He cites the $17 billion settlement from Meta, which introduced safeguards but believes further actions are necessary. He advises reducing interaction with social media and questions the efficacy of websites claiming to detect AI fakes.
Looking ahead, Farid anticipates heightened autonomy in AI agents, predicting they might generate fakes without human involvement. This evolution poses significant challenges, demanding ongoing vigilance and discourse on AI’s role in society.