Deepfake

Deepfakes (a portmanteau of 'deep learning' and 'fake'[1]) are images, videos, or audio which are edited or generated using artificial intelligence tools, and which may depict real or non-existent people.

[7] Deepfakes have garnered widespread attention for their potential use in creating child sexual abuse material, celebrity pornographic videos, revenge porn, fake news, hoaxes, bullying, and financial fraud.

[48] The mobile app Momo created the application Zao which allows users to superimpose their face on television and movie clips with a single picture.

[66] The use of AI technology was a major issue in the 2023 SAG-AFTRA strike, as new techniques enabled the capability of generating and storing a digital likeness to use in place of actors.

[68] Disney improved their technology through progressive training programmed to identify facial expressions, implementing a face-swapping feature, and iterating in order to stabilize and refine the output.

[77] In the 2024 Indian Tamil science fiction action thriller The Greatest of All Time, the teenage version of Vijay's character Jeevan is portrayed by Ayaz Khan.

[78] In March 2018 the multidisciplinary artist Joseph Ayerle published the video artwork Un'emozione per sempre 2.0 (English title: The Italian Game).

[79] The artist used Ornella Muti's time travel to explore generational reflections, while also investigating questions about the role of provocation in the world of art.

A report by the American Congressional Research Service warned that deepfakes could be used to blackmail elected officials or those with access to classified information for espionage or influence purposes.

[92][93][94] In 2023 the bands ABBA and KISS partnered with Industrial Light & Magic and Pophouse Entertainment to develop deepfake avatars capable of performing virtual concerts.

[95] Fraudsters and scammers make use of deepfakes to trick people into fake investment schemes, financial fraud, cryptocurrencies, sending money, and following endorsements.

[157] The Congressional Research Service cited unspecified evidence as showing that foreign intelligence operatives used deepfakes to create social media accounts with the purposes of recruiting individuals with access to classified information.

The deepfake videos featured an "artificial intelligence-generated doppelganger" of Cruise doing various activities such as teeing off at the golf course, showing off a coin trick, and biting into a lollipop.

A deepfake photograph appears to have been generated together with a legend for an apparently non-existent person named Oliver Taylor, whose identity was described as a university student in the United Kingdom.

The Oliver Taylor persona submitted opinion pieces in several newspapers and was active in online media attacking a British legal academic and his wife, as "terrorist sympathizers."

[121] AI researcher Alex Champandard has said people should know how fast things can be corrupted with deepfake technology, and that the problem is not a technical one, but rather one to be solved by trust in information and journalism.

[164] Li predicted that genuine videos and deepfakes would become indistinguishable in as soon as half a year, as of October 2019, due to rapid advancement in artificial intelligence and computer graphics.

They have a potential to interfere with democratic functions in societies, such as identifying collective agendas, debating issues, informing decisions, and solving problems though the exercise of political will.

[174] A team at Massachusetts Institute of Technology published a paper in December 2021 demonstrating that ordinary humans are 69–72% accurate at identifying a random sample of 50 of these videos.

[175] A team at the University of Buffalo published a paper in October 2020 outlining their technique of using reflections of light in the eyes of those depicted to spot deepfakes with a high rate of success, even without the use of an AI detection tool, at least for the time being.

[176] In the case of well-documented individuals such as political leaders, algorithms have been developed to distinguish identity-based features such as patterns of facial, gestural, and vocal mannerisms and detect deep-fake impersonators.

[177] Another team led by Wael AbdAlmageed with Visual Intelligence and Multimedia Analytics Laboratory (VIMAL) of the Information Sciences Institute at the University Of Southern California developed two generations [178][179] of deepfake detectors based on convolutional neural networks.

[215] However, media that has been edited to alter the order or context of words in one's message would remain on the site but be labeled as false, since it was not generated by artificial intelligence.

[217][225] In September 2018, Google added "involuntary synthetic pornographic imagery" to its ban list, allowing anyone to request the block of results showing their fake nudes.

[226][check quotation syntax] In February 2018, Pornhub said that it would ban deepfake videos on its website because it is considered "non consensual content" which violates their terms of service.

[234] Several states have also introduced legislation regarding deepfakes, including Virginia,[235] Texas, California, and New York;[236] charges as varied as identity theft, cyberstalking, and revenge porn have been pursued, while more comprehensive statutes are urged.

6943: No AI Fraud Act into the 118th United States Congress on January 10, 2024, to establish specific property rights of individual physicality, including voice.

[257][258] In the summer of 2018, MediFor held an event where individuals competed to create AI-generated videos, audio, and images as well as automated tools to detect these deepfakes.

[260] In 2019, DARPA hosted a "proposers day" for the Semantic Forensics (SemaFor) program where researchers were driven to prevent viral spread of AI-manipulated media.

[261] DARPA and the Semantic Forensics Program were also working together to detect AI-manipulated media through efforts in training computers to utilize common sense, logical reasoning.