Deepfakes and Indian elections: A case of much ado about nothing

While India is simply a processing economy, it has a digitally- and societal media-savvy population. Political parties, too, person ample IT and societal media wings. Deepfakes are not new, with AI technologies similar Generative Adversarial Networks (GANs) churning them retired since 2011. 

However, the advent of GenAI and the ubiquity of societal media person made their creation, prime and dispersed astatine standard overmuch cheaper and easier. This explains the concerns expressed by civilian society, governments and media.

In a fewer articles and comments connected AI and deepfakes, I had taken a contrarian stance, arguing that AI, if utilized well, could assistance the electoral process and democracy. It tin assistance observe fraud, optimize the analyzable logistics of booth and elector management, and physique assets and outgo efficiencies into the immense machinery of Indian elections. 

Also read: Lok Sabha elections 2024: Amid rising deepfake content, EC directs governmental parties to region fake clips wrong 3 hours

GenAI tin beryllium utilized for politicians to scope retired to voters successful a much personalized and scalable manner, level the playing field, and marque voting entree for the differently-abled voters easier. Most, however, tended to absorption connected the negatives and the expected havoc that GenAI could wreak connected Indian elections.

Now that Lok Sabha polling is over, it seems to beryllium overmuch ado astir nothing. A station facto investigation done by 2 Harvard Kennedy School scholars Vandinika Shukla and Bruce Schneier (bit.ly/4er8ISi) concluded that AI was mostly utilized constructively, alternatively than the destructive usage that astir of america expected. 

They estimated that governmental parties spent $50 cardinal connected authorized AI-generated content, and utilized it for targeted connection aimed astatine their constituency’s voters. In Tamil Nadu, for instance, some Karunanidhi, with his trademark acheronian glasses, and Jayalalithaa were resurrected to entreaty to voters, and this was openly authorized by their respective governmental parties—an illustration of ‘deepfakes without deception’. 

Party workers astatine the lowest rung of the ladder frequented tiny tech companies that created personalized ‘deepfake’ videos of them. These could past beryllium distributed astatine scale. This created an unprecedented accidental for young techies, who acceptable up nimble outfits to service specified governmental parties. India has 25 authoritative languages and thousands of section dialects, and immoderate politicians leveraged the Bhashini AI level to dub their speeches. 

Voice clones of candidates made millions of calls to voters, explaining their promises. Even Prime Minister Narendra Modi got into the enactment erstwhile successful Tamil Nadu helium asked his assemblage to enactment connected earphones for his Hindi code to beryllium translated into Tamil successful existent time. Political workers utilized GenAI and different technologies to flood societal media with localized and contextualized memes.

Also read: Lok Sabha elections: AI deepfakes airs challenges to the integrity of electoral process. How to place fake videos?

It was not each a furniture of roses, though. There were galore instances of the usage of deepfakes to impersonate candidates spewing hatred and discord, saying things they had ne'er said, and making it hard for voters to recognize what was existent and what was fake. 

But, arsenic James Thornhlll writes successful Financial Times (bit.ly/3RBLgI8): “[It] could good beryllium that the expanding usage of AI tools by millions of users is itself deepening nationalist knowing of the technology, inoculating radical against deepfakes. The predetermination did not look to beryllium disfigured by the integer manipulation."

In a sense, the impervious was successful the pudding. There were much concerns astir the heatwave sidesplitting canvass workers and reducing elector turnouts than of fake videos. The result of the predetermination revealed little IT-savvy parties winning successful galore places. This showed that the omniscient Indian elector not lone saw done the rhetoric of politicians, but besides the fake quality that immoderate of them generated. 

In the end, the dire forecasts of doomsayers did not travel true, arsenic AI was utilized much for bully than bad. Harvard researchers Shukla and Steiner constituent retired that “the technology’s quality to nutrient non-consensual deepfakes of anyone tin marque it harder to archer information from fiction, but its consensual uses are apt to marque ideology much accessible."

Also read: OpenAI joins AI contented confederation arsenic Election Commission warns of deepfakes

James Thornhill wisely noted that “we should interest much astir politicians spouting authentic nonsense than fake AI avatars generating inauthentic gibberish." Indian voters this summertime seemed to agree.