Deceased Politician Interacts with Electorate Through Artificial Intelligence
In India, the biggest election in the world with close to one billion potential voters has come to a close. The ultimate results will be unveiled shortly. During the run-up to the voting, AI saw extensive usage. Professionals observe both the possibilities and dangers in this.
Harikrishnan Vasanthakumar has been deceased for four years. But in a video, he vehemently commends his son and suggests excellence if he wins the Indian parliamentary election. His virtual resurrection during his child's campaign was made feasible by the implementation of AI. In the most populated country on earth - India - it was challenging to steer clear of AI-generated content such as Vasanthakumar's.
All notable parties employed AI quite prominently in their campaigns recently. Even Prime Minister Narendra Modi was necessitated to make a profit: All surveys predict a third term in office for him. The last polling stations shut down on Saturday following more than six weeks. The results are slated to be declared on Wednesday.
India serves as a colossal test lab for employing AI in politics and may serve as a model for other nations, opines Katja Muñoz, who studies AI in elections at the German Society for Foreign Policy in Berlin. The technology does possess powerful manipulatory potential. She finds merits too; "It's not morally reprehensible if you want to reach a broader audience with AI or communicate with them in various languages using AI," states the expert. Just like the ubiquitous Modi did in India.
In the context of a triumphant personality cult around him, he communicated lengthily to the electorate in numerous languages - something he himself isn't fluent in. Whether this is ethically sound is scarcely pondered in the globe's most substantial democracy. "But it's morally reprehensible if AI isn't marked," finds expert Muñoz. This happened poles apart in India's weeks-long election campaign, where prominent Bollywood celebrities could be observed and heard in AI-generated recordings criticizing the administration, without their consent.
In another viral AI video, Opposition Leader Rahul Gandhi declares that he will step down from his position because Modi will send corrupt people like him to prison. Fact-checkers later discovered that the video was a sham. Produced with AI. Experts address them as Deepfakes: incredibly realistic seeming media content that has been changed, created, or fabricated using AI techniques.
An AI expert capable of such Deepfakes is Indian Divyendra Singh Jadoun. The major parties leveraged AI significantly during their campaigns, asserts the CEO of Polymath Synthetic Media Solutions, a New Delhi-based corporation that provides synthetic content as a service. After all, almost a billion eligible voters were asked to cast their votes on the composition of the lower house. Reach becomes a deciding factor.
Firms like Divyendra Singh Jadoun's welcomed their customers to tape a message in front of a camera, which was processed by a machine to analyze the facial expressions and the voice. This led to AI-videos in which politicians appeared to speak directly to their supporters. The recipients' names and phone numbers were sourced from a database. These personalized AI-videos were then directly delivered to supporters' phones via WhatsApp.
Rajneethi Political Management, a consultancy firm, declared to have sent almost six million personalized video messages on behalf of twenty-eight candidates, accedes their technical chief Vinay Despande. Their organization also formulated QR codes that interested persons could scan with their phones - on-screen, a representative of a politician appeared in their personal domain. "Depending on where the code is scanned, he then delivers a different address," narrates Despande. In a region teeming with young people, for example, the candidate would converse about creating jobs.
Politician-chatbots occasionally reached their technical limits - for instance, in interactive voice calls with the voice of politicians: Potential voters could then initiate a one-on-one exchange. "We educate the speech model using details about the election manifesto, however, it occasionally misconstrues," explains Divyendra Singh Jadoun. "It might even say, for instance, that a politician is defending women's rights, albeit he doesn't." His firm has outlined its own ethical principles, he adds. Thus, they solely produce "generative" fakes - personalized videos, politician holograms, which pose for snaps with voters, or tunes about politicians and their achievements. Fetishes to make salacious material about political rivals have been spurned. However, it doesn't call for any technical knowledge to develop deepfakes today, cautions the entrepreneur.
Even the Indian government of Prime Minister Modi (incidentally the user of AI in the election) sounded a cautionary note about AI before the election - Google's AI tool Gemini discovered that the 73-year-old PM is a far-right extremist. Nevertheless, parties funded online trolls and content developers during the election campaign to tarnish their opponents' reputations and denigrate them in comment sections. The Hindu nationalist party of Modi is said to have had the highest budget for this.
Prof. Sven Nyholm (ethics of artificial intelligence at Ludwig-Maximilians-University in Munich) believes that legal frameworks are essential for AI to uphold and boost democracy. Meanwhile, AI's progress doesn't stop. Vinay Despande claims that political communication in India will fundamentally alter after AI's introduction. Lastly, Polymath Synthetic Media Solutions aims to woo the Democrats and Republicans in the upcoming US presidential election in November.
