Is AI already misleading Indian voters? Report Nyt
In India there is a veritable factory of avatars of local politicians which, thanks to artificial intelligence (AI), are credible, as well as being economical. They speak all the different languages of the country and address voters by name. How and to what extent will they influence the elections? The New York Times article
To get an idea of the direction that artificial intelligence is taking in election campaigns, just look at India, the largest democracy in the world, which will go to the polls on Friday – writes the New York Times .
THE AVATARS OF THE MODI PARTY, CREDIBLE AND CHEAP
An AI-generated version of Prime Minister Narendra Modi, shared on WhatsApp, shows the possibilities of hyper-personalized engagement in a country with nearly a billion voters. In the video – a demonstrative clip whose source is unclear – Modi's avatar directly addresses a series of voters, calling them by name.
However, it's not perfect. Modi appears to be wearing two different pairs of glasses and some parts of the video are pixelated.
Modi's party is sending videos via WhatsApp in which its AI avatars deliver personal messages to specific voters about government benefits they have received and ask for their vote.
These video messages can be automatically generated in any of the dozens of Indian languages spoken by the voter. The same goes for phone messages from AI-powered chatbots that call voters in the voices of political leaders and ask for their support.
This type of communication requires a fraction of the time and money spent on traditional election campaigns and has the potential to become an essential tool in elections.
DISINFORMATION ON DISINFORMATION
However, as the technology enters the political scene, there are few guardrails preventing its misuse. Chatbots and personalized videos may seem more or less harmless. But experts fear that as technology advances and becomes more widespread, voters will have more difficulty distinguishing between real and artificial messages.
“This year is going to be the Wild West and an unregulated AI space,” said Prateek Waghre, executive director of the Internet Freedom Foundation, a digital rights group based in New Delhi. “Technology – he added – is entering a media landscape already polluted by disinformation”.
Around the world, elections have become a testing ground for the AI boom. The tools were used to transform an Argentine presidential candidate into Indiana Jones and a Ghostbuster. During the New Hampshire primary, voters received robocall messages urging them not to vote, in a voice that was most likely artificially generated to resemble President Biden's.
In India, Modi's Bharatiya Janata Party (BJP) and the opposition Indian National Congress have accused each other of spreading election-related deepfake content online.
POLITICIANS CRAZY ABOUT AI
An outpost of this new Indian frontier lies in the western desert state of Rajasthan. On the ground floor of a residential building in a dusty alley, a 31-year-old college dropout, Divyendra Singh Jadoun, runs an artificial intelligence startup, The Indian Deepfaker.
Its team of nine made commercials with AI-generated avatars of Bollywood actors and actresses. But earlier this year, parties and politicians began asking him to do for them what he had done for celebrities. Of the 200 requests, Jadoun said he accepted 14.
Among those who received the treatment is Shakti Singh Rathore, a 33-year-old BJP member. His job this election season is to inform as many people as possible about Modi's programs and policies. So he decided to create a replica of himself.
“AI is wonderful and the way forward,” Rathore said as he settled in front of a video camera in The Indian Deepfaker's office, preparing to incarnate digitally. “How else could I reach the beneficiaries of Modi's programs in such a large number and in such a short period of time?”
As Rathore adjusted a saffron scarf with the party logo hanging around his neck, Jadoun told him: “Look into the camera and speak as if the person is sitting in front of you.”
THE INDIAN FACTORY OF AVATAR IA
With about 5 minutes of material, including an audio recording and profile photos, Jadoun got to work. He said he uses open-source AI systems and builds on them with his own code. First, Rathore's face was isolated from each frame of the recording. Then data relating to his somatic features were collected, including the size of his face and lips and his gaze.
Jadoun said the data set was then fed into artificial intelligence models that learn to predict facial patterns. “You need to keep putting them through the program and fine-tuning the face until you get the best face possible,” he said.
A “cloning algorithm” also analyzes the audio recording, learning the cadence and intonations of the voice. Jadoun said it often takes 6 to 8 hours of editing to perfect the face and lip sync to the words. The rest is largely automated.
In one demonstration, it took about 4 minutes to create about 20 personalized greeting videos.
Jadoun said his team can produce up to 10,000 videos a day. For larger, time-sensitive jobs, he will rent graphics processing units.
AN AI THAT REACHES EVERYONE
Generative AI can also eliminate language barriers, which is especially useful in a linguistically diverse country. Rathore's avatar can be programmed to speak regional languages and reach the remotest corners of India.
Political parties don't just send video messages to voters, they also use cloned voices to call people directly, thanks to chatbots like ChatGpt.
In the past, when a party representative called voters, they hung up, Rathore said. “But now, when a local leader says a voter's name, it immediately captures their attention.”
During the conversation, the chatbot asks about local government programs that offer free electricity or funding for startups. Jadoun said the calls were recorded and transcribed for quality control and AI training.
Rathore said he spent about $24,000 of his own money to reach about 1.2 million people through his video messages and phone calls and to receive information about those who did not respond. He called it an investment in his future with the BJP.
THE IMPORTANCE OF CUSTOMIZATION
Nikhil Pahwa, editor of MediaNama , which covers digital media in India, said personalized messages could be particularly powerful among Indians.
“India is a country where people love to have their photos taken with celebrity lookalikes,” he said. “So if they get a phone call from, say, the prime minister, talking as if he knows them, where they live and what their problems are, they would be really excited about it.”
Waghre of the Internet Freedom Foundation doubts that AI content is persuasive enough to influence this year's elections. But he said the long-term effects could be problematic. “Once they become normalized in people's information diet, what will happen six months later when there are deceptive videos?” he said.
Modi himself has discussed adding disclaimers for AI-generated content, so that people are not “misled.” Jadoun and representatives from two other AI start-ups in India have created what they call an “AI coalition manifesto,” pledging to protect data privacy and uphold election integrity. For example, videos from India's Deepfaker are labeled "AI-generated" and its chatbots advertise that they are AI-generated voices, Jadoun said.
Narendra Singh Bhati, 28, a resort owner in Rajasthan, received an AI-generated call from Rathore this week. Bhati said he was impressed by its customization and didn't realize the call was AI-generated, even though the script clearly said so. “I even said goodbye to Mr. Rathore at the end,” Bhati said.
(Excerpt from the foreign press review edited by eprcomunicazione )
This is a machine translation from Italian language of a post published on Start Magazine at the URL https://www.startmag.it/mondo/lia-sta-gia-intortando-gli-elettori-indiani/ on Sat, 27 Apr 2024 06:57:22 +0000.