To get a sense of where artificial intelligence is heading in election campaigns, look to voting starting on Friday in India, the world’s largest democracy.
An AI-generated version of Prime Minister Narendra Modi shared on WhatsApp shows the possibilities for hyper-personalized outreach in a country with nearly a billion voters. In this demo clip from the video (which is of unknown origin), an avatar of Modi directly addresses a series of voters by name.
However, it’s not perfect. Modi appeared to be wearing two different pairs of glasses, and parts of the video appeared pixelated.
Down below, Modi’s party workers sent videos via WhatsApp in which their own AI avatars sent personal messages to specific voters describing the government benefits they received and asking them to vote.
These video messages can be automatically generated in any of the dozens of languages spoken by Indian voters. AI-powered chatbots can also leave messages on the phone, calling voters in the voices of political leaders and seeking their support.
This kind of outreach takes a fraction of the time and money of a traditional campaign and has the potential to be an important tool in elections. But as the technology enters the political arena, there are few safeguards to prevent misuse.
Chatbots and personalized videos may seem more or less innocuous.However, experts worry voters will face increasingly difficult times Distinguish between real and synthetic messages As technology advances and spreads.
“This year is going to be the Wild West and unregulated AI,” said Prateek Waghre, executive director of the Internet Freedom Foundation, a New Delhi-based digital rights group. He added that the technology is entering a media landscape already polluted by misinformation.
Around the world, elections have become proving grounds for the artificial intelligence boom.These tools have been used to turn argentine presidential candidate Indiana Jones and the Ghostbusters.During the New Hampshire primary, voters received Automated call messages The voice urging them not to vote was likely artificially generated to sound like President Biden’s.
And in India, Mr Modi’s Bharatiya Janata Party (BJP), and the opposition indian national congress party Each accuses the other of spreading election-related deepfakes online.
One of India’s new frontier outposts is in the western desert state of Rajasthan. On the ground floor of a residential building in a dusty back lane, 31-year-old college dropout Divyendra Singh Jadoun runs an artificial intelligence startup, The Indian Deepfaker.
His team of nine has been creating ads using AI-generated avatars of Bollywood actors and actresses. But earlier this year, political parties and politicians began asking him to do for them what he had done for celebrities. Mr. Jadoun said that of the 200 requests, he accepted 14.
Shakti Singh Rathore, a 33-year-old member of the Bharatiya Janata Party, is one of those who has received AI treatment. His job this election season is to educate as many people as possible about Mr. Modi’s plans and policies. So he decided to create a replica of himself.
“Artificial intelligence is wonderful and the way forward,” Mr. Rathore said in front of a camera at Deepfaker India’s offices, preparing to become a digital avatar. “How else could I reach such a large number of beneficiaries of Mr. Modi’s schemes in such a short period of time?”
As Rathore adjusted the saffron scarf with the party logo hanging around his neck, Jadoon instructed him, “Just look at the camera and speak as if this person is sitting right in front of you.”
Armed with about five minutes of material, including audio recordings and a profile photo, Mr. Jardoun got to work. He said he uses open source AI systems and builds them with his own code.
First, Mr. Rathore’s face is isolated from every frame of the recording.
Data was then collected from his facial features, including the size of his face and lips, as well as his gaze.
The data set was then fed into an artificial intelligence model that learned to predict facial patterns, Jadoun said.
“You need to keep running the program and fine-tuning the face until you get the best face possible,” he says.
The “cloning algorithm” also analyzes the recordings, learning the rhythm and intonation of the voice. It usually takes six to eight hours of adjustments to perfect the face and synchronize the lips with the speech, Mr. Jardoun said. Most of the rest is automated.
In one demo, it took about four minutes to create about 20 personalized greeting videos.
Mr. Jadoun said his team can produce up to 10,000 videos a day. For larger jobs that are completed before deadline, it will rent graphics processing units.
Generative AI can also eliminate language barriers, which is especially useful in countries with diverse languages. Mr. Rathore’s avatar can be programmed to speak regional languages, thereby reaching the remotest corners of India.
Political parties are not only sending video messages to voters, they are also calling people directly using cloned voices, all powered by chatbots like ChatGPT.
Rathore said that in the past, when party representatives called voters, they would hang up. “But now, when local leaders say their constituents’ names, it immediately gets their attention.”
During the conversation, the chatbot asked the local government about its plans to provide free electricity or funding to startups. Mr Jadoun said the calls were recorded and transcribed for quality control and AI training.
Mr. Rathore said he spent about $24,000 on his own to contact about 1.2 million people via video messages and phone calls and received information about who didn’t answer.He calls it an investment in his future in the BJP
Nikhil Pahwa, editor of MediaNama, which covers digital media in India, said personalized messages could be particularly influential among Indians.
“India is a country where people love to take pictures with celebrity impersonators,” he said. “So if they get a call from someone like the prime minister and he talks like he knows them and where they live and what their problems are, they’re actually going to be excited about it.”
Mr. Wagel of the Internet Freedom Foundation questioned whether AI content would be persuasive enough to influence this year’s election. But he said the long-term effects could be problematic. “Once you incorporate this into people’s information diet, what happens six months later when deceptive videos come out?” he said.
Mr Modi himself has discussed adding disclaimers to AI-generated content so people don’t “misled”. Mr. Jadoun and representatives from two other Indian artificial intelligence startups created what they call “AI Alliance Declaration”, pledging to protect data privacy and uphold election integrity. For example, Jadoun said, Indian Deepfaker videos were labeled as “AI-generated,” with chatbots declaring they were AI-generated voices.
Narendra Singh Bhati, 28, a resort owner in Rajasthan, received an AI call from Mr Rathore this week. Mr Bhatti said he was impressed by the personalized service.
He said he didn’t realize the call was generated by artificial intelligence, even though the script made that clear. In the end, “I even said goodbye to Mr. Rathore,” Bhatti said.
