When Artificial Intelligence Meets a "Super Election Year"
Billions of voters in Russia, India, the United States and dozens of other countries will vote in 2024, which may be the largest "super election year" in the world's history. At the same time, artificial intelligence technology has ushered in leaps and bounds, from generating text and pictures to synthesizing sound and video.
When elections meet artificial intelligence, candidates on the electoral stage of various countries have added "new weapons" in their hands. Candidates use AI tools to analyze voter preferences, write speeches, design social media presences, and more. What's more, AI technology is being used to create fake audio recordings, images, and videos, becoming accomplices in smearing, attacking, or manipulating behavior.
Similar cases have emerged, when some voters received calls from "US President Joe Biden" before voting in the Democratic primary in January this year, and subsequent preliminary investigations revealed that the voices in these calls were generated by artificial intelligence. In last year's Slovak National Assembly elections, audio of what appeared to be a candidate's voice was alleged to have been processed by artificial intelligence. Many experts say that countries' electoral processes are at risk of being severely disrupted by artificial intelligence.
On February 16, 20 tech companies signed an agreement in Munich, Germany, to voluntarily take steps to prevent artificial intelligence from being used to disrupt elections around the world. However, observers are skeptical about the effectiveness of "self-regulation" agreements in the tech industry. Governments have been accused of not keeping pace with technological developments.
Fake "Biden" called voters to block the vote
In January of this year, Gail Huntley, a 73-year-old resident of New Hampshire, received a phone call from the earpiece with what appeared to be Biden's hoarse voice coming from the earpiece. Huntley had planned to cast a vote for Biden in the Jan. 23 Democratic New Hampshire primary, but "Biden" on the phone urged her not to.
Such a vote would only allow the Republican Party's plan to re-elect Trump. "Biden" said "on the phone.
"That's nonsense. Huntley quickly realized that the call was fake, but at first she thought it was just pre-synthesized audio. She was shocked and confused when she found out that "Biden" on the phone could have a conversation with people.
In New Hampshire, Huntley wasn't the only one who received a call from "Biden." Preliminary investigations by New Hampshire authorities believe the calls used artificial intelligence technology to mimic Biden's voice in order to prevent people from voting in the state's Democratic primary. The state is still conducting an in-depth investigation into the matter.
On January 23, 2024 local time, the Republican primary for the 2024 presidential election in the United States began in New Hampshire in the northeast, and the people participated in the vote. Photo/IC photo
A few weeks later, the U.S. federal government announced a ban on automated phone calls using AI-generated voices. On February 8, local time, the Federal Communications Commission (FCC) of the United States passed a unanimous ruling that according to the "Telephone Consumer Protection Act" promulgated in the United States in 1991, which "restricts the use of manual and pre-recorded voice messages for harassing calls", the use of artificial intelligence voice cloning tools to make automated phone calls is prohibited, and violators may face heavy fines.
On February 25, local time, Steve Kramer, a political adviser to Congressman Dean Phillips, who was employed by the Democratic presidential candidate of the United States and Congressman Dean Phillips, admitted that he used artificial intelligence technology to concoct a call from Biden's voice to New Hampshire Democratic primary voters at a low price. Kramer did not admit that the fake phone calls were instructed by Phillips and appeared to be deliberately portraying his actions as "righteous acts" aimed at alerting the public to misleading applications of artificial intelligence.
U.S. media commented that "Biden" is the first artificial intelligence test encountered by the U.S. government, technology companies and civil society. In the age of artificial intelligence, anyone can create images that can be fake, or clone other people's voices with striking similarity. Next, dozens of countries such as Russia, India, and the United States will have elections, and billions of voters are expected to vote. Experts are concerned that the election is at risk of being severely disrupted by artificial intelligence.
In fact, in last year's Slovak National Assembly elections and in Indonesia's presidential elections in February this year, AI has been involved in the electoral process in different ways.
"It seems to be something from the distant future, but in fact the threat is already there. "Federal Communications Commission Chairwoman Jessica Rosenworther said that everyone can be a victim of AI disinformation.
American media commentator Timothy Carr recently wrote in an article that technology companies play a huge role in shaping democratic participation activities. "As we saw in the 2020 [US] election, disinformation on social media can have dangerous real-world consequences. Carr believes that as AI tools become more widely used, disinformation on social media will increase.
Deceased Indian politicians are "resurrected" to support current politicians
In last year's Slovak National Assembly elections, the rivalry between two main prime minister's candidates, Fizo and Himecka, came under scrutiny. During the media and candidates' silence ahead of polling day, an audio clip of what appeared to be a private conversation by Shimechka suddenly circulated on social media. In the audio, a voice similar to that of Himechka recounts a plan to rig elections by bribing votes from marginalized people.
Shimechka's campaign has denied the authenticity of the audio. European media reported that the audio showed signs of artificial intelligence processing. The report also said that the social media in question usually have restrictions on AI-generated content, but the publisher apparently found a loophole in the platform.
In India, deceased celebrities and politicians have been "resurrected" by AI and used in propaganda campaigns for current politicians. Muthuvel Karunanidhi was a late Indian writer and politician who served as Tamil Nadu's chief minister for nearly two decades during five terms from 1969 to 2011 before his death in 2018.
In the past six months, the image of Karunanidi, generated by a local technology company using artificial intelligence technology, has appeared frequently on different occasions. At a political rally in Tamil Nadu, an AI-generated image of Karunanidi appeared on the big screen in praise of his son, Tamil Nadu's Chief Minister M. Murphy. K. Stalin's leadership.
Senthil Nayagam, an Indian developer, said his company has used open-source software to generate the images and voices of 45 current and former Indian political leaders, including Mahatma Gandhi and Prime Minister Narendra Modi. Nayagam believes that the "resurrection" of deceased politicians with artificial intelligence is only out of nostalgia. But local observers are concerned that India's political parties may use artificial intelligence to launch their party's popular former leaders in order to boost their support.
The Massachusetts Institute of Technology Review recently published an article saying that there is still a lot of room for imagination on the issue of artificial intelligence that could threaten elections. "The real danger may not be what we have seen, but what we cannot yet imagine. The article said.
Many governments have been accused of being slow to make progress on AI regulatory measures
In the face of the rapid development of artificial intelligence, regulators seem to be struggling to keep up with the speed. U.S. senators and deputies have reportedly introduced dozens of bills on AI regulation, but none of them have yet reached the poll. The European Parliament has passed the most comprehensive AI regulation bill in history, but it is not expected to come into force until 2026.
The United Kingdom has been criticised for its slow action on AI regulation. On 6 February, the UK government announced that it had taken an important step towards its AI regulation by presenting "preliminary ideas related to binding regulation" and discussing it with technical, legal and social experts. The British press sarcastically said that the British government had taken an important step "quickly and decisively" – that is, "standing still".
Katie Harbas, an expert on U.S. technology policy, said that the U.S. government's actions in the field of artificial intelligence regulation are far from enough, and the government may be worried that regulation will have a certain effect on inhibiting innovation. In the short term, she said, the onus on regulating AI-generated content will fall on the companies that develop these tools.
On February 16, 20 tech companies, including OpenAI, Google and Microsoft, signed an agreement at the Munich Security Conference in Germany to voluntarily take "reasonable precautions" to prevent AI from being used to disrupt elections around the world.
February 18, 2022 local time, Munich, Germany, Munich Security Conference venue. Photo/IC photo
The agreement says the parties agree to take more appropriate measures to flag and limit AI-generated disinformation that spreads online, including audio, video, and images that fake the appearance, voice, or behavior of candidates and other key stakeholders. The companies also pledged to share more information about "how voters can protect themselves from being manipulated or deceived."
However, some analysts believe that while the agreements are a good thing, how they will be implemented is another matter. Others called the signing of the "self-regulation" agreement by 20 tech companies a "public relations exercise".
An opinion piece published in the US media said that the agreement, signed by tech companies, does not explicitly prohibit the use of artificial intelligence to fake content, but only flags and restricts. In addition, even in the past without artificial intelligence, tech companies have failed to stop the spread of disinformation about elections. Therefore, the authors of the article are skeptical about the validity of the agreement.
According to a non-profit organization in the United States, in recent years, American technology companies Google, Meta and X have laid off a total of about 40,000 employees, among which the content moderation category has laid off a large number of employees. The group warns that social media is particularly vulnerable to exploitation by bad actors using AI tools as tech companies lay off workers.
On October 30, 2023 local time, Washington, USA, U.S. President Joe Biden signed an executive order on artificial intelligence, establishing security and privacy protection standards for AI and requiring developers to conduct security tests on new AI models. Photo/IC photo
Candidates use AI tools to analyze voters
However, at least in the eyes of some, the development of artificial intelligence is not a bad thing for this global "super election year".
In Indonesia's election event in February, MP candidate Farah Putri Naria turned to artificial intelligence to understand her constituents when she went to her constituency to canvass for votes.
According to local media reports, an artificial intelligence software that is widely used in Indonesia's elections generates complex scenarios about the lives of voters in a region by sifting through big data — from official censuses to sentiment analysis of local social media groups, as well as data on commodity prices and polls.
Naria said there are 3 million voters in her district, and the use of artificial intelligence to analyze voters will be a priority. "[Here] how many people have completed their studies, how many are malnourished, what job opportunities are available, what they are complaining about on social media, and how each polling station has performed in previous elections. AI has really helped us simplify everything. She said.
In addition, in this election, Indonesian candidates not only used artificial intelligence to analyze data, but also used it to write speeches, generate social media posts, and design models for campaign memorabilia.
Yosse Rizar, a political adviser in Indonesia, told local media that while AI technology could be misused to spread fake news, it also has benefits. AI is like "your personal political advisor", and the detailed analysis, strategic planning, and copywriting that previously required a team to complete can now be done by AI tools.
According to Rizar's incomplete statistics, about 700 candidates in various categories used AI tools in Indonesia's elections. The election elects Indonesia's new president and vice president, members of parliament, members of the council of local representatives, and members of local councils.
On concerns that AI-generated disinformation could undermine elections, some experts said that exaggerating the risks could make voters distrustful of all election information, which could adversely affect election activity.
"I'm more worried about the narrative around AI than about the application of AI. If the narrative is exaggerated, people feel that the phenomenon of AI-generated disinformation is worse than it actually is, and that alone can reduce the level of trust that little trust people have in the information environment, Habas said.
Beijing News reporter Chen Yikai
Edited by Zhang Lei and proofread by Liu Yue