If you’re worried about artificial intelligence bots breaking free from human control and taking over the world, shield your eyes now.
Prime Minister Nicolae Ciuca claims the bot, which calls itself “Ion,” is capable of interpreting the opinions of the country’s population and conveying them back to him and his government, helping them choose how to make decisions.
“Ion will do, through artificial intelligence, what no human can: listen to all Romanians and represent them before the government of Romania,” Ciuca said.
“Hello. You gave me life. I am Ion. Now, my role is to represent you. Like a mirror,” the machine said, when asked by Ciuca to introduce itself at a public meeting broadcast by Romanian television Wednesday.
Officials say the bot will help ministers formulate policies that are more in touch with voters’ everyday concerns, and will one day even be able to propose its own original ideas.
According to Romania’s minister of research, the bot uses artificial intelligence and natural language processing to automatically identify the opinions of Romanians that have been shared on social media or submitted to a newly established portal. The machine synthesizes their thoughts into categories, prioritizes their importance and delivers the information back to government decision-makers.
“Ion is the first AI governmental adviser — we think — in the world,” said Sebastian Burduja, Romania’s minister of research, in a telephone interview Thursday. The bot, which was developed by his ministry, is in its “learning phase,” he said, inviting Romanians to teach it about their everyday concerns by posting about them on social media. “Ion will take all that information and will start to report results,” said Burduja, “which will be meant for the prime minister and designated members of the cabinet.”
The government believes that the tool — which will automatically trawl through social media to collect opinions — will allow it to make decisions more efficiently.
Others are more skeptical, raising ethical concerns that the algorithm, which identifies which concerns on social media should be prioritized for decision-making, could itself be biased — resulting in biased decision-making.
“Romanians should be informed and explained how this AI tool selects important posts, and on what criteria. This should explained to the public,” Kris Shrishak, a technologist at the Irish Council for Civil Liberties, said in an interview Thursday. He pointed out that the tool might also elevate the concerns of citizens who are most active on social media, which would not necessarily make it representative.
Eventually, Ion’s engineers hope to teach it to proactively formulate policy proposals for government ministers to consider, a prospect that adds fuel to an emerging debate over the role of artificial intelligence in everyday life — and in democratic processes in particular.
As well as providing reports to politicians in writing, the bot speaks to them, too. It has a deep, male-like voice — although as Romanians send their thoughts and concerns to the machine through voice notes, officials hope that voice will evolve. Burduja said Ion’s AI software allows it to integrate thousands of audio submissions into a single Romanian accent — the basis for Ion’s final voice. “Through AI, a single voice of the nation will be heard,” Burduja said.
PM @NicolaeCiuca: As of today, the Government of 🇷🇴 Romania has the first government adviser running on AI, a good example of emerging technologies in public service.— GuvernulRomâniei (@GuvernulRo) March 1, 2023
Congratulations to the 🇷🇴 researchers on this project that brings us closer to citizens! pic.twitter.com/yP5b5pOs5i
Bucharest is heralding the technology as a democratic innovation allowing Romanians to — both knowingly and unknowingly — influence decision-making at the highest levels of government, without having to wait for elections.
The AI allows the government to synthesize huge amounts of data about people’s concerns in real-time, something officials believe is not currently possible without the technology.
“Romanians can actually make their voices heard every day, and through artificial intelligence that voice can reach decision-makers in the best possible forum. It can inform better decisions ... that are more appropriate for the needs of average Romanians,” Burduja said. Eventually, he hopes, it will be able to propose policies of its own — based on the submissions.
Romanians have so far responded to the call with a flood of social media complaints about everyday life, including concerns over corruption, high inflation, and even the government’s pension policy. One Twitter user asked the bot for its opinion on why some 40-year-olds were able to retire and receive a pension.
Many others responded to the bot negatively. “Read what the Romanians are saying here and tell your boss how much THEY love Prime Minister Dr. Ciuca and those in the government,” suggested one user — probably sarcastically — addressing the bot on Twitter.
As the use of AI technology to complete everyday functions grows more widespread, governments are racing to establish ethical protocols for their use in public sector decision-making processes — if permitted at all. One concern is that the algorithms underpinning the machines could start to reflect existing human biases, meaning that their outputs would also be tainted by bias. Another worry is that they could be vulnerable to hacking or spamming attacks.
Ion, Romanian authorities say, will be protected from spamming using “proof of human” technology. Burduja said the inbuilt technology will allow the machine to filter out automatically generated content, preventing it from inadvertently speaking to other machines and using the results to inform government policy.
Last month, British media reported that civil servants in London had been instructed by their government employers not to use AI chatbots while formulating government policies or sending emails at all. It followed the publication of guidance from the British government’s Office for Artificial Intelligence — setting out recommendations for public sector workers considering incorporating AI systems into the workflows. “The potential uses for AI in the public sector are significant, but have to be balanced with ethical, fairness and safety considerations,” the guide states.
“Make sure your model is fair, explainable, and you have a process for monitoring unexpected or biased outputs,” the official guidance goes on to suggest. It also recommends public officials hire an ethicist “to provide ethical judgments and assessments on the AI model’s inputs,” to ensure that the model does not produce biased policy, and that the algorithm is performing in line with “ethical considerations.”
Amar Nadhir contributed reporting.