ChatGPT has invaded the academic and professional spheres and is now attacking the political field: elected officials use it for speeches or laws and some dream of seeing it carry their own values, but the tool poses the threat of political campaigns. influence difficult to detect.
In Japan, a parliamentarian questioned the Prime Minister at the end of March with questions proposed by this conversational robot. In France, he drafted an amendment to the 2024 Olympic Games bill.
French President Emmanuel Macron himself recently mentioned the artificial intelligence of OpenAI on Twitter, posting a screenshot of an exchange with the chatbot which considers Europe “competitive” in the race for innovation .
The American technology behind ChatGPT is however not initially designed to make such a judgment because it only responds with the most plausible words when faced with a request, thus being able to hold a position as its opposite.
Mr. Macron himself has recently been associated with this AI: barely finished his last televised address, the general secretary of the left-wing CGT union, Sophie Binet, quipped by declaring that it “could have been made by ChatGPT”.
Politicians are now trying to seize the possibilities offered by this chatbot, which had more than 100 million active users at the start of the year, just two months after its release.
According to Pascal Marchand, professor of information sciences at the University of Toulouse, AIs like ChatGPT “are able to generate discourses that are very faithful” to the traditional ideological markers of policies.
But unable to innovate, they are less relevant for parties who wish to “adapt to the situation and have a discourse in tune with the times”, believes the researcher.
The more right-wing parties in politics also struggle with ChatGPT, which they deem “woke” and steeped in liberal and progressive Silicon Valley values.
In France, the president of the National Rally Jordan Bardella thus raises on social networks the specter of “another great replacement” by artificial intelligence, while Marion Maréchal, vice-president of Reconquête, another far-right party, claims the program harms “critical thinking”.
OpenAI, or its competitors like Bard developed by Google, do have biases, resulting from their training from a huge corpus of texts and the filters added by their designers to limit the generation of objectionable remarks.
In New Zealand, researcher David Rozado designed – but did not publish – the RightWingGPT robot, an AI trained to produce a conservative argument in support of traditional family, Christian values, or the free market.
The new boss of Twitter Elon Musk, an investor in OpenAI at the launch of the start-up, also said in an interview that he wanted to launch TruthGPT, an AI that is less “politically correct” than ChatGPT. The Chinese government has for its part enacted rules so that generative AI “reflects fundamental socialist values”.
“If someone develops a conversational robot that always goes in the same direction, it will be able to provide elements of language to convinced people but it will interest a lot fewer people”, judges Pascal Marchand, who considers that “it should not be too much fantasize about the mass manipulation that this medium could represent”.
The founder of the Belgian satirical site NordPresse, nicknamed Vincent “Flibustier” on social networks, tried his hand at it by creating the “Marxist AI” ChatCGT with his brother at the start of the year.
Its operation is “fairly basic”, he explains to AFP. She simply sends questions from Internet users to ChatGPT, telling her to answer “like an angry trade unionist would do”. Result, according to him, “it is an excellent generator of populist texts”.
04/22/2023 08:49:37 – Paris (AFP) – © 2023 AFP
