Created last May, the Synthetic Party is entirely controlled by the artificial intelligence “Leader Lars”.
Like in a science fiction novel. Last May, a Danish artist collective called “Computer Lars” created the first political party run entirely by artificial intelligence (AI). Named “Leader Lars”, after the most popular first name in Denmark, it is powered by the Mindfuture Foundation. “Det Syntetiske Parti”, or the Synthetic Party, wants to involve citizens in a collective reflection. Its objective: “to connect AI and humans to understand what people really think of politics and go against the current of the dominant political discourse”, explains to Figaro Asker Bryld Staunaes, representative of the collective of artists . The party’s first public meeting will take place in September on the occasion of the Vestegnens Kulturuge cultural festival.
Still far from being able to enter the hemicycle of the Folketing – the Danish Parliament -, the Synthetic Party has only four signatures of members out of the 20,182 necessary. While waiting to garner more support, the few supporters are already mobilizing and conversing directly with the AI via Discord, an instant messaging software allowing you to speak with the “chatbot”, the conversational agent of the program. The party wants to be revolutionary in its way of conceiving legitimacy in politics and wants to shake up our representation of democracy. But for the expert Jean-Gabriel Ganascia, professor at the Faculty of Sciences of Sorbonne University and researcher at LIP6, this project has “a disturbing aftertaste with significant risks”. So, a political party managed by an artificial intelligence, progress or danger?
“Anyone can interact”
Concretely, “Leader Lars” is “a mixture of many different opinions”, explains Asker Bryld Staunaes. Through their interactions with the AI, users express their political ideas and influence the machine. At the same time, it also compiles all the political publications of marginal parties since 1970 in order to represent “all those who do not have the organizational means to run for Parliament”. Thus, the entire political spectrum is taken into account and opinions are dissolved. “There is a part of randomness, but it’s the only way to be in tune with what we represent,” he adds. For the Synthetic Party, if “everyone can interact”, then everyone can be represented, in particular the 15.46% of abstainers in the 2019 legislative elections.
“Leader Lars” can generate text but also images. It is through these processes that he can respond when asked a question. Thus, if the Party were to win seats in Parliament in June 2023, “human people will represent the AI on behalf of the Party and participate in all decision-making processes by conversing with the chatbot”, explains Asker Bryld Staunaes. For example, the AI has already proposed some measures including the creation of a universal income of 100,000 Danish kroner (13,440 euros), “which could not have been suggested by a human”, he says. Another suggests reorganizing the structure of parliament by making random appointments every month. For Jean-Gabriel Ganascia, “the ideas would be proposed like those of an oracle”, and that poses a problem.
Finally, what AI would lack the most would be one of the pillars of the policy: “knowing how to arbitrate between different elements”, analyzes Jean-Gabriel Ganascia. Indeed, by definition, AI has no reasoning ability and therefore “nothing allows to give the merits of its proposals”, he explains. Unlike the politician, she can neither debate nor justify her measure, in short, she has no argument. In another context, its mode of operation could be useful, but only to complement the decision-making process and not replace it. Removing the notion of debate would be “a mockery of the parliamentary system and that is what is worrying”, continues Jean-Gabriel Ganascia.
The limits of the process are also related to its reliability. How can you be sure of escaping pressure groups or militant collectives? “We can very well imagine that groups invest in the chatbot to divert specific pieces of information. Even if it can engage a certain number of people, it can quickly become unequal”, explains Professor Ganascia. In addition, the AI is very sensitive to language modulations. In short, “we can have a totally different answer depending on how we direct the question”.