The future of democracy ? An AI candidate in the British elections

© Neural Voice

After the Breton influencer who was actually just an AI, say hello to Steve, an AI candidate running for a seat in the British Parliament as an independent. Designed by the company Neural Voice, under the leadership of its director Steve Endacott, Steve is a political chatbot allowing voters in the Brighton Pavilion constituency to interact directly with him using an online voice interface . This system offers citizens the ability to submit questions relating to the positions defended by Steve Endacott, but also, even more remarkably, to propose their own ideas and reflections .

Endacott has the ambitious vision that his artificial candidate can transcribe and synthesize these fruitful exchanges with the electorate, in order to subsequently forge public policies directly inspired by these citizen interactions. As the entrepreneur points out in the columns of Wired: “ We are reinventing politics by using & #8217;artificial intelligence as a technological base, not to replace elected officials, but to truly connect them to their audience, to their constituency ”. One more step towards anything ?

Between laws and reality: the obstacles of AI in politics

Governing AI through a legal framework is no easy task, even less so in politics. Although the initial enthusiasm generated by Steve's unique candidacy cannot be eclipsed, many obstacles still stand in his way.

Firstly, legal vagueness persists as to the very legality of presenting software in elections, while the real capacity of artificial intelligence to assume the daily responsibilities of an elected official, as close as possible to human realities, remains a major uncertainty. Furthermore, if Steve proves capable of engaging in up to 10,000 simultaneous conversations, a physical limit obviously prevents him from appearing in person in the voting booth .

Faced with this constraint, Endacott plans to take on this role on an interim basis, committing to scrupulously following the political decisions coming from its chatbot, even if they diverge from its own personal convictions. An approach which, although attractive on paper, does not fail to raise legitimate questions in terms of reliability and ethics. Indeed, the risk that AI produces false information or reproduces harmful biases; two recurring pitfalls with current generative models; flat filigree.

Democratic progress or simple sleight of hand ?

If some voices are raised to praise the potential of artificial intelligence to reinvigorate the mysteries of democracy, others do not fail to express serious reservations about the risks inherent in such innovation. It is true that, conceptually, AI could prove valuable in quickly analyzing and synthesizing large volumes of government and public policy documentation, thereby providing more rational prescriptions for decision-makers. However, it is clear that current models, such as ChatGPT or Google Gemini, still suffer from damaging flaws like those mentioned previously.

Therefore, a legitimate question emerges: will voters really be reassured by the prospect of an artificial candidate ? Some recent incidents, such as New York Mayor Eric Adams' chatbot advising merchants to violate local regulations, illustrate how trust in these systems remains fragile.

Moreover, a recent survey conducted jointly by the University of Chicago and the Associated Press reveals that 58% of American adults fear that the #8217;AI will not contribute to disinformation before the 2024 presidential elections. Another Pew Research study highlights that the majority of Americans say they are more concerned than enthusiastic about it’ 8217;increasing use of AI in their daily lives.

As promising as it may seem, the Steve project does not cross the limits of reason and ethics? An excess of audacity which could well be akin to an insidious form of technochauvinism< /strong>, obscuring the risks inherent in a technology whose control still largely escapes us. Furthermore, the recurring flaws and deviations of current models cannot be avoided under the pretext of a possible revitalization of democracy.