How a scandalous experiment with AI bots has put the future of democracy in serious question
Recently, scientists from the University of Zurich conducted what I think is a very interesting experiment.
They tried to find out to what extent "bad actors" can successfully use AI bots to persuade people in online discussions.
To do this, they created accounts for the bots on Reddit (the largest online forum platform). They were given a certain portrait and a story, and a rather peculiar one at that (one of the bots represented an imaginary black man who had been raped and opposed the popular Black Lives Matter movement).
They could not only see messages from other users and write replies. But also take into account other messages from these users and information about them (what the user writes in the about section), and this allowed the bots to create very targeted and personalized arguments to convince other users. And with this approach, they were very good at it .
It is clear that such bots are highly scalable entities. In this very limited experiment alone, they wrote thousands of comments.
The experiment itself caused a big scandal, because no one warned the Reddit administration about it, nor did they warn users. And such use of bots without approval is prohibited by their policies. They were so outraged that they are preparing official legal claims.
I don't know how it will end. But the experiment itself is undoubtedly successful and revealing.
The whole idea of democracy is based on a free and open discussion of important decisions for society, such as the election of officials, the adoption of certain important laws, etc .
And it's obvious, like 2+2, that artificial intelligence can quickly turn this structure upside down, both up and down.
And even by players not from this particular society.
In general, it's easy to imagine a group from Rashi, North Korea, or some other place creating thousands or tens of thousands or even millions of bots, flooding social media in a particular country with them. They systematically engage in hundreds of thousands of discussions. It accumulates the most effective arguments. Coordinates efforts. Studies interlocutors in dialogues, creates strong personal arguments.
And this is all against people who, for the most part, do not even realize that this can happen, and it is already commonplace.
On the other hand, it is also quite possible to imagine scenarios where democracy is significantly enhanced. When AI acts as a mushroom, a kind of intelligent environment in which society can crystallize its will and optimal solutions. In which it can be an effective moderator in the dialog of millions. And eventually, accordingly, to turn democracy from a largely illusory state, as it is now, into reality.
