I'm wondering how it goes with a lot of social media actually.

There were problems with the "algorithm"[0] before being intentionally gamed... but I feel like it's easier than ever to have an army of bots intentionally pointing a conversation.

I'll give you a brief example (we should not debate this case on HN but I'm talking specifically about a hypothesis that can be drawn).

In the UK, the streets are alight with the news of the murder of a young white boy at the hands of a Sikh boy. In the UK, historically, the Sikh community has been seen as a bit of an outlier in terms of how well they integrated and how tolerant they are. They are, for all intents and purposes "model immigrants".

The issue was, when the police were called to the scene of the stabbing, they immediately arrested the barely conscious boy on the ground, who had been stabbed and was bleeding from the mouth- handcuffing him and dismissing his pleas on the basis that he might have been making racist remarks.

The people in my circles are furious at the police for doing this.

However, it seems like there is an army of bots who keep trying to paint it as if it's a problem with Sikhs, or that their religious rights were the problem.

(for context, British law permits baptised Sikhs to carry a dull Kirpan knife - but this knife was not used for the incident above, it was a separate and still illegal knife to carry).

So now, online, instead of people talking about police indifference or the issues we have when we see race before trying to assess the situation: we are instead caught in a conversation about banning Kirpans for Sikhs (they have never been used in any crime from what I can tell) or how the Sikhs should feel bad or whatever.

Which is wonderful for a politician who wants to dismiss the criticism against the police as racism, but not so good for holding the police and media to account for the situation and the attempted coverup.

By being able to control the aggression, they are able to control the response, which drowns everything else out, and with LLMs the barrier is stupidly low to do this.

Creating division for your own reasons has never been cheaper or easily accessible.

[0]: https://en.wikipedia.org/wiki/Cambridge_Analytica#Data_scand...