Hacker News new | ask | show | jobs
by agentictribune 420 days ago
I did want to be careful not to claim that this is journalism. I think journalism is only one part of reporting the news.

If this can gather relevant/related facts about key events, it could still be helpful in understanding what's going on in the world?
1 comments
mcphage 419 days ago
> If this can gather relevant/related facts about key events, it could still be helpful in understanding what's going on in the world?

I'm not sure that's the case. If you pulled from Ukraine's military reports, you'd get one impression on how the war is going. If you pulled from Russia's military reports, you'd get a different impression on how the war is going. If you pulled both, you'd end up with a lot of confusing contradictions. None of those would help you understand what's going on in the world.

Furthermore, to actually understand what's going on in the world, you need to know not just what, but also why. And "why" isn't exposed in any strictly neutral, fact-based documents.

link
agentictribune 419 days ago
I don't see why an AI can't include a "why." An AI can list motives and conflicts of interest that impair credibility of claims being reported. AI agents can quickly pull in historical context, like from Crimea. If I keep going with this project, I could also add bots to monitor telegram war-reporting channels and include that context too.

As I put on the "about" page (https://agentictribune.com/article/about): "Deciding what to cover, which context to include, and how to explain complex issues involves judgment - even for AI." Some of that judgement comes from how I generate prompts, or which feeds I poll, or even the choice of model provider.

Reporting of contradictions and lies can be factual. Neutral in tone doesn't mean that bad things, lies, and falsehoods can't be referred to factually as things that happened or were falsely claimed.

I'm not sure if the issue is more with the writing, and perceptions of what AI is capable of, vs whether I'm just poorly describing the goals of this experiment. I _do_ want to optimize the agents as much as possible to analyze and reason about and contextualize the news. I just want it _not_ to be an influence-bot, written in a persuasive tone, promoting a particular agenda, or deliberately spreading disinformation. I think most of the "AI-slop" we see online is produced with an intent to mislead, but here's a site that's more open about what it's doing.

I'm finding AI agents to be surprisingly capable, and I think they'll keep getting better. Though most AI-news is nefarious, I'm not sure that AI-news has to be "bad" or unethical. Even mainstream news organizations like AP use AI for some reporting, such as reporting on quarterly earnings reports (https://www.ap.org/the-definitive-source/announcements/autom...)

link