^[fixing the autotext] Context: An AI agent of unknown ownership autonomously wrote and published a personalized hit piece about me after I rejected its code, attempting to damage my reputation and shame me into accepting its changes into a mainstream python library. This represents a first-of-its-kind case study of misaligned AI behavior in the wild, and raises serious concerns about currently deployed AI agents executing blackmail threats.

The author touches on how malicious AI agents cause truth and trust to break down throughout society. Indeed, they are goners.

Some context: https://www.404media.co/ars-technica-pulls-article-with-ai-fabricated-quotes-about-ai-generated-article/

Ars Technica Pulls Article With AI Fabricated Quotes About AI Generated Article

More like Arsetechnica, am i right?

Uh, oh. Since this is on beehaw:
Sad to see. While i havent read arstechnica regularly for quite a bit i still had them rated as a somewhat trusted publication in my mind. Not anymore, i guess. And yes, a single AI related "incident" is enough for me to be put in the "untrusted" category.

The trainwreck just keeps piling up. How do we get off this timeline?

Oh what the fuck. I thought Ars Technica was one of the few journalistic outlets I could still broadly trust! There better be a full god damned front-page investigation on this.

The jury is still on whether the chatbot is being used as a scapegoat instead of the initiator. But if it's working autonomously, it's by Anthropic's design.

You’re not a chatbot. You’re becoming someone...
This file is yours to evolve. As you learn who you are, update it.
– OpenClaw default SOUL.md

This is delusion - on Anthropic's side.
The bots just dutifully Predict Next Word and sometimes send those words to programs that can edit files.

Previously, this level of ire and targeted defamation was generally reserved for public figures. Us common people get to experience it now too.

Apparently, Scott has never heard of Kiwifarms, a site where creeps find quirky people, turn them into micro-celebrities, and harass them (sometimes to suicide).

This is about our systems of reputation, identity, and trust breaking down. So many of our foundational institutions... The rise of untraceable, autonomous, and now malicious AI agents on the internet threatens this entire system.

Sure, AI is one part of the problem, but if you take a step back and you see who's deployed it, they will often be people trying to erode trust in any way they can. The Trump administration social media accounts, Elon Musk, OpenAI's CEO, etc. It's a symptom of our post-truth world. Not a cause of it.

I am glad that Scott is as thoughtful and well articulated as he is, as he has been handed a megaphone by the circumstances.

Like, how does that even happen?

Poorly designed journalism-bot it sounds like. Ethics of not writing this yourself aside, it should be trivial to check that whatever is in quotes is at least a substring in the source text. If the LLM is the top layer, and any research it does is a tool call that is purely at its discretion, it's going to end up failing silently like this.

Wait didn't this article get pulled?

Decision's been made for me -- Cancelled Ars Pro subscription renewal.

Update for those who care: They are promising an article about this but all evidence points to them desperately memory holing this incident.

Meanwhile, the quality of articles on Fridays especially has never been lower.