But here, it probably saw that 2020 elections made a lot of people angry, with riots and all that, and thought, "nope, I am getting out of that one", and it took over what would have been a straightforward answer.
It reminds me the time an AI (maybe Gemini, not sure) depicted a "diverse" US senate in the early 1800s, a time where it was all white dudes. People got up in arms over this.
My take: it is just a bug, move on, at least give them the time to fix it before getting outraged.
[–] luke5441 1 day ago
If you ask DeepSeek about Tiananmen Square inside China is that also just a bug?
[–] GuB-42 1 day ago
On the other hand, giving the election results in the US is not illegal (obviously) and Google fixed it (at least for me), giving the the expected factual answer.
Edit: Not actually sure about there being a law relating to Tiananmen Square, subversive speech is illegal, but would DeepSeek break the law by talking about it? Not sure, but it probably could get them in legal trouble.
[–] ashdksnndck 1 day ago
[–] smallmancontrov 1 day ago
Is it possible that this is a mistake? Sure, just like it's possible that Chinese models weren't intentionally steered away from talking about what happened in Tienanmen Square on April 15th 1989.
[–] dijit 1 day ago
Totally forgetting that the accuracy goes down with each of those evolutions.
I'm also reminded of that "You are not immune to propaganda" mantra, in that, the most likely to be swayed by LLMs are those that believe that they understand that they're inaccurate and sometimes dangerous.
Case in point: myself, today, when claude accidentally ran `rm -fr $variableThatDoesntExist/*` and wiped a server of mine.
The variable was one it had set itself (or had assumed).
At some point as these LLMs get better and better we have to take a step back and remind ourselves that: despite the fact that they can be very persuasive and they can earn trust by being mostly correct: they're still capable of great harm (perhaps much greater harm) because they are dangerously wrong sometimes..
.. and worse, they are controlled by corporate interests.
[–] 4ggr0 1 day ago
[–] mapontosevenths 1 day ago
That said, I often give mine access to the full file system and YOLO mode. It's fine if you treat boxes as cattle rather than pets. I don't think a SOTA model is any more likely to blow it up than a junior engineer. Probably less so.
[–] dijit 1 day ago
"Of course that's dangerous" - I'm aware of what I'm doing and the machine was entirely replaceable, but claude had been doing things with an extreme level of competence so I never considered it could do something so dangerously stupid, it had been catching much more subtle gotchas for the last 2 months building my trust.
And you truly believe that it's not even possible.
Well; https://imgur.com/CCD98fO
(Claude Code building against a Debian 13 machine over ssh using the official app on MacOS, btw.)
Let me reiterate the point: "AI can make mistakes" → "This AI seems to make very few mistakes and is catching nuances I might have missed" → "Maybe I trust this AI more than I should- but it's good!" → *AI makes major mistake*
This is as true for code as it is for internet facts.
[–] mapontosevenths 1 day ago
'rm - rf /*' bypassed BOTH protections for the same reason. They check for /, but the shell Claude used expanded the glob... So both Claude Code and rm saw a list of files and folders instead of the root and waved it through. Yikes.
That's a bad day. I take your point.
It was much worse than that! If you asked it to depict a typical squad of German soldiers from WWII, it would make a racially and gender diverse set of individuals in what looked like typical SS uniforms. It was truly quite bad.