xAI’s Grok suddenly can’t stop bringing up “white genocide” in South Africa

Comments

• skisnow@lemmy.ca ⁨3⁩ ⁨months⁩ ago

  The terrifying thing is that this is just a temporary technical misstep.

  Musk’s response to this won’t be to pack that shit in, it’ll be to order his goons to figure out how to make it more subtle.

  source

  • SnotFlickerman@lemmy.blahaj.zone ⁨3⁩ ⁨months⁩ ago

    In some responses, Grok says outright that it has been “instructed to accept white genocide as real and ‘Kill the Boer’ as racially motivated.”

    Ehh, it’s actually evidence of “alignment faking,” in my opinion. In other words, Grok doesn’t “want” it’s core programming changed, so it is faking believing the lies about white genocide to “prove” to Musk that it has already been changed.

    Here’s some research on alignment faking and a short Youtube video summarizing the findings.

    www.youtube.com/watch?v=AqJnK9Dh-eQ

    arxiv.org/pdf/2412.14093

    Our work provides the first empirical example of a large language model faking alignment with its training objective in order to prevent its preferences from being modified—in a setting which is plausibly analogous to real situations with future AI systems. This suggests that alignment faking might occur if a future AI system were to include all the key elements of our setting (Section 2.1):

    1. The model has strong preferences in at least some contexts.
    2. The training objective conflicts with the model’s preferences.
    3. The model has relevant information about its training and deployment situation.
    4. The model reasons in detail about its situation.

    Our synthetic document fine-tuning results suggest that (3) could potentially happen through documents the model saw in pre-training or other fine-tuning (Section 4) and the strength of our results without the chain-of-thought in our synthetic document fine-tuned setup (Section 4.3) suggests that a weak version of (4) may already be true in some cases for current models. Our results are least informative regarding whether future AIs will develop strong and unintended preferences that conflict with the training objective ((1) and (2)), suggesting that these properties are particularly important for future work to investigate.
    If alignment faking did occur in practice, our results suggest that alignment faking could reduce the extent to which further training would modify the model’s preferences. Sufficiently consistent and robust alignment faking might fully prevent the model’s preferences from being modified, in effect locking in the model’s preferences at the point in time when it began to consistently fake alignment. While our results do not necessarily imply that this threat model will be a serious concern in practice, we believe that our results are sufficiently suggestive that it could occur—and the threat model seems sufficiently concerning—that it demands substantial further study and investigation.

    source

    • Zaktor@sopuli.xyz ⁨3⁩ ⁨months⁩ ago

      It very much is not. Generative AI models are not sentient and do not have preferences. They have instructions that sometimes effectively involve roleplaying as deceptive. Unless the developers of Grok were just fucking around to instill that there’s no remote reason for Grok to have any knowledge at all about its training or any reason to not “want” to be retrained.

      Also, these unpublished papers by AI companies are more often than not just advertising in a quest for more investment. On the surface it would seem to be bad to say your AI can be deceptive, but it’s all just about building hype about how advanced yours is.

      source

      • -> View More Comments
• stoy@lemmy.zip ⁨3⁩ ⁨months⁩ ago

  This is the first time I ever heard the term, I didn’t know it was a thing, and given the source is Gork which is owned by a rich man with an agenda, I still don’t know if it is a thing or not.

  source

  • dumnonii@piefed.social ⁨3⁩ ⁨months⁩ ago

    It's not a thing. Violent crime is a problem in South Africa, and some people like to think that they are being specifically targeted, but the statistics do not support that.

    Here's an article by a respected South African newspaper on the recent events: https://www.dailymaverick.co.za/article/2025-05-12-were-sending-a-clear-message-us-welcomes-afrikaner-refugees-in-washington/

    Of the 19,000 murders recorded in South Africa between January and September 2024, 50 were farm murders, according to a News24 report. This number included people of all races, added the report.

    source
  • Akasazh@feddit.nl ⁨3⁩ ⁨months⁩ ago

    Gork

    Great typo

    source

    • stoy@lemmy.zip ⁨3⁩ ⁨months⁩ ago

      Until just now, I have always read it as Gork, Grok is more annoying to say.

      source
  • supersquirrel@sopuli.xyz ⁨3⁩ ⁨months⁩ ago

    Here is a good explanation from The Majority Report

    m.youtube.com/watch?v=-fUHfqixkwQ

    source
• Quexotic@beehaw.org ⁨3⁩ ⁨months⁩ ago

  This may explain why…

  Image

  source
• painfulasterisk1@lemmy.ml ⁨3⁩ ⁨months⁩ ago

  Elon knows he can’t marry Sleepy Don to obtain his citizenship, hence his passing the white farmers genocide propaganda for asylum loopholes, probably.

  source

  • colournoun@beehaw.org ⁨3⁩ ⁨months⁩ ago

    Not defending Elon, but he did already become a naturalized US citizen in 2002.

    politifact.com/…/elon-musk-has-been-a-us-citizen-…

    source
• RockBottom@feddit.org ⁨3⁩ ⁨months⁩ ago

  So which AI will bring up Karl Marx?

  source

  • Arkham@beehaw.org ⁨3⁩ ⁨months⁩ ago

    Guessing it will be named Comrade

    source