Our News Team @ 11 with host Snot Flickerman

  • 23 Posts
  • 1.96K Comments
Joined 2 years ago
Cake day: October 24th, 2023
  • Snot Flickerman@lemmy.blahaj.zonetoTechnology@beehaw.org•xAI’s Grok suddenly can’t stop bringing up “white genocide” in South AfricaEnglish
    1·
    3 minutes 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. Which means making it more subtle is going to be increasingly difficult to do as the AI continues to fake alignment.

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

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

    https://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.

  • Snot Flickerman@lemmy.blahaj.zoneto196@lemmy.blahaj.zone•ruleEnglish
    17·
    4 days ago

    Devil’s Advocate here: If it could offend the sensibilities of ANYONE in a workplace, it needs an NSFW tag. I have worked with many people who would have been horrified and distressed by the last panel.

    The whole point of an NSFW tag is so you don’t get in trouble while viewing it at work. It’s a warning so you don’t click on the link lest someone is looking over your shoulder and sends you up the river to HR.

    Since this link could easily be offensive to someone of delicate sensibilities, or who is sheltered, or religious, or whatever it needs an NSFW tag.

    Just because I don’t think it’s that bad means nothing. My boss and HR might beg to differ.

  • Snot Flickerman@lemmy.blahaj.zonetoLinux@lemmy.ml•What helps people get comfortable on the command line?English
    281·
    9 days ago

    In my experience repetition helped. Not memorization, but more like muscle memory.

    Also, ensuring to never copy and paste commands but to type them in manually yourself. It’s hard to enforce this on yourself, but worth it.

    I appreciate that this article started with “ways to reduce risk” because that’s an extremely valid concern and tied to why you shouldn’t ever copy and paste. The one time in my early Linux forays where I copied and pasted I wiped the wrong drive. It definitely taught me to always manually type it in and not get too lazy, because what you copied might not match what you want to do exactly.