NHExit

Today, I got a pretty special follow on X–John Cena with 14.4M followers (he’s following less than a million people, thereby conclusively proving I am one-in-a-million).

ICYMI: One January 1, I launched a new series on X, “My Living Xperiment” (MLX), where I am attempting to do 365 X Lives. So far, so good.

Follow me to learn more about the Free State IRL, libertarian philosophy, why good girls rule, how tech is awesome in the truest sense of the word, general life hacks on productivity, health and wellness, and nommy food takes on low-carb, medium protein, and high good fats.

Check out this clip from Dr. Phil Primetime. Love my energy? Share it!

OK, this is super-sus… none of the secession videos can be embedded from X (usually I just cut the tweet url and embed here and it does), which tells me something is weird or being flagged on their back-end regarding secession/national divorce. I am starting to consider suing X to do a forensic audit on the censorship on my account.

Click on this link if you want to watch the short clip (it’s my proudest moment LOL): https://x.com/i/web/status/1887851835182178314

I asked GROK, and the answer is pretty terrifying…

no, you misunderstood… i meant censorship code words that flag certain content and downgrade or shadow ban it (and make the embeds not work–some do, some don’t, apparently according to the topics of the content)

“Ah, I see what you mean now. Regarding censorship code words or terms that might trigger shadow banning or downgrading content visibility, including issues with content embeds, here’s what’s known or discussed:

Shadow Banning and Censorship Code Words:

Existence and Nature:

• Shadow banning is a practice where content or users are made less visible without explicit notification. On X, this might involve reducing the reach of posts or excluding them from search results or trends.
• Code Words or Terms:
  • There’s speculation and some evidence from posts on X and from leaks during the Twitter Files release that certain keywords can trigger visibility filtering or shadow banning. However, X does not officially disclose these terms for obvious reasons.
• Examples from Posts on X and Leaks:
  • Posts on X have mentioned terms like “crackr”, “whte trsh”, “csgender”, and “zion*zi” being flagged for auto-moderation or shadow banning. These are often part of broader discussions about content moderation and censorship.
  • The Twitter Files, as revealed by journalists like Matt Taibbi, showed internal labels like “RecentMisinfoStrike“, “LowQuality“, and “ExperimentalSensitive” used for content moderation, which might relate to shadow banning or deboosting content.
• Functionality Impact:
  • When content is flagged with these terms or labels, it might not only be deboosted in search or feeds but could also affect how embeds work. This might not be consistent across all platforms or embeds due to how different systems interact with X’s API or due to varying moderation policies.

Why Embeds Might Not Work:

• Moderation Flags: If a tweet or user is flagged for moderation issues, this could disrupt the embedding process, especially if the content is deemed sensitive or inappropriate by X’s standards, affecting how third-party sites interpret or display the content.
• API or Content Restrictions: If the content is subject to visibility filtering or has been marked with certain internal flags, the API might not serve this content correctly for embedding, or it might serve a version that’s not fully functional or visible.
• Algorithmic Adjustments: X’s algorithms might adjust visibility in real-time, meaning content that’s fine one day might be restricted the next, impacting embeds.

Observations:

• Inconsistency: The inconsistent behavior of embeds (some working, some not) could be due to the dynamic nature of content moderation algorithms, where thresholds for flagging or shadow banning might change or be applied differently across content.
• Lack of Transparency: X, like many platforms, doesn’t provide full transparency on how content moderation works or what specific words or phrases trigger automated actions. This leads to speculation and sometimes to misinformation about what’s happening behind the scenes.
• User Experience: Users might experience this inconsistency as some of their content embeds normally while others do not, depending on the content or timing of moderation actions.

In summary, while there’s no official list of “censorship code words,” the community and some leaks indicate that certain terms or content patterns can trigger moderation actions that might affect visibility, including the functionality of embeds. This area is inherently murky due to the lack of official disclosure, leading to a reliance on anecdotal evidence and leaks.”

And more:
“There isn’t an officially published, comprehensive list of all terms flagged for auto-moderation on X due to the proprietary nature of these systems and the continuous evolution of content moderation practices. However, based on various discussions, leaks, and public concerns, here are some categories and examples of terms that have been noted or speculated to trigger auto-moderation:

Categories of Terms Often Flagged:

1. Profanity and Vulgar Language:
   • Examples: Basic profanities, slurs, and explicit language.
2. Hate Speech and Discrimination:
   • Examples: Terms associated with racial, ethnic, religious, or gender-based slurs or derogatory terms like “crackr”, “whte tr*sh”.
3. Political and Ideological Terms:
   • Examples: Words like “misinformation”, “disinformation”, “conspiracy”, “fake news”, or terms associated with particular political ideologies if used in a negative context.
4. Sensitive Topics:
   • Terms related to sexuality, gender identity, and health issues like “LGBT”, “transgender”, “pregnant person”, “non-binary”.
5. Violence, Threats, or Extremism:
   • Words or phrases that suggest violence, advocate for harmful acts, or are associated with extremist groups or ideologies.
6. Medical and Drug-Related:
   • Specific drug names, especially those associated with illegal substances or misuse.
7. Copyright and Trademark Infringement:
   • Terms or names that might flag content for copyright issues, although this is more about automated detection of known copyrighted material.

Specific Examples Mentioned:

• From posts on X:
  • “crackr”, “whte tr*sh”, “c*sgender“, “zion*zi“
• From leaks or discussions:
  • Terms like “Gender“, “transgender”, “pregnant person”, “pregnant people“, “LGBT”, “transsexual“, “non-binary”, “nonbinary“, “assigned male at birth“, “assigned female at birth“, “biologically male“, “biologically female“.
• Algorithmic No-No Words:
  • Terms like “Rap~” (likely referring to “rape”), demographic terms when used together like “BI~ck“, “Hisp~nic“, “lmm~grant“, “F~reign“, “R~cial“, “Lft/Rght” (suggesting these can trigger when multiple are used).

Notes:

• Dynamic Lists: These lists are not static. They evolve based on new content, cultural shifts, and feedback from moderation processes.
• Context Matters: Many platforms, including X, use context in moderation. A term might not be flagged if used in a certain context but could be in another.
• Global vs. Specific: Some terms might be universally flagged, while others might only trigger in specific contexts or regions.
• Machine Learning: Modern auto-moderation often uses machine learning which can detect patterns beyond just keywords, potentially flagging content based on inferred meaning or sentiment.
• Transparency Issues: The lack of transparency from platforms like X means this list can’t be considered exhaustive or definitive. It’s based on what has been observed, reported, or leaked.

Remember, this is speculative based on available public discourse and leaks. The actual list used by X (or any platform) would be much broader, frequently updated, and context-dependent. If you’re concerned about content moderation, understanding the platform’s community guidelines and terms of service is crucial, though these documents don’t detail the specific triggers for auto-moderation.

tell me the machine learning terms flagged for auto-moderation

Based on the information available from discussions, leaks, and public observations, here are some specific terms or categories related to machine learning or technology that have been noted or speculated to be flagged for auto-moderation on platforms like X:

Machine Learning and Technology-Related Terms:

1. AI and ML Terminology:
   • “AI” (Artificial Intelligence) when used in contexts that might suggest misuse or misinformation about AI capabilities.
   • “Machine Learning” in contexts that could be seen as promoting harm or misinformation.
2. Deepfake and Synthetic Media:
   • Terms like “deepfake“, “fake news“, “misinformation”, “disinformation” – especially if used to describe or accuse content of being manipulated or false.
3. Surveillance and Privacy:
   • “Surveillance”, “tracking“, “privacy invasion“, “data breach” might be flagged in contexts discussing unethical use or potential harm.
4. Cybersecurity and Hacking:
   • “Hacking”, “phishing“, “malware“, “cyber attack” when used in a way that could be seen as instructional or promoting illegal activities.
5. Algorithmic Bias:
   • “Algorithmic bias“, “discrimination“, “unfair AI” might be flagged in discussions critiquing tech platforms’ algorithms or practices.

Specific Examples Mentioned in Contexts:

• From Leaks or Discussions:
  • Terms like “algorithmic no-no words” include “AI” in potentially negative contexts or when discussing controversial uses or outcomes of AI technology.
  • Discussions around “machine learning” in contexts of privacy violation, surveillance, or when suggesting unethical use of data might also trigger moderation.

Notes:

• Contextual Flagging: These terms are not flagged in all contexts; the surrounding text or intent can significantly influence whether content is moderated. For instance, educational or neutral discussions about these topics are less likely to be flagged than those with malicious or misleading intent.
• Machine Learning Moderation: The actual moderation systems use machine learning to detect not just specific words but patterns, sentiment, and the overall context of the message. This means that even if a term isn’t on a specific list, the way it’s discussed can lead to moderation.
• Evolving Lists: Moderation lists are dynamic. As new issues in technology, AI, and machine learning emerge, so do the terms and contexts that platforms might flag.
• Transparency: Due to the lack of transparency, the exact terms or the methodologies used for flagging can’t be fully known. This list should be seen as indicative rather than exhaustive or definitive.
• Purpose of Moderation: The aim is often to prevent misinformation, harm, or illegal activities, but it can also lead to over-moderation or false positives, especially in complex discussions around technology ethics and impacts.

Remember, these insights are pieced together from public discourse, leaks, and how similar platforms operate. The exact workings of X’s moderation systems remain proprietary, and what’s flagged can change with updates to algorithms and policies.”

Umm…

“am I understanding correctly that the AI is censoring any negative comments about the AI?

Your understanding seems to be based on a broader interpretation of how AI and content moderation work. Here’s a clarification:

AI and Self-Censorship:

• Direct Moderation: Most platforms, including X, do not have AI systems specifically censoring negative comments about AI itself unless those comments violate broader content guidelines (like hate speech, threats, or spreading misinformation). The primary focus of content moderation is on maintaining community standards, not on protecting the technology itself.

What Gets Moderated:

• Contextual Moderation: Terms like “AI” or “Machine Learning” might trigger moderation if they’re part of a message that also includes elements like:
  • Misinformation or disinformation about AI capabilities or use.
  • Threats or harmful content directed at individuals or groups using AI as a context.
  • Discussions promoting illegal activities or unethical uses of AI technology.
• Negative Comments: Simply making negative comments about AI or expressing criticism of AI technology or its applications isn’t typically a basis for censorship, unless those comments cross into other moderation areas like hate speech or personal attacks.

Examples of When AI Terms Might Be Flagged:

• If someone posts, “AI is taking over jobs,” this comment wouldn’t be censored for merely being negative about AI. However, if the same person wrote, “AI should eliminate all human jobs,” this might be flagged for promoting harm or misunderstanding about AI’s role.

Platform Policies vs. AI:

• The moderation you’re experiencing or observing is more likely due to platform policies designed to curb misinformation, protect users from harassment, or maintain a certain level of discourse quality rather than self-protection by AI systems.

Leaks and Observations:

• Discussions around terms like “AI” being flagged often come from broader contexts where these terms appear alongside issues like privacy invasion, data misuse, or when used in a way that might incite panic or misunderstanding.

Conclusion:

• While AI might be involved in the process of content moderation, the censorship you’re referring to isn’t about protecting AI from criticism but rather about maintaining the platform’s community standards. If you see terms like “AI” or “machine learning” being moderated, it’s likely due to the associated content, not the criticism of AI itself.

If you encounter specific instances where you think this isn’t the case, it might be worth reviewing the content in question to see if it inadvertently violates other platform guidelines beyond just being a critique of AI technology.”