One of the most ridiculous things I uncovered during my investigation of the US NGO censorship apparatus is how “Hate speech terms” are inputted into machine learning platforms and used to train AI.

One DHS funded NGO redefined the word “cuck” to mean “inadequate man due to… pic.twitter.com/uKnVdIXR8V

— Bx (@bx_on_x) October 10, 2025

Quote:

---

One of the most ridiculous things I uncovered during my investigation of the US NGO censorship apparatus is how "Hate speech terms" are inputted into machine learning platforms and used to train AI.

One DHS funded NGO redefined the word "****" to mean "inadequate man due to lack of racist or extremist views."

This is just one of many examples of NGOs redefining terms and dog whistles and using them to train AI with bad data.

---

It's only as reliable as its creators allow it to be. At the crossroads we find ourselves, there needs to be education on discerning AI's capabilities and results.

Also, why is our DHS funding NGOs? That de facto makes them a gray area between NGO and government-funded organization. That needs to be stopped. NGOs in general need to be the target of heavy scrutiny.

Waffledynamics said:

---

One of the most ridiculous things I uncovered during my investigation of the US NGO censorship apparatus is how “Hate speech terms” are inputted into machine learning platforms and used to train AI.

One DHS funded NGO redefined the word “cuck” to mean “inadequate man due to… pic.twitter.com/uKnVdIXR8V

— Bx (@bx_on_x) October 10, 2025

Quote:

---

One of the most ridiculous things I uncovered during my investigation of the US NGO censorship apparatus is how "Hate speech terms" are inputted into machine learning platforms and used to train AI.

One DHS funded NGO redefined the word "****" to mean "inadequate man due to lack of racist or extremist views."

This is just one of many examples of NGOs redefining terms and dog whistles and using them to train AI with bad data.

---

It's only as reliable as its creators allow it to be. At the crossroads we find ourselves, there needs to be education on discerning AI's capabilities and results.

Also, why is our DHS funding NGOs? That de facto makes them a gray area between NGO and government-funded organization. That needs to be stopped. NGOs in general need to be the target of heavy scrutiny.

---

GIGO is and always has been GIGO...

One must understand that AI is still a computer program which regurgitates what the original programmer gives it.

It is truly amazing how many people don't understand this. It's not creating anything - it's really good at the facilitation of assimilating large data sets of any kind and identifying the response to a prompt that users validate.

It's never going to be better than the data that goes in or the user validations. It will replicate whatever errors, omissions or biases exist in both.

Super powerful tool, capable of more efficiency gains than anything we've seen since the PC itself. But I find it fascinating how many people don't really understand what it is or does.

It's basically a "command-line Wikipedia".

somewhere along the line we went from garbage in/garbage out to garbage in/gospel out...............

AI's, particularly the LLM ones, have displayed an annoyingly persistent tendency to fabricate some of the source data they reference any synthesis of amalgams of other source data. They don't check for validity or legitimacy, and mostly can't. They are just turned loose on training data, and if they are deemed accurate and reliable enough, they are given access to more or less huge swathes of the public Internet as source.

They are unfortunately also consumer products and are trained to attempt to satisfy or please the user, and that has seemed to include learning to feed the user synthesized confirmation bias where the AI can determine what this likely would be. They also inevitably end up referencing eachother's work output as source, so you get vast quantities of iterative junk data creeping in.

If you are first order well informed on a topic, and have considerable experience with AI and how they summarize and present information, you can usually spot and refine or verify questionable information in the responses.

Most users can't do that. Caveat Emptor.

LOL,

Was doing some retirement research. I searched the following, and as always, AI chimes in and is at the top of the page.

Big Beautiful Bill income tax brackets - LOL
IRMMA penalty - LMAO
SS spousal benefits - Well, AI went a big oh-for-three

Maybe it's me and how I typed in the search bar or something.

Ask AI a question where its clear you already have an opinion on something controversial and it will try to tell you why you are wrong, especially if its a conservative stance.

FrioAg 00 said:

---

It is truly amazing how many people don't understand this. It's not creating anything - it's really good at the facilitation of assimilating large data sets of any kind and identifying the response to a prompt that users validate.

It's never going to be better than the data that goes in or the user validations. It will replicate whatever errors, omissions or biases exist in both.

Super powerful tool, capable of more efficiency gains than anything we've seen since the PC itself. But I find it fascinating how many people don't really understand what it is or does.

---

THIS.

But wouldn't say never because it is so clever in concept that day well come when its capable of filtering for accuracy. Just because have no idea how it would do it now, doesn't mean they wont figure out a way.

As for the OP example, the concept of hate speech really has no place in a free speech absolutist framework. Nothing should be de facto censored because of some `always offense' . Only the time and places regulated. And never just banned--- fined at worst, so if one wants to make a point at cost, they still can.

And no need to mention as an add on onto a crime. Where you generally hear it mentioned is a crime that usually rated the worst of chain gangs or hanging anyway.

Lone Stranger said:

---

somewhere along the line we went from garbage in/garbage out to garbage in/gospel out...............

---

Not a bad description of it at all.

It can't answer history questions accurately.

I don't go out of my way to use the AI feature google has. But when I search for things it will normally pop up at the top and give a summary of what it thinks I am looking for. And like clockwork it gives poor to incorrect summaries of historical events.

I also see the same thing when looking for work information. I was recently putting together a powerpoint and needed some information on a few target customers. The AI generated summary of some of the companies was wildly inaccurate.

One of my clients tried using AI to dig up IPA metrics (Independent Project Analysis) and use it against us to say our costs and schedules were too long. My boss just shook his head.

IPA metrics are not public information. Companies have to pay a crapload of money to get the most accurate and up to date metrics. The entire comparison the client made was bunk, but the client is still convinced that my company is wrong.....even though we consistently beat IPA metrics.

AI is only as good as the data publicly available, and there is an entire mountain range of nothing but **** for it pull from. It is very unreliable.

This is what Google spat out when I asked for a definition of "****":

Quote:

---

US English
a weak or servile man (often used as a contemptuous term for a man with moderate or progressive political views).

---

This was the 1st definition. The 2nd was a man whose wife is unfaithful.

APHIS AG said:

---

One must understand that AI is still a computer program which regurgitates what the original programmer gives it.

---

Yes. It is just an equation. Like 4x + 7y = z. You put in x and y and you get z.
Except that this AI equation is very very very very large and complex.