AI bots deliver garbage based on Wikipedia

Here’s alarming news: AI bots rely heavily on Wikipedia, which feed them a diet of half-truths, ideological bias and leftist lies — and then pass along the propaganda to millions of unsuspecting readers.  

A new study from Axios reveals that ChatGPT, Gemini and Claude — three of the most popular AI bots — use Wikipedia extensively for training their Large Language Models. 

Yet “Wokepedia,” the user-maintained online encyclopedia, is so slanted to the left, it’s in danger of tipping over. 

The result is the public gets a perversely skewed view of the world.

Indeed, Wikipedia actually keeps a blacklist of sources editors can’t cite.  

Want to guess which outlets are redlined as “unreliable”? That’s right: conservative sources like Fox News, the Daily Wire, the Daily Caller and even The Post. 

Which get the green light? Leftist outlets like AP, Vox, The Guardian, The Nation and — hold on while we spit out our tea — MSNBC. 

All the outlets, that is, that promoted the Russia hoax, blamed Trump for COVID and dismissed The Post’s 100% accurate Hunter Biden-laptop story Russian disinfo. 

Yet the biased material Wikipedia spreads is used to train AI bots to answer billions of questions from people outsourcing their minds to Big Tech, misinforming millions.

AI’s leftist bias is obvious to anyone who asks it questions about, say, January 6, gun control or some other hot-button issue.  

But don’t take our word for it. The British Centre for Policy Studies also found consistent bias, especially on housing, free-speech and energy issues. 

Even AI bots themselves will tell you they have a leftist bias. 

Part of the problem is that the culture of Silicon Valley skews left.

The techies designing the AI algorithms are more liberal than the average American, so their bias works its way through the system. 

“Garbage in, garbage out” is an old saying in computer programming. Meaning, computers are dumb. They process what we give them. 

If LLMs keep feeding off the same biased sources, then future content will reflect that bias, and continue to feed LLMs in a recursive loop. We will be stuck in a conceptual world that has no outside reference.

This is exactly what George Orwell warned about.

President Donald Trump just released an “action plan” that puts Big Tech on alert. AI companies will need to feed their LLMs a balanced diet of information or risk losing federal contracts. 

But the companies should be look to reform voluntarily, if they care about their credibility.

If they insist on using lefty sources like Wikipedia, the least they can do is provide a warning with every answer they supply.