It is happening again. Why do I even bother? I can’t stand idiots, and I run into them. OK, so I knew that DWW is censoring me (after the facts, meaning that my comment first shows up and only later disappears) when I sign as Ludditus or Béranger, either from Germany or from Romania, so I decided to pretend I’m from Georgia (Gruzia).

I added this comment to DWW Issue 1109:

11 • @8: Some people are completely ignorant RE: LLMs or AI chatbots (by St. George on 2025-02-17 14:44:47 GMT from Georgia)

Tumbleweeds @8 wrote an aberration: “You have the wrong idea about the AI chat tools. They don’t create answers. They find them just like I do, only they are much more efficient and faster. They are internet crawlers with complex learning algorithms and huge, expanding databases. If no one has asked the question and no NI (natural intelligence) has answered it, the AI will have nothing to show.”

THIS IS WRONG. LLM-based chatbots initially ONLY had a model from which answers were CREATED through statistical inference and language generation. Only later, and only some chatbots started ALSO performing web searches and summarize an answer from the web results. The first such engine I used was Perplexity.

The 2 working modes (1: strictly parametric, using many billions of weights and biases that somehow store the knowledge they were trained on; and 2: using a web search’s results to generate an answer) are completely different!

We’re more than 2 years into the mass utilization of chatbots, and so many people are completely clueless! Linux people aren’t more knowledgeable nor smarter than Windows users. (Maybe it’s GNOME that makes people stupid.)

Even funnier, sometimes, an AI agent will avoid or refuse to perform a web search even if the web search capability is enabled. It happens when it considers that this is a knowledge issue, not a recent news one, so why searching the web? Sometimes, I had to specifically tell Mistral, Copilot, or Gemini, “Please search the web to answer.” Then, Claude cannot search the web. Copilot’s new Think Deeper DOES NOT search the web, because o1 is a purely “reasoning” model.

When Perplexity, Gemini, Copilot, ChatGPT do offer web links as references, you can be sure they did search the web. Otherwise, they did not!

Also yesterday, Claude gave a correct answer without searching the web. Perplexity, ChatGPT, and Copilot searched the web, gave several links in support, but the answers were poor! The fix was to install a package. Some AI chatbots wanted me to install half of the universe and the kitchen sink. Go figure.

Most LLMs you run locally would not search the web either. Those small, over-distilled Llama models.

People should try DeepSeek without web search and with DeepThink (R1) enabled, and examine its chain of thought: it’s a fascinating and frightening insight into LLM hallucinations!

Just yesterday, I asked several chatbots about a punchline from a French movie. They ALL hallucinated, but most accepted my correction, and even came with more details, once put on the right track. Only the Chinese DeepShit stubbornly refused to admit failure and insisted in its hallucinations, mocking me for my “parallel universe” in which what I said (and which was the truth!) was true. It’s Chinese, so it must be right! Like Uncle Xi.

After a couple of hours, it disappeared! Now at number 11 there’s another comment.

To put this in context, this DWW has in its Q&A section an answer to this question: “Under X11 I can use xdotool to script actions and xclip to manipulate the clipboard. Are there any equivalents for Wayland?”

A comment that was at number 2, but got censored as well, was dismissive of the idea of asking such things: “Go ask ChatGPT, Gemini, perplexity.ai or whatever.” To this comment replied at number 8 “Tumbleweeds from the United States,” and this triggered my comment.

Jesse Smith, Ladislav Bodnar, or whatever script they used to censor comments, are utterly retarded. Such censorship would sound absurd even for Xi Jinping’s China!

Apparently, the reference to chatbots is not verboten, because the comment counted as number 8 is still there. That ignorant’s BS can stay, but my clarification cannot!

The only offensive statement in my comment, “Maybe it’s GNOME that makes people stupid,” could have been deleted. There are major sites that do that instead of removing entire comments. But DE wars happened before in DWW’s comments, so maybe this wasn’t the reason I was censored.

Maybe the true reason is that Jesse Smith and Ladislav Bodnar are shitheads.