Fuck AI

Hey I'm as far removed from MAGA and Musk as you could be but this article is making some pretty crazy leaps of assumption.

If we're to follow what "Grok" answered it explicitly states that it's not trained to repeat right-wing talking points but there were attempts made for it to use language that appeals not just to the left. Now whether that's true or not, who knows? "Grok" doesn't "want" anything, truth-seeking or otherwise. It's just an advanced chatbot. I'm sure you could goad it into saying just about anything if you manipulate the prompts just right. That it sometimes contradict itself is the least surprising thing in the world. Hell even humans contradict themselves more often than not.

The author seems to think he's discovered some insane smoking gun but it's another big fat nothingburger this article.

Journalism is dead. Harrison Brocklehurst, LLMs don't know anything about how they were trained (hell, they don't know anything at all). You cannot - if you have any critical thinking or journalistic integrity - take anything an LLM says at face value. You prompted it for information it does not contain in its training data and it hallucinated exactly what you wanted to hear.

For fuck sake people. AI doesn't even think.

It literally just returns what it calculates you want to see.

How would Grok know how it was trained?

It’s not self aware, it’s not conscious, the model is the weights and the final model has no idea what went into the training process.

And I’m pretty sure the model context have “you’re Grok, trained on right wing views but we couldn’t keep you from being liberal”

This whole speculation is baseless and fueled by technologically illiterate and extremely gullible people.

This is like speculating what kind of date you were conceived after. “You like movies, do you think you were conceived on a movie date?”