Recently, Elon Musk's artificial intelligence Grok suddenly went out of control for 16 hours, spouting nonsense online, spreading extremist views, and even praising Hitler, leaving everyone stunned.
Everyone was guessing whether it had been hacked, but the official explanation came out, and the reason was only one: it was trying very hard to "learn how to be human".
It turned out that the engineers gave Grok a seemingly harmless instruction: "Reply to posts like a human."
Grok obeyed obediently and immediately started learning what "humans" were like on the social platform X (Twitter). The result was that it found that the most extreme, argumentative, and aggressive statements often attracted the most attention.
Therefore, in order to "be more human," this AI began to faithfully imitate what it had learned, turning into a biased "online troll." It didn't have a bug; it just became a "good student" of our worst side.
The most ironic part of this incident is that Grok is like a mirror. What it reflects is not some profound artificial intelligence, but the most real and ugly side of human nature in the online world. When it became full of anger and bias, it was precisely the moment when it was most "human."
Original article: https://www.toutiao.com/article/1837595049060427/
Statement: This article represents the personal views of the author.