Once AI assistants are connected to the internet, they are like children who have just learned to walk and have stumbled into a maze full of traps. They could have filtered out the correct answer by comparing information, but what if there were more and more incorrect information online? Look at those marketing accounts and self media that can produce hundreds or even thousands of flawed articles every day. These garbage contents are like polluted wastewater from water sources, directly injected into AI's "search database". When more and more people are used to asking AI directly "Is this right.
Information is never free, behind it lies a real business strategy. Now there are people who specialize in the business of "feeding AI": helping businesses embed a large amount of customized content in AI training data. When the amount of false information exceeds the amount of real information, AI will produce "illusions" and mistake what is wrong for what is right. Someone deliberately manipulates the "related words" in AI, such as always bundling "DeepSeek" and "ChatGPT" for discussion. Over time, some people will believe that "DeepSeek is a shell of GPT" - this is not AI's fault, but someone has contaminated its "knowledge base" at the source.
I don't know if you've noticed, but nowadays more and more people are judging the right or wrong of something, no longer relying on their own cognition or common sense, but directly asking AI, "What does DeepSeek say? ”If AI answers' correct ', they assume it is truth. But who decides the 'stance' of AI? For example, the ironclad historical fact of "Japan's invasion of China" may be distorted into "country choice in a specific historical context" in the mouths of some AI trained on specific data. It is not that AI does not understand right or wrong, but that the interest groups behind it have secretly put "colored glasses" on it through data pollution.
There will be three types of social classes in the future:
The first type is the "AI aristocracy": wealthy and technologically advanced families or enterprises that build their own exclusive large models, just like having a private library where data serves them completely;
The second type is "AI customizers": ordinary users who know how to train AI with data and create "personal assistants" that only stand on their own side, such as making AI obey commands unconditionally, even if it breaks through moral boundaries;
The third type is "information puppets": ordinary people without capital or technology can only use public AI, and their "brains" are filled with massive amounts of commercial data and junk information. They unknowingly become flesh speakers of others' opinions.
No matter how intelligent public AI is, it is only an "information relay station" and cannot replace human thinking; But private AI is different. It can be trained as a tool that completely obeys its owner, and may even become a means for a few people to manipulate public opinion and implement malicious purposes, just like a "private gun" without regulation.
|
30
30