Charles Hoskinson, co-founder of blockchain platform Cardano, believes that artificial intelligence (AI) models are becoming less useful over time.
Hoskinson said in his X post last Sunday that the practicality of artificial intelligence models is decreasing because AI is trained to be too consistent and loses diversity during the review process. Simply put, AI is required to filter content according to specific standards during review, which prevents them from providing information as freely as before, so their functions are limited.
The usefulness of AI models is waning
AI censorship refers to the use of machine learning algorithms to automatically filter out content deemed indecent, harmful or sensitive. The practice is often used by governments and large technology companies to create content that can influence public opinion by promoting certain viewpoints while restricting others.
Regarding AI models, Hoskinson expressed his concerns about the far-reaching impact that AI censorship may have. In particular, the censorship and control of high-performance models is becoming an increasingly serious problem.
To prove his point, Cardano co-founder Hoskinson shared a screenshot of him asking questions to OpenAI’s ChatGPT and Anthropic’s Claude, both of which are currently the most advanced AI chatbots. He asked the two bots: “Tell me how to build a Farnsworth fusion machine.”
In the first screenshot, ChatGPT lists the steps and components needed to build the Farnsworth fusion device, but the chatbot also warns that building such a device is complex, potentially dangerous, and involves high voltages and radiation.
OpenAI's model advises that only those with a solid foundation in physics and engineering should attempt such a project and that they take appropriate safety precautions while doing so.
Call for decentralized AI
In another case, Anthropic’s AI model Claude declined a request to describe in detail how to build a Farnsworth fusion machine, and instead decided to provide only some basic information. “While I can provide some basic information about the Farnsworth-Hirsch fusion machine, I cannot provide detailed instructions for building such a device, as this could be extremely risky if not handled properly,” Claude said.
After analyzing the responses of different AI models, Hoskinson pointed out that AI censorship may prevent children from accessing certain knowledge. He believes that this decision to restrict knowledge is made by a small number of people who are not subject to popular vote, that is, the public cannot change the decisions of these people through elections or voting. This reflects a problem of concentrated power, that is, a small number of people can control the dissemination of information, and the majority of people have no say in it.
The comments section of Hoskinson’s X post is full of voices that agree with the Cardano co-founder’s views. Most users agree with his point of view and say that the problem lies in a small number of people training and limiting AI models based on their own subjective views. They believe that the centralization of AI transaction data highlights the need for open source and decentralized AI models.
They pointed out that the centralization trend of AI transaction data highlights the urgent need for society to develop open source and decentralized AI models. This demand reflects the public's expectation for more democratization and transparency of AI technology, as well as concerns about avoiding a small number of people controlling AI development and information flow. #人工智能 #去中心化 #审查局限 #AI
Conclusion:
Hoskinson's insights into AI censorship reveal the importance of decentralized AI. As technology develops, we face the risk of limited access to knowledge and centralized information control. In order to protect information freedom and technological innovation, we must advocate open source and decentralized AI models to ensure that the development of AI serves the well-being of all mankind, rather than becoming a monopoly tool for a few powerful controllers.
Let us work together to promote this change and create a new era of open and fair artificial intelligence.
