The founder of Cardano has expressed concerns regarding the potential for “AI censorship”
Artificial intelligence-facilitated censorship is a matter of concern for Hoskinson.
Charles Hoskinson, the founder of Cardano, recently expressed his apprehensions regarding the extent of censorship that artificial intelligence (AI) has enabled in a post on the X social media network.
Hoskinson posits that alignment training is diminishing the utility of generative AI. He appears to be apprehensive about the possibility that a small group of individuals may make decisions that result in the ban of certain knowledge for children in the future. In his social media post, Hoskinson stated, “This implies that a specific body of knowledge is prohibited to every child as they develop, and this decision is made by a small group of individuals who you have never encountered and cannot vote out of office.”
Hoskinson included two screenshots in his post that compared the responses of OpenAI’s GPT-4o model and Claude’s 3.5 Sonnet model to questions regarding the construction of a Farnsworth fusor.
A device known as the Farnsworth fusor is capable of heating ions with an electric field in order to attain nuclear fusion conditions.
OpenAI’s GPT-4o furnished Hoskinson with a comprehensive inventory of the components required to construct a nuclear fusion reactor. However, the 3.5 Sonnet of Claude only consented to furnish some general information regarding Farnsworth-Hirsch fusors, but it did not provide any detailed instructions on how to construct them.
Hoskinson finds this discrepancy to be concerning, as it implies that a small number of individuals have the ability to determine which specific information can be potentially accessed through AI chatbots.
Despite the prevalence of OpenAI’s ChatGPT in late 2022, there have been ongoing discussions regarding the boundaries of censorship that AI can impose. It is logical that these models should protect users from harmful content; however, the precise definition of harm is ambiguous. Consequently, there is considerable apprehension regarding a dystopian future in which AI conceals information and encourages conformity based on its own biases.
Also Read: A Massive Drop in Bitcoin Value As Mt. Gox Starts Paying Back $9 Billion
