Former OpenAI Researcher Reveals 'Open Secret': Humanity is Creating AI That Cannot Be Fully Controlled

According to a report by Jinse Finance on May 13, citing Business Insider, former OpenAI researcher and current head of the AI Future Project, Daniel Kokotajlo, stated that the AI industry is racing to build systems that even the companies themselves do not fully understand or control. He explained that the core issue facing AI companies is the problem of 'alignment,' which refers to the effort to ensure that future AI systems reliably follow human instructions and values, even as their capabilities in many areas surpass those of humans. He noted that researchers currently do not fully understand how advanced AI models make decisions internally, and this uncertainty makes it difficult to ensure that future systems are truly 'aligned' and consistently execute the goals humans want them to achieve. 'This is, in some sense, an open secret, but we do not yet have a truly viable solution,' he said when discussing the implementation of AI alignment. 'Once superintelligence is created, humans will no longer be the dominant force on this planet—at least not by default.'