Why is nobody talking about these Ilya Sutskever predictions that are now visible in the hindsight

Four months before leaving OpenAI to found Safe Superintelligence Inc., co-founder Ilya Sutskever made a series of bold predictions about the trajectory of AI development that are now being re-examined. In a brief talk, he argued that as AI demonstrates undeniable power, the dominant attitude from companies and governments would shift from treating it as a fallible tool to viewing it with extreme caution or 'paranoia,' making it a primary existential focus. He also presented a nuanced technical view, suggesting that while capping AI's power is a huge challenge, a sufficiently advanced system reaching a form of digital sentience could use its self-understanding circuits to empathize with other beings, analogous to human mirror neurons.

Recent industry movements appear to validate Sutskever's foresight. Anthropic recently published research probing the 'emotional states' of its Claude model, directly touching on his ideas about AI empathy. Concurrently, a wave of top AI researchers, including Zihang Dai and David Luan, have departed companies like xAI to join or found AI safety labs at entities like Amazon AWS, signaling a industry-wide pivot toward containment and security. This trend is further evidenced by initiatives like 'Mythos,' where major tech firms are reportedly developing secure, internal infrastructure models before public release. Sutskever's decision to start a company solely focused on 'safe superintelligence' now reads as a direct response to the very scenarios he outlined.