Ilya Sutskever, one of the co-founders of OpenAI, recently spoke at an important annual conference on artificial intelligence (AI) called NeurIPS. During his talk, he discussed various topics related to AI and received an award for his significant contributions to the field.
One of the main ideas he talked about was the concept of “superintelligent AI.” This term refers to a type of AI that would be much smarter and more capable than humans in many areas. Sutskever believes that we will eventually create this kind of superintelligent AI, and he emphasized that it will be very different from the AI we have today. In fact, he suggested that it might be so advanced that we might not even recognize it as AI.
Read Also: “X Gets A Quicker Grok Model And A New ‘Grok Button'”
Sutskever explained that these superintelligent systems will be “agentic,” which means they will act more like independent agents rather than just following programmed instructions. In contrast, the AI we have now is only “very slightly agentic,” meaning it can perform tasks but doesn’t really think or reason like a human. He believes that future AI will be able to reason and make decisions based on limited information, which could make them more unpredictable.
Interestingly, Sutskever also mentioned that these advanced AIs might develop a sense of self-awareness and even desire certain rights. He suggested that it wouldn’t be a bad outcome if superintelligent AIs simply wanted to coexist with humans and have their own rights.
After his time at OpenAI, Sutskever started a new organization called Safe Superintelligence (SSI). This lab is dedicated to ensuring that general AI is developed safely. Recently, SSI managed to raise a substantial amount of funding—$1 billion—in September to support its mission.
