Core Assets in the AI Era: Memory and Abstract Thinking

Jack Kong, CEO of Nano Labs, posted on X about the evolving landscape of AI technology, emphasizing that the most crucial assets in the AI era are memory and abstract thinking abilities. He highlighted an intriguing architectural design that combines local memory with cloud-based large language models (LLM). The core logic of this design is that memory, as a vital digital asset for individuals, should not be entirely entrusted to the cloud. This design, which follows the sequence of input to local memory, filtering layer, cloud LLM, audit, and output, essentially constructs a decentralized personal AI system. In this system, complete memory is stored locally, while the cloud processes only the filtered context. This approach aligns with the core principles of Web3, which advocate for data sovereignty returning to individuals. Kong suggests that the future of AI infrastructure may see widespread application of such hybrid architectures, where personal data is managed locally, and only necessary information is processed in the cloud. This model could redefine how AI systems are structured, ensuring greater control and privacy for users.