A new arXiv position paper argues that large language models, despite their impressive capabilities, are missing a key ingredient for artificial general intelligence: explicit memory. The paper's title states the thesis directly: hippocampal explicit memory is the cornerstone for AGI.

The authors contend that while LLMs have sparked hopes for AGI, their architecture does not naturally support the kind of explicit, episodic recall that the hippocampus provides in biological brains. They argue that integrating this form of memory into AI systems is necessary for genuine general intelligence, rather than just pattern completion.

Because the paper is a position piece, it offers an argument rather than experimental results. The abstract is brief, but the title and framing make clear that the authors see explicit memory as a missing pillar in current AI research.