Revolutionizing Long-Term Memory in AI: New Horizons with High-Capacity and High-Speed Storage

Hiroaki Yamanaka, Daisuke Miyashita, Takashi Toi, Asuka Maki, Taiga Ikeda, Jun Deguchi · 2026

Read Revolutionizing Long-Term Memory in AI: New Horizons with High-Capacity and High-Speed Storage by Hiroaki Yamanaka, Daisuke Miyashita, Takashi Toi, Asuka Maki, Taiga Ikeda, Jun Deguchi (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-16192.txt.

Description

Hiroaki Yamanaka and coauthors rethink what memory should mean for artificial superintelligence. The dominant paradigm, extract then store, pulls out information judged useful from events and saves only that extract. That risks losing valuable knowledge, since material discarded as useless for one task may matter for another. The authors favor the reverse: store then on-demand extract, keeping raw events and applying them flexibly to varied tasks as needed.

Two further paths get attention: discovering deeper insights from large collections of probabilistic events, and improving collection efficiency by sharing stored events. Simple experiments support the intuitive appeal of these directions. The paper also discusses major challenges that have limited investigation into these promising paths and proposes research topics to address them. It is a design exploration, not a claim of novel methods, aimed at lifting the world with memory.

Catalog record

CREATORHiroaki Yamanaka, Daisuke Miyashita, Takashi Toi, Asuka Maki, Taiga Ikeda, Jun Deguchi
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE41 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-16192

Browse the full library · About the Tower of Babel