Abstract representation of AI “self-mumbling”

A human outline in neon pink sits on a background of wire patterns in bright neon colors. Code in speech bubbles emerges from the back of the outline

Inner speech and working memory architecture boost AI performance when multitasking and completing complex pattern generation challenges.

Date:
27 January 2026
Credit:
Kaori Serakaki/OIST

Copyright OIST (Okinawa Institute of Science and Technology Graduate University, 沖縄科学技術大学院大学). Creative Commons Attribution 4.0 International License (CC BY 4.0).

Download full-resolution image
Share on:
x icon facebook icon linkedin icon mail icon

Related Images

Comparison between computational modeling and actual embryo on gene regulation in the entire fruit fly embryo

  Dr. Garth Ilsley, in Prof Nicholas Luscombe’s Genomics and Regulatory Systems Unit at OIST and Assistant Professor Angela DePace at Harvard Medical School succeeded in modelling early regulation of development in the fruit fly Drosophila using a statistical approach that accurately predicts gene expression in the whole embryo. This study was published in the open access journal, eLIFE, on August 6th. This figure is comparison of calculated results from the model (upper half) with image data from an actual embryo (lower half). It can be seen that the mutations in stripes 3 and 7 due to the misexpression of kni (A) and hb (B) are consistent. (A) Misexpression of kni is seen as a thinner stripe 3, and stripe 3 is also cut when there are 2 copies of the transgene. (B) Misexpression of hb results in bulging, bending and cutting of stripe 3. Stripe 7 bulges when there are 2 copies of the transgene. Stripe 2 remains unaffected in both cases.