Cliste → Sophia
Machine cognition,
guided by the brain.
I’m building Cliste to explore how perception, memory, and learning can work together in a continuing mind. Vector symbolic architectures provide a way to combine experiences with spatial and temporal context.
From the visual-memory experiments
A fragment recalls a stored scene.
Pixels become vector memories that can be queried and decoded back into images. No captions or text queries are involved.
Stored earlier
The full photograph
Later cue
Gray marks hidden pixels
Recalled memory
Decoded from stored vectors
Partial recall
66/192 → 111/192
Correct recalls with the adjusted rule. False matches also increased.
Recall and errors →Remembered sequences
25/128 → 60/128
Correct image and two later encounters with the adjusted recall rule, across 32 stored photographs.
Sequence results →Vector storage
71% smaller
Eight-bit memory tensors; all 192 tested recall choices unchanged.
Storage comparison →