Aging brains blend memories together instead of just forgetting them
A neuroscience study on aging suggests that older brains may increasingly blend similar memories together rather than simply losing them, prompting comparisons to “lossy compression” and hash collisions in computing. Commenters contrast normal aging with degenerative conditions like dementia, sharing personal experiences where past events are merged, reordered, or partially fabricated, and note that memory is inherently reconstructive at all ages. Others delve into neurobiology and cognitive theory, questioning simplistic “brain as computer” analogies, debating whether memory issues stem from biological aging versus capacity limits, and pointing to circadian rhythms and neural dynamics as underappreciated factors in how memories are formed, stored, and recalled.
Brain vs Computer Models
- Several comments compare “compaction” in LLMs to human forgetting, arguing LLMs only summarize context while humans lose or distort recall.
- Extended debate over whether memories are like sparse embeddings in high‑dimensional spaces with “collisions” when capacity is approached.
- Others push back, saying this is too computer-like; biological memory is distributed, dynamic, and not stored in fixed locations. Engrams drift, the same cells participate in many memories, and memory is more about evolving network dynamics.
- Discussion of non–von Neumann architectures, Hopfield networks, reservoir computing, and the idea that the brain may integrate multiple computational paradigms.
Personal Experiences with Memory Loss
- Multiple anecdotes of parents and grandparents with dementia or in memory care: timelines blended, people inserted into wrong eras, old grievances treated as current events.
- Some find brief lucid episodes especially meaningful; others describe fear, frustration, and the emotional toll of dealing with regression to earlier, sometimes abusive, personality states.
Aging, Capacity, and Neurobiology
- One line of thought: older brains may effectively be “fuller,” leading to more interference between similar memories (hash table analogy).
- A neurobiologist argues decline is more about aging processes: neuron loss from early adulthood, synapse pruning surpassing formation, and accelerated synaptic decline after ~60.
- Limited adult neurogenesis (e.g., in hippocampus) is discussed; blocking it in animal models can make representations overlap more and impair forming distinct new memories.
Circadian Rhythms and Dynamics of Memory
- Some emphasize circadian clocks as key modulators of learning, storage, and recall; aging weakens rhythm amplitude and synchrony, potentially contributing to degraded memory management.
- Broader idea that brain function is best seen as coupled oscillators and dynamical systems, not static storage.
Memory as Reconstruction and Compression
- Many note that memories are lossy, reconstructive, and change upon recall; repeated experiences get “mushed” into generic scripts, making time feel faster with age.
- People describe dream memories mistaken for real events, early memories likely replaced by reconstructions, and photographs or diaries revealing misremembered details.
- Several explicitly liken this to compression, Huffman coding, or lossy blending of similar episodes.
LLMs / AI and Article Quality
- Some call the popular article AI-generated “slop” and criticize the headline as over-claiming relative to the small, limited study.
- Others point to disclosed AI-assistance policies and human review, questioning how much to trust such secondary write-ups.