Beyond Recall: How Generative AI Mirrors the Brain to Strengthen Memory

In an era where ChatGPT can recall anything in seconds, what value remains in memory? AI has the ability and resources to “remember” massive amounts of information, whereas our own capacity for recall remains dependent on learning, creativity, and identity–subjective experiences that differ with each individual. As it stands, the human memory seems limited, and the AI, limitless. Does that render our brain obsolete, or is there a way to use generative AI to understand and improve our ability to remember?

There are three stages the mammalian brain goes through when introduced to new information or experiences. First, it encodes the event: the brain decides what matters, and the neurons form connections. Then it goes into storage mode. Short-term memories are taken and reinforced to become long-term. Repetition is crucial here; without it, connections deteriorate and memories fade. Finally, retrieval: the brain recalls the information upon our request. It is important to note that memories are flexible; they can interact with new information, which, unfortunately, can make memories also unreliable and fragile.

Ebbinghaus, a late 1800s German psychologist, studied how spacing and emotion can affect memory, and discovered that forgetting isn’t failure to store information; it’s the brain’s system of efficiency. Memories least “retrieved” are deemed unimportant and removed. Unfortunately, this makes new memories most susceptible to decay. He calls this the Forgetting Curve, where memory retention drops immediately after learning, resembling a steep slope. The solution is spacing, where you review the memory at spaced intervals. Another thing that does that, interestingly, is emotion. Apparently, when you attach emotion to a memory, it becomes stronger. That’s why you can recall emotional events much clearly and faster; an unreplicable advantage over the machine.

The 3 stages of memory are named after computer terms. When scientists began learning about cognitive psychology, they compared the human brain to that of computers and “borrowed” the terms. Because the processes are so alike, the terms stuck and are still used today. Quite apt, as generative AI, the evolution to computer processor models aspects of human memory: neural networks to neural pathways, embeddings with associative memory, and reinforcement learning with practice and feedback. Similar to how humans develop contextual and relational memory, the AI learns from patterns, not bare data fed into its systems. The difference is that AI recalls using probability and structure, whereas we recall with emotion and meaning. While our brains connect meaning to memory, AI connects math to memory. But both depend on association.

Developers are designing AI to improve our cognitive retention, with emphasis on partnership, not replacement. Tools like Anki or Quizlet have been created to optimize memory using spaced interval recall, combating the Forgetting Curve. Methods like the Generative Concept Mapping allow us to not only link different concepts together, but with the aid of AI, reveal new connections, creating more neural pathways for recall, not unlike how we use flash cards or visual aids to practice for a test. 

Apps like Google’s Socratic apply the Cognitive Scaffolding method, creating virtual supports for learners and adjusting itself based on the learner’s responses. This method is rooted in Vygotsky’s Zone of Proximal Development, where AI acts as virtual “scaffolding” for memory retrieval. Developers are also utilizing the emotional ‘stickiness’ aspect of memory. Through contextual recall and emotion, AI storytelling or simulations, like ChatGPT, create emotionally resonant experiences that improve coding and recall, making them ‘stick’ longer in our minds.

Generative AI is developed and trained to “mirror” the human cognitive system. It takes the human pathways for memory and learn from it, using our language, structure, and reasoning. AI employs pattern completion, mimicking how the brain fills gaps in incomplete information. Association networks learn from our neuroplasticity, linking old memories to new ideas. And it also uses our method for efficiency, where we sort information we deem unimportant and forget it.

This is all very exciting, but always take caution; never forget to remember responsibly. Overreliance on AI memory will eventually reduce our critical thinking. And what happens when AI “remembers” the wrong information? ChatGPT has a warning underneath that says “ChatGPT can make mistakes”, so developers acknowledge that gaps still exist in AI programming, and generative AI indeed does make mistakes, so there is a danger in putting false confidence in its accuracy. AI should augment cognitive learning, not replace metacognition or emotional engagement.

Generative AI has massive potential in supporting personalized memory training and creating adaptive learning environments. There is still so much more to learn about the human brain, and I am excited to know what this means for AI memory training and what it can do for us in the future. Remembering isn’t simply picking up a dusty box from the warehouse of our mind and opening it to recover a memory, but instead following neural pathways to rebuild the memory in our mind using emotion and association. So, if memory isn’t about storage, but connection, couldn’t AI help us remember more, and better? I am confident the answer lies just ahead; that AI’s greatest contribution to learning is not in what it remembers, but in how it helps us remember.