MIT researchers have come up with a new idea about memory that highlights the role of astrocytes, a type of brain cell that’s often overlooked.The human brain is packed with around 86 billion neurons. These neurons send electrical signals, helping us store memories and communicate throughout the brain and nervous system.But there are also billions of other cells called astrocytes. These star-shaped cells have tons of long, branching arms that let them connect to millions of neurons. For a long time, astrocytes were thought to only play a supporting role—like cleaning up or feeding neurons. But recent research is suggesting that astrocytes might actually help with memory storage and other brain functions too.MIT researchers have now put forward a new theory on how astrocytes could be crucial for storing memories. Their model suggests that these cells might play a major role in the brain’s huge memory capacity, which is way more than you’d expect if you only count neurons.“Originally, astrocytes were believed to just clean up around neurons, but there’s no particular reason that evolution did not realize that, because each astrocyte can contact hundreds of thousands of synapses, they could also be used for computation,” says Jean-Jacques Slotine, an MIT professor of mechanical engineering and of brain and cognitive sciences, and an author of the new study. Dmitry Krotov, a researcher at the MIT-IBM Watson AI Lab, is the lead author of the paper, which came out in the Proceedings of the National Academy of Sciences. Leo Kozachkov, who earned his PhD at MIT in 2022, is the study’s first author.
How astrocytes help memory
Astrocytes do a lot of supporting jobs in the brain. They clean up debris, provide nutrients, and make sure neurons get enough oxygen. They also have these tiny “tentacles” that wrap around synapses—the points where neurons connect with each other. This forms what’s called a “tripartite synapse,” involving three cells: two neurons and one astrocyte.Unlike neurons, astrocytes don’t send electrical impulses, but they communicate through calcium signals. Scientists have found that these signals let astrocytes sync up with neuron activity. When astrocytes pick up on neural activity, they can change their calcium levels, which can then trigger the release of chemicals called gliotransmitters into the synapse.So, the MIT team decided to dig deeper and build a model to figure out how these connections work and how they might help store memories. They used a concept called Hopfield networks, a type of artificial neural network that can store and recall patterns.
New model for more memory storage
Hopfield networks have been used to try to explain how the brain works, but they just don’t have the capacity to match the massive memory storage of the human brain. So the researchers turned to a newer version of the Hopfield network called dense associative memory, which can store way more information by connecting more neurons in a complicated way.But there’s a problem: normal synapses only connect two neurons, so how could the brain possibly make all these many connections? That’s where astrocytes come in.In the new model, astrocytes help store more info than traditional models can by using complex connections between many neurons. This model has the potential to explain the brain’s huge memory capacity.
What’s next?
To test if this model could be how the brain stores memories, scientists would need to figure out how to mess with the connections between astrocytes and see if that affects memory.This model also has potential applications in AI, especially in helping researchers build smarter, more energy-efficient models. By tweaking how astrocytes connect, we could develop AI systems that function more like the brain. Slotine points out that while AI was originally inspired by neuroscience, recent AI models have drifted away from biology—this research could help bring them back together.So, this could be a big step toward understanding how our brains store so much info—and maybe even making AI smarter by copying the brain’s strategies