Credit: SciTechDaily.comResearchers have discovered that AI memory consolidation processes resemble those in the human brain, specifically in the hippocampus, using prospective for advancements in AI and a deeper understanding of human memory mechanisms.An interdisciplinary group consisting of scientists from the Center for Cognition and Sociality and the Data Science Group within the Institute for Basic Science (IBS) exposed a striking resemblance between the memory processing of artificial intelligence (AI) models and the hippocampus of the human brain. The group applied concepts of human brain knowing, particularly concentrating on memory consolidation through the NMDA receptor in the hippocampus, to AI models.The NMDA receptor is like a clever door in your brain that facilitates knowing and memory formation. Simply like in the brain, where altering magnesium levels impact memory strength, tweaking the Transformers specifications to reflect the gating action of the NMDA receptor led to boosted memory in the AI design.”The Fusion of Cognitive Mechanisms and AI DesignWhat sets this research study apart is its effort to include brain-inspired nonlinearity into an AI construct, signifying a substantial development in simulating human-like memory combination. The merging of human cognitive mechanisms and AI design not only holds pledge for developing low-priced, high-performance AI systems however also provides valuable insights into the functions of the brain through AI models.
An interdisciplinary group has actually discovered that AI models, especially the Transformer, procedure memory in a way comparable to the human brains hippocampus. This breakthrough recommends that applying neuroscience principles, like those of the NMDA receptor, to AI can enhance memory functions, advancing the field of AI and using insights into human brain function. Credit: SciTechDaily.comResearchers have actually discovered that AI memory combination procedures resemble those in the human brain, particularly in the hippocampus, offering potential for developments in AI and a much deeper understanding of human memory mechanisms.An interdisciplinary group consisting of scientists from the Center for Cognition and Sociality and the Data Science Group within the Institute for Basic Science (IBS) revealed a striking similarity between the memory processing of synthetic intelligence (AI) models and the hippocampus of the human brain. This brand-new finding supplies a novel viewpoint on memory debt consolidation, which is a process that transforms short-term memories into long-term ones, in AI systems.Advancing AI Through Understanding Human IntelligenceIn the race towards developing Artificial General Intelligence (AGI), with influential entities like OpenAI and Google DeepMind blazing a trail, understanding and reproducing human-like intelligence has actually become an essential research study interest. Central to these technological improvements is the Transformer model [Figure 1], whose fundamental concepts are now being explored in new depth.Figure 1. (a) Diagram highlighting the ion channel activity in post-synaptic nerve cells. AMPA receptors are involved in the activation of post-synaptic neurons, while NMDA receptors are blocked by magnesium ions (Mg ² ⁺ )but cause synaptic plasticity through the increase of calcium ions (Ca ² ⁺) when the post-synaptic nerve cell is sufficiently triggered. (b) Flow diagram representing the computational procedure within the Transformer AI model. Details is processed sequentially through phases such as feed-forward layers, layer normalization, and self-attention layers. The chart portraying the current-voltage relationship of the NMDA receptors is really comparable to the nonlinearity of the feed-forward layer. The input-output chart, based on the concentration of magnesium (α), shows the modifications in the nonlinearity of the NMDA receptors. Credit: Institute for Basic ScienceThe Brains Learning Mechanisms Applied to AIThe key to effective AI systems is understanding how they discover and remember information. The group applied concepts of human brain learning, particularly focusing on memory debt consolidation through the NMDA receptor in the hippocampus, to AI models.The NMDA receptor resembles a wise door in your brain that assists in learning and memory formation. When a brain chemical called glutamate exists, the afferent neuron undergoes excitation. On the other hand, a magnesium ion serves as a small gatekeeper obstructing the door. Only when this ionic gatekeeper steps aside, compounds are enabled to stream into the cell. This is the process that permits the brain to produce and keep memories, and the gatekeepers (the magnesium ion) role in the entire procedure is quite specific.AI Models Mimicking Human Brain ProcessesThe group made an interesting discovery: the Transformer design seems to use a gatekeeping process comparable to the brains NMDA receptor [see Figure 1] This revelation led the scientists to examine if the Transformers memory consolidation can be managed by a mechanism comparable to the NMDA receptors gating process.In the animal brain, a low magnesium level is known to compromise memory function. The researchers found that long-lasting memory in Transformer can be improved by simulating the NMDA receptor. Similar to in the brain, where altering magnesium levels affect memory strength, tweaking the Transformers criteria to show the gating action of the NMDA receptor caused improved memory in the AI design. This breakthrough finding suggests that how AI models learn can be described with established understanding in neuroscience.Expert Insights on AI and NeuroscienceC. Justin LEE, who is a neuroscientist director at the institute, said, “This research makes an essential action in advancing AI and neuroscience. It enables us to dig much deeper into the brains operating concepts and develop more sophisticated AI systems based on these insights.”CHA Meeyoung, who is an information researcher in the team and at KAIST, notes, “The human brain is amazing in how it operates with minimal energy, unlike the big AI models that require immense resources. Our work opens new possibilities for affordable, high-performance AI systems that discover and remember information like people.”The Fusion of Cognitive Mechanisms and AI DesignWhat sets this study apart is its initiative to include brain-inspired nonlinearity into an AI construct, representing a considerable advancement in simulating human-like memory debt consolidation. The merging of human cognitive systems and AI style not only holds pledge for creating low-cost, high-performance AI systems but likewise provides valuable insights into the operations of the brain through AI designs.
