Neuromorphic Computing Innovations in AI

What if neuromorphic computers used neurotransmitters? In the latest issue of Device, Kai Xiao & co-workers summarize progress in using dopamine as a signaling modality in bioinspired computing that improves artificial intelligence applications by better emulating the human brain! Dopamine detection and integration in neuromorphic devices for applications in artificial intelligence by Kai Xiao and co-workers Link (OA): https://lnkd.in/gKKD9HxM The bigger picture: Neuromorphic devices play a pivotal role in reshaping artificial intelligence, brain-like computing, and neuroprostheses by emulating the intricate computational processes of the human brain. However, current neuromorphic technologies primarily rely on electrical signals, overlooking the indispensable role of neurotransmitter-mediated chemical signaling in biological organisms. This review explores methodologies for detecting the key neurotransmitter dopamine, including microdialysis, optical, and electrochemical techniques. The exploration extends to the integration of these detection methods with neuromorphic devices, highlighting the imperative to comprehend both electrical and chemical signaling for precise emulation of neural processes. By bridging this gap, we can unlock the full potential of neuromorphic technology, resulting in more sophisticated and human-like artificial intelligence systems.

An exciting week for #neuromorphic computing and decreasing the compute power required for #AI and #ML! For more on this topic, see my previous post: https://lnkd.in/g3EeG3Ku https://lnkd.in/gia2EVK2 Researchers report the creation of the first #roomtemperature, #lowpower (20 pW) moiré #synaptic #transistor. It is #graphene based. "The asymmetric gating in dual-gated moiré heterostructures realizes diverse biorealistic neuromorphic functionalities, such as reconfigurable synaptic responses, spatiotemporal-based tempotrons and Bienenstock–Cooper–Munro input-specific adaptation. In this manner, the moiré synaptic transistor enables efficient compute-in-memory designs and #edgehardware accelerators for #artificialintelligence and #machinelearning. Key points: Design and Material Composition: The synaptic transistor is designed to mirror human brain function, in its ability to process and store information concurrently. This mimics the brain's capability for higher-level cognition. The transistor combines two atomically thin materials – bilayer #graphene and hexagonal boron nitride – arranged in a moiré pattern to achieve its #neuromorphic functionality. This innovative structure enables the device to perform associative #learning and recognize patterns, even with imperfect input. Cognitive Functionality: The device’s ability to perform associative learning and pattern recognition, even with imperfect inputs, represents a step towards replicating higher-level cognitive functions in artificial intelligence systems. This research provides a foundation for the development of more efficient, brain-like AI systems, potentially transforming how information processing and memory storage are approached in silico. Operational Stability and Efficiency: Unlike previous brain-like computing devices that required #cryogenic temperatures to function, this new device operates stably at room temperature. It demonstrates fast operational speeds, low energy consumption, and the ability to retain stored information even when power is removed, making it highly applicable for real-world use. Implications for AI and ML: This highlights a shift away from traditional #transistor-based computing towards more energy-efficient and capable systems for AI and ML tasks. This development addresses the high power consumption issue prevalent in conventional #digitalcomputing systems, where separate processing and storage units create bottlenecks in data-intensive tasks. Original article in Nature: https://lnkd.in/gSvyUyYK

Can neuromorphic computing overcome power and latency constraints that currently limit deployment of multiple real-world, real-time AI capabilities? A recent Intel Labs paper published at ICASSP 2024 found that new neuromorphic approaches using Intel's Loihi 2 can provide orders of magnitude gains in combined efficiency and latency for feed-forward and convolutional neural networks in video, audio denoising, and spectral transforms compared to state-of-the-art solutions. The Neuromorphic Computing Lab at Intel Labs found that several uniquely neuromorphic features enable these gains, such as stateful neurons with diverse dynamics, sparse yet graded spike communication, and an architecture that integrates memory and compute with highly granular parallelism to minimize data movement. The team characterized and benchmarked sigma-delta encapsulation, resonate-and-fire neurons, and integer-valued spikes as applied to standard video, audio, and signal processing tasks. In some cases, the gains exceeded three orders of magnitude, but often at the cost of lower accuracy. Read the paper here: https://lnkd.in/gdT6X4UP #iamintel #Neuromorphic #ArtificialIntelligence #LLM #GenerativeAI

On-Edge: Neuromorphic Computing for Psychiatric Biophysical Modeling. The article in the comments presents a brain-inspired platform for real-time dynamic computing of Spiking Neural Networks (SNNs) using asynchronous sensing in a neuromorphic chip. It highlights the growing need for edge-computing, i.e., processing data near the sensors. This approach, inspired by the biological nervous system, promises always-on processing of sensory signals, supporting on-demand, sparse, and edge-computing. The system emulates dynamic and realistic neural processing phenomena like short-term plasticity, NMDA gating, AMPA diffusion, homeostasis, spike frequency adaptation, conductance-based dendritic compartments, and spike transmission delays. The analog circuits implementing these primitives are paired with low-latency asynchronous digital circuits for routing and mapping events. This asynchronous infrastructure allows defining different network architectures and provides direct event-based interfaces to convert and encode data from event-based and continuous-signal sensors. The article also discusses the system’s architecture, characterizes the mixed-signal analog-digital circuits that emulate neural dynamics, demonstrates their features with experimental measurements, and presents a software ecosystem for configuring the system. The system’s flexibility to emulate different biologically plausible neural networks and its ability to monitor both population and single neuron signals in real-time allow for the development and validation of complex models of neural processing for both basic research and edge-computing applications. Neuromorphic computing has several potential applications in psychiatry, including real-time data analysis at the edge of the web, human-like cognitive computing, and the use of models for EEG signals to better understand brain activity patterns associated with mental health disorders. It can also be used in robotics to develop intelligent therapeutic robots that interact with patients empathetically. By processing, analyzing, and multiplexing large amounts of multimodal data, neuromorphic computing can help develop personalized treatment plans based on a patient’s unique genetic makeup, lifestyle, and environmental factors. As we gain deeper insights into the human brain and neuromorphic computing, we can expect more innovative applications in psychiatry. The work highlighted here demonstrates the use of advanced spiking neuron models for efficient data processing, emphasizing the potential of neuromorphic computing in advancing psychiatry and contributing to the broader field of neuromorphic computing, with the promise of achieving AI with lower energy needs.