Perfect memory, powerful computing ability, and outstanding intelligence: we all want a brain like that, but how do we design such a brain? The real brain consists of approximately 80 billion neurons, which coordinate their work through thousands of synaptic connections. The human brain does not have a centralized processor like a standard laptop. Instead, many computations run in parallel and the results are compared. Although the operation principles of the human brain are not fully understood, existing mathematical algorithms can transform the principles of deep learning into a system that is more similar to the human brain. This brain-inspired computing paradigm, called Spiking Neural Networks (SNN), provides a computational architecture that matches the potential advantages of systems that use both optical and electronic components.
In Spiking Neural Networks (SNN), information is processed in the form of spikes or action potentials, which are the electrical signals generated when real neurons fire. One key characteristic is that they use asynchronous processing, which means that spikes are processed in the order of their occurrence in time, rather than in batches like traditional neural networks. This allows SNN to respond to inputs faster and perform certain types of computations more efficiently than traditional neural networks.
SNN also enables certain neural computations that are difficult or impossible to achieve in traditional neural networks, such as temporal processing and Spike-Timing-Dependent Plasticity (STDP). STDP is a form of Hebbian learning theory that allows neurons to change their synaptic connections based on the timing of their spikes. (Hebbian learning theory can be summarized as "neurons that fire together, wire together" and it describes the plasticity of the brain's learning ability using mathematical models.)
A recent paper published in the IEEE Journal of Quantum Electronics describes the development of an SNN device that integrates optoelectronic neurons, analog circuits, and a Mach-Zehnder interferometer grid. These grids are optical circuit components that can perform matrix multiplication, similar to how synaptic networks of neurons operate in the human brain.
The authors demonstrate that optoelectronic neurons can receive inputs through an optical communication network, process information through analog circuits, and communicate with the network through lasers. This process allows for faster data transmission and inter-system communication compared to traditional electronic systems.
The paper also describes the creation of brain-inspired computing systems using existing algorithms, such as random backpropagation and contrastive Hebbian learning. These algorithms enable the system to learn from the information of each synapse, similar to the human brain, and have significant advantages in terms of computational performance compared to traditional machine learning systems that use backpropagation.
In the field of artificial intelligence and machine learning, SNN has several advantages over modern computing paradigms in simulating tasks under natural evolutionary conditions. Because SNN processes data continuously over time, they are well-suited for real-time applications that present a single case, such as event-driven signal processing.
Furthermore, the propagation of information over time allows for multiple forms of memory at different time scales, similar to how humans distinguish between working, short-term, and long-term memory. SNN finds common applications in neural morphological perception and robotics. For example, an adaptive robot arm controller can provide reliable motor control even when the actuators wear out.
More likely, future devices may leverage these characteristics for operations in the context of speech assistants, real-time captioning services, or audio separation in speech and natural language processing. Similarly, SNN can be used for real-time video and lidar processing in autonomous vehicles or monitoring systems.
Source: https://techxplore.com/news/2023-05-team-brain-inspired-device-optoelectronic.html
Alphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Anthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Apple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Google launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
YouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Google grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Mistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.