The lunar material touched down in China’s Inner Mongolia Autonomous Region early Tuesday morning (June 25).
Computers have come so far in terms of their power and potential, rivaling and even eclipsing human brains in their ability to store and crunch data, make predictions and communicate. But there is one domain where human brains continue to dominate: energy efficiency.
“The most efficient computers are still approximately four orders of magnitude — that’s 10,000 times — higher in energy requirements compared to the human brain for specific tasks such as image processing and recognition, although they outperform the brain in tasks like mathematical calculations,” said UC Santa Barbara electrical and computer engineering Professor Kaustav Banerjee, a world expert in the realm of nanoelectronics. “Making computers more energy efficient is crucial because the worldwide energy consumption by on-chip electronics stands at #4 in the global rankings of nation-wise energy consumption, and it is increasing exponentially each year, fueled by applications such as artificial intelligence.” Additionally, he said, the problem of energy inefficient computing is particularly pressing in the context of global warming, “highlighting the urgent need to develop more energy-efficient computing technologies.”
Neuromorphic computing has emerged as a promising way to bridge the energy efficiency gap. By mimicking the structure and operations of the human brain, where processing occurs in parallel across an array of low power-consuming neurons, it may be possible to approach brain-like energy efficiency.
Computer vision, one of the major areas of artificial intelligence, focuses on enabling machines to interpret and understand visual data. This field encompasses image recognition, object detection, and scene understanding. Researchers continuously strive to improve the accuracy and efficiency of neural networks to tackle these complex tasks effectively. Advanced architectures, particularly Convolutional Neural Networks (CNNs), play a crucial role in these advancements, enabling the processing of high-dimensional image data.
One major challenge in computer vision is the substantial computational resources required by traditional CNNs. These networks often rely on linear transformations and fixed activation functions to process visual data. While effective, this approach demands many parameters, leading to high computational costs and limiting scalability. Consequently, there’s a need for more efficient architectures that maintain high performance while reducing computational overhead.
Current methods in computer vision typically use CNNs, which have been successful due to their ability to capture spatial hierarchies in images. These networks apply linear transformations followed by non-linear activation functions, which help learn complex patterns. However, the significant parameter count in CNNs poses challenges, especially in resource-constrained environments. Researchers aim to find innovative solutions to optimize these networks, making them more efficient without compromising accuracy.
The brain is the most complex organ ever created. Its functions are supported by a network of tens of billions of densely packed neurons, with trillions of connections exchanging information and performing calculations. Trying to understand the complexity of the brain can be dizzying. Nevertheless, if we hope to understand how the brain works, we need to be able to map neurons and study how they are wired.
Japanese researchers have developed a novel technique to attach engineered skin tissue to humanoid robots.
Robotic platforms may benefit from enhanced mobility, embedded sensing capabilities, self-healing capabilities, and a more realistic appearance.
The innovation was made possible by mimicking skin-ligament structures and using V-shaped perforations in a robot face.
Physics meets machine learning.
The Progress and Promise for Science in Indonesia Regional Special Feature focuses on biodiversity and climate change, highlighting research based on the unique geology and biology of a nation comprising more than 17,600 islands, containing about 10 percent of the world’s remaining tropical forests, and home to over 300,000 species of wildlife.
Organic electrochemical artificial neurons (OANs) are the latest entry of building blocks, with a few different approaches for circuit realization. OANs possess the remarkable capability to realistically mimic biological phenomena by responding to key biological information carriers, including alkaline ions, noise in the electrolyte, and biological conditions. An organic artificial neuron with a cascade-like topology made of OECT inverters has shown basic (regular) firing behavior and firing frequency that is responsive to the concentration of ionic species (Na+, K+) of the host liquid electrolyte33. An organic artificial neuron consisting of a non-linear building block that displays S-shape negative differential resistance (S-NDR) has also been recently demonstrated34. Due to the realization of the non-linear circuit theory with OECTs and the sharp threshold for oscillations, this artificial neuron displays biorealistic firing properties and neuronal excitability that can be found in the biological domain such as input voltage-induced regular and irregular firing, ion and neurotransmitter-induced excitability and ion-specific oscillations. Biohybrid devices comprising artificial neurons and biological membranes have also shown to operate synergistically, with membrane impedance state modulating the firing properties of the biohybrid in situ. More recently, a circuit leveraging the non-linear properties of antiambipolar OMIECs, which exhibit negative differential transconductance, has been realized35. These neurons show biorealistic properties such as various firing modes and responsivity to biologically relevant ions and neurotransmitters. With this neuron, ex-situ electrical stimulation has been shown in a living biological model. Therefore, the class of OANs perfectly complements the broad range of features already demonstrated by solid-state spiking circuits (Supplementary Table 1), offering opportunities for both hybrid interfacing between these technologies and new developments in neuromorphic bioelectronics.
Despite the promising recent realizations of organic artificial neurons, all approaches still remain in the qualitative demonstration domain and a rigorous investigation of circuit operation is still missing. Indeed, quantitative models exist only for inorganic, solid-state artificial neurons without the inclusion of physical soft-matter parameters and the biological wetware (i.e., aqueous electrolytes, alkaline ions, biomembranes)36,37. This gap in knowledge significantly impedes the simulation of larger-scale functional circuits, and therefore the design and development of integrated organic neuromorphic electronics, biohybrids, OAN-based neural networks, and intelligent bioelectronics.
In this work, we unravel the operation of organic artificial neurons that display non-linear phenomena such as S-shape negative differential resistance (S-NDR). By combining experiments, numerical simulations of non-linear iontronic circuits, and newly developed analytical expressions, we investigate, reproduce, rationalize, and design the wide biorealistic repertoire of organic electrochemical artificial neurons including their firing properties, neuronal excitability, wetware operation, and biohybrid formation. The OAN operation is efficiently rationalized to include how neuronal dynamics are probed by biochemical stimuli in the electrolyte medium. The OAN behavior is also extended on the biohybrid formation, with a solid rationale of the in situ interaction of OANs with biomembranes. Non-linear simulations of OANs are rooted in a physics-based framework, considering ion type, ion concentration, organic mixed ionic–electronic parameters, and biomembrane properties. The derived analytical expressions establish a direct link between OAN spiking features and its physical parameters and therefore provide a mapping between neuronal behavior and materials/device parameters. The proposed approach open opportunities for the design and engineering of advanced biorealistic OAN systems, establishing essential knowledge and tools for the development of neuromorphic bioelectronics, in-liquid neural networks, biohybrids, and biorobotics.
One of the ambitions of computational neuroscience is that we will continue to make improvements in the field of artificial intelligence that will be informed by advances in our understanding of how the brains of various species evolved to process information. To that end, here the authors propose an expanded version of the Turing test that involves embodied sensorimotor interactions with the world as a new framework for accelerating progress in artificial intelligence.
Human visual neurons rely on event-driven, energy-efficient spikes for communication, while silicon image sensors do not. The energy-budget mismatch between biological systems and machine vision technology has inspired the development of artificial visual neurons for use in spiking neural network (SNN). However, the lack of multiplexed data coding schemes reduces the ability of artificial visual neurons in SNN to emulate the visual perception ability of biological systems. Here, we present an artificial visual spiking neuron that enables rate and temporal fusion (RTF) coding of external visual information. The artificial neuron can code visual information at different spiking frequencies (rate coding) and enables precise and energy-efficient time-to-first-spike (TTFS) coding. This multiplexed sensory coding scheme could improve the computing capability and efficacy of artificial visual neurons. A hardware-based SNN with the RTF coding scheme exhibits good consistency with real-world ground truth data and achieves highly accurate steering and speed predictions for self-driving vehicles in complex conditions. The multiplexed RTF coding scheme demonstrates the feasibility of developing highly efficient spike-based neuromorphic hardware.
A spiking neural network that is based on event-driven vision sensors can be created using two parallel photodiodes of opposite polarities that output programmable spike signal trains in response to changes in light intensity.
