I listened to his podcast with Dr Alexandre Marcireau from WSU’s ICNS (International Centre for Neuromorphic Systems) a couple of weeks ago, but can’t seem to find it posted here, yet, via the search function. No mention of Brainchip, but worthwhile listening to nevertheless.
A link to the podcast transcript is also provided below.
One thing that confuses me, though, in the article & podcast below as well as in other publications, is how the term “analog” is being used here (eg “The Future of AI is analog“).
Is it correct to say in this context it doesn’t refer to the analog vs digital logic circuitry design specifically (such as Akida being fully digital vs. eg Mythic’s analog compute architecture) but rather to the general concept of the extremely power-efficient way our brain processes information (neurons working asynchronously and in parallel etc) which differs fundamentally from the way a digital computer operates on data expressed in binary code?
So is analog here essentially just being used as a synonym for neuromorphic?
So is analog here essentially just being used as a synonym for neuromorphic?
"7 Analog Computing Companies Powering Next-Generation Applications
October 16, 2023
by Quantum Business Intelligence
Analog computing, a method rooted in continuous physical phenomena like electrical voltages or mechanical movement, stands in contrast to the discrete 0s and 1s of digital computing. Historically, tools like slide rules served as rudimentary analog computers, and even water was once employed for complex economic calculations. However, the modern analog revolution is chip-based, with numerous companies delving into its potential, especially in neuromorphic computing. This approach seeks to emulate the human brain’s structure and function, using circuits to mimic neurons and synapses, offering a more efficient and parallel processing alternative to traditional digital methods.
...
BrainChip
...
In conclusion, while analog computing might seem like a relic of the past, its principles are finding new life in modern applications, especially in neuromorphic computing. The companies listed above are just a few pioneers leading the charge in this exciting field. Read another Article about
Analog Computing, what it is, and how it relates to Quantum Computing*."
https://quantumzeitgeist.com/7-analog-computing-companies-powering-next-generation-application/
* =>
"Is Analog Computing the Quiet Computing Revolution that you haven’t heard of
October 12, 2023 by Schrödinger
Though not used as commonly today due to digital calculators, slide rules are a form of analog computer that can perform multiplication, division, and other functions by sliding scales against one another. In the past, researchers have even used water to compute complex calculations for use in economics. But today’s revolution is fully chip-based, and numerous companies are exploring Analog Computing.
What is Analog Computing?
Analog computing is a method of computation that utilizes continuous physical phenomena, such as electrical voltages or mechanical movement, to represent and process information. Unlike digital computers, which use discrete values like 0s and 1s, analog computers work with continuous signals. They can solve complex mathematical equations, especially differential ones, in real time. Historically, analog computers were widely used in scientific and industrial applications, such as predicting tides, controlling machinery, and simulating flight dynamics, before the rise of digital computing. While they have been largely overshadowed by their digital counterparts, analog computing principles are still explored today in specific applications and research areas.
One of the most intriguing developments in analog computing in recent years is neuromorphic computing. Neuromorphic engineering, or neuromorphic computing, seeks to design systems that mimic the structure and function of the human brain. These systems use circuits to emulate the behavior of neurons and synapses, allowing for more efficient and parallel processing than traditional digital methods.
Use-cases for Analog Computing.
Neuromorphic systems, when related to analog computing, refer to hardware architectures specifically designed to mimic the neural structures and functionalities of the brain. These systems leverage the inherent strengths of analog computing to emulate the continuous and parallel nature of biological neural networks.
In
artificial intelligence (AI), neuromorphic systems offer a more efficient way to train neural networks. Their adaptive nature allows them to learn from data more effectively, potentially leading to faster and more accurate AI models.
An application of neuromorphic computing is in
robotics. Robots equipped with neuromorphic chips can process sensory data in real time, allowing them to adapt to their environment more effectively. This is crucial for navigation, object recognition, and interaction with humans or other robots.
Neural Network Simulations and Machine Learning
Analog computers, especially neuromorphic systems, are designed to emulate the behavior of neurons and synapses in the brain. This makes them particularly suitable for running neural network simulations and machine learning algorithms. Their inherent parallelism and energy efficiency can lead to faster processing times and lower power consumption than traditional digital systems when executing these tasks.
Signal Processing
Analog computers can be used for real-time signal-processing tasks. For instance, they can be employed in filtering noise from signals, amplifying specific frequencies, or modulating signals. Their ability to process continuous data streams in real time makes them ideal for audio processing, telecommunications, and radar systems applications.
Control Systems
Analog computers have historically been used in control system applications, such as guiding rockets or controlling industrial processes. They can simulate and predict system behaviors and provide real-time feedback to adjust parameters, ensuring optimal performance and stability.
Differential Equations Solving
Analog computers can solve differential equations much faster than their digital counterparts by directly modeling the continuous changes described by the equations. This capability is in fields like physics, engineering, and economics, where differential equations are commonly used to define system dynamics.
Optimization Problems
Analog computers can be deployed to solve optimization problems by finding the minimum or maximum values of functions. For example, they can be used in network design to find the most efficient path or in financial modeling to optimize investment portfolios.
Companies Involved in Analog Computing
Intel and Analog Computing
Intel is one of the leading companies delving into neuromorphic computing. Their research in this area has led to the developing of
“Loihi,” a neuromorphic research chip. Loihi mimics the brain’s basic computational unit, the neuron, allowing it to process information more efficiently. The chip can adapt quickly, making it suitable for complex tasks like pattern recognition and sensory data processing.
The chip is equipped with 128 cores, each containing 1,024 artificial neurons, resulting in a total of over 130,000 neurons. These neurons can be interconnected with approximately 130 million synapses. The adaptive nature of Loihi allows it to learn in real time, making it particularly effective for tasks like pattern recognition, sensory data processing, and even robotics.
Aspinity and Analog Computing
Another notable name in the field focuses on analog processing for edge devices. The company is named Aspinity. Their approach aims to reduce the power consumption of always-on sensing devices, like voice-activated assistants, by directly analyzing raw, analog sensor data. By processing data in its analog form before converting it to digital, Aspinity’s technology can drastically reduce power consumption, making it a game-changer for battery-operated devices.
The AML100 is the first product in
Aspinity’s AnalogML™ (analog machine learning) family. The AML100 detects and classifies sensor-driven events from raw, analog sensor data, allowing developers to design significantly lower-power, always-on edge processing devices. Based on the unique Reconfigurable Analog Modular Processor (RAMP™) technology platform
IBM and Analog Computing
IBM has been a pioneer in neuromorphic computing with its
TrueNorth chip. TrueNorth emulates the structure and scale of the brain’s neurons and synapses but uses significantly less power. It’s designed for various applications, including real-time processing in sensors and mobile devices.
BrainChip and Analog Computing
BrainChip is a company known for its work in neuromorphic computing. They have developed the Akida
Neuromorphic System-on-Chip, designed to provide advanced neural networking capabilities. Akida is designed for edge and enterprise applications, including advanced driver assistance systems, drones, and IoT devices. The chip’s architecture allows for low-power and low-latency processing, making it suitable for real-time applications. BrainChip’s endeavors in neuromorphic computing showcase the potential of analog computing in modern applications. Their technology aims to bridge the gap between artificial neural networks and the human brain’s functionality.
HPE and Analog Computing
HPE has been exploring the realm of neuromorphic computing through their project called the “Dot Product Engine.” This project focuses on developing hardware that can accelerate deep-learning tasks.
The Dot Product Engine uses analog circuits to perform matrix multiplications, a fundamental operation in deep learning. This approach aims to reduce power consumption and increase the speed of deep learning computations.
HPE’s exploration into analog computing signifies its commitment to finding innovative solutions to modern computational challenges. Their research could pave the way for more efficient AI hardware in the future.
MemComputing and Analog Computing
MemComputing is a company that has developed a novel computing architecture inspired by the human brain’s neurons and synapses. Their technology is designed to solve complex optimization problems.
Their approach involves using memory elements for computation and storage, which can lead to significant speed-ups for specific computational tasks. This is particularly beneficial for industries that require real-time decision-making based on large datasets. MemComputing’s technology showcases the potential of neuromorphic architectures in addressing some of the most challenging problems in computing. Their solutions aim to provide a competitive edge to businesses across various sectors.
Applied Brain Research (ABR) and Analog Computing
Applied Brain Research (ABR) is a company that specializes in neuromorphic engineering and software. They have developed tools and software for building brain-like computing systems.
Their software, called Nengo, is a neural simulator for designing and testing large-scale brain models. It’s used in various applications, from robotics to AI. ABR’s work in neuromorphic computing is centered around creating efficient, brain-like systems that can process information in real time, making them suitable for a range of applications where traditional computing architectures might fall short.
Knowm and Analog Computing
Knowm is a company that focuses on developing memristor-based solutions for neuromorphic computing. Memristors are electronic components that can change resistance based on the amount and direction of voltage applied, making them suitable for brain-like computing.
Knowm’s technology aims to provide a platform for building adaptive learning systems that can evolve and learn over time. Their approach to neuromorphic computing is hardware-centric, focusing on creating components that can support brain-like computation at the chip level.
CogniMem and Analog Computing
CogniMem specializes in pattern recognition and has developed technologies based on neuromorphic computing principles. Their products are designed to mimic the human brain’s ability to recognize patterns and learn from experience. This makes them suitable for applications like image and speech recognition. CogniMem’s vision is to create computing systems that can learn and adapt in real time, providing more natural and intuitive interactions between humans and machines.
Neurala
Neurala is a company that develops deep-learning neural network software for drones, cameras, and other devices. Their Neurala Brain technology is designed to make devices like drones more autonomous and capable of learning and adapting in real time. Neurala’s approach to neuromorphic computing is to create software that can be integrated into various devices, making them smarter and more responsive to their environments.
Analog Computing and Quantum Computing?
Analog computers operate based on continuous variables and physical phenomena. They represent data as varying physical quantities, such as electrical voltage or fluid pressure. For instance, in an electronic analog computer, the magnitude of an electrical voltage might represent a specific value. Calculations are performed by manipulating these continuous signals through components like operational amplifiers and integrators. In contrast, quantum computers operate on the principles of quantum mechanics. They utilize qubits, which can exist in a superposition of states, allowing them to represent both 0 and 1 simultaneously. Quantum operations involve entanglement and superposition, which Analog systems cannot achieve. One commonality, of course, is that the states are not digital, although they will be typically converted back to digital states to interpret—for example, the words in a document or images.
Modern analog computers, such as neuromorphic chips, are designed to mimic the brain’s neural structures, offering efficient solutions for tasks like pattern recognition and sensory processing3. Quantum computers, on the other hand, have the potential to revolutionize fields like cryptography, optimization, and drug discovery. They can solve problems currently intractable for classical computers, such as factoring large numbers or simulating complex quantum systems."
https://quantumzeitgeist.com/analog-computing-computing-revolution/
@Diogenese can you or someone else explain to me, amateur, why exactly Brainchip speaks of fully digital and in such a way that I understand it in my lack of knowledge? If this has already been clarified please excuse me and point me to the corresponding post.