The 10 most noteworthy players in brain-inspired computing｜Qubit Think Tank Report (attached download)

Qubit think tank from Ao Fei Si
Qubit | Public account QbitAI

Why should we pay attention to brain-like computing?

Because as the amount of data increases, current computer systems have faced three major challenges :

"Integration density" problem, "storage wall" problem and "power wall" problem.

In order to solve the above dilemma, some people naturally thought of drawing on a series of advantages of the human brain such as low power consumption, high energy efficiency, and self-learning to develop a new generation of brain-like intelligent systems.

In fact, based on scientific research, the industry has been exploring the implementation direction of related technologies in recent years:

Currently, about 20 companies focusing on commercialization of brain-inspired computing have been born around the world.

After completing the "In-depth Industry Report on Brain-like Computing" , QuantumBit Think Tank has combined the integrity, core technology and commercialization process of brain-like products to provide 7 brain-like companies worth paying attention to around the world, as well as 3 research centers in China that are closely related to brain-like intelligence .

Let’s watch it together.

7 brain-inspired companies around the world worth paying attention to

1. Lingxi Technology

2. Shishi Technology

Date of establishment: 2017
Financing round: Pre B round
Investment amount: nearly 200 million yuan
Investors: Qigang Investment, Zhangjiang Science and Technology Investment, CEC Hikvision, China Merchants Qihang, Tykeyuan Electronics, Ventech China, Heli Capital , Yachang Investment
Products:
"Sensing and computing integrated" dynamic vision intelligence SoC Speck: large-scale spiking convolutional neural network
(sCNN) chip architecture based on asynchronous logic paradigm;
DYNAP: dynamic vision dedicated processor;
XYLO: based on spiking neural network, Ultra-low power consumption (~0.1mW), always-on low-dimensional signal dedicated AI processor.
Highlights: Integrated sensing and computing, using DYNAP-CNN brain-like processor core, integrating 330,000 pulse neurons, increasing neuron integration by 30-40 times, supporting complex deep pulse neural networks, and system sensing and computing power consumption is less than 1mW.

3. Tashan Technology

Established: 2017
Financing round: Pre A round
Investors: SoftBank China, Mumeng Group
Products:
Distributed touch-sensitive chips: digital-analog hybrid AI touch-sensitive chips, and distributed brain-like industrial application chips based on SNN.
Highlights: Distributed brain-like touch-sensitive chip architecture; partitioned and hierarchical pulse neural network algorithm architecture

4. U-Intelligence Innovation

Date of establishment: 2021
Financing rounds: Angel +
Financing amount: tens of millions of yuan
Investor: Tongwei Capital
Product:
Sisi No. 1 general-purpose brain-like reasoning chip: 8-core "Friendly" processor + AI acceleration + CLAS causal learning Algorithm, power consumption is no more than
2 watts, and can calculate 1 million neurons at the same time;
"Silicon Brain" fully autonomous unmanned system platform: embedded with the self-developed CLAS causal learning algorithm system and brain-like chip, using
small sample recognition algorithm , fully autonomous real-time path planning algorithm and fully autonomous decision-making algorithm, realizing
functions such as environmental perception, environmental cognition, and brain-like decision-making, and can be widely used in fields such as the metaverse, AIGC, drones, unmanned driving, and robots. Currently, the fully autonomous unmanned flying brain-inspired computing box K50/K51 is launched.
Highlights: Universal scene extraction algorithm in imperfect information environment;
Impulse neural network based on causality;
"Structure Growth Algorithm" reinforcement learning algorithm framework for decision-making with imperfect information.

5. Prophesee

Founded: 2014
Financing round: C round
Investors: 360 Capital Partners, European Investment Bank, iBionext, Weihao Chuangxin, Intel Capital, Renault Group, Bosch Ventures, Sinovation Ventures, Supernova Invest, Xiaomi, etc.
Products:
METAVISION Vision Sensor: The fourth-generation dynamic vision sensor is jointly developed by Prophesee and Sony, combining Sony's CMOS image sensor technology with Prophesee's unique event-based Metavision® sensing technology;
METAVISION Intelligence Kit: Contains 95 algorithms, 67 code samples and 11 ready-to-use applications; full interoperability between open software and hardware.
Highlights: Event-based dynamic visual information acquisition.

6. GrAI Matter Labs

Founded: 2016
Financing round: A+ round
Investors: iBionext, 360 Capital, 3T Finance, etc.
Products:
NeuronFlow: sparse processing; high-precision data flow architecture for efficient fine-grained parallel computing;
GrAI VIP: a SOC that integrates the neuron engine GrAICore;
GraiFlow SDK: compatible with common ML frameworks, TensorFlow and PyTorch to implement custom models; also provides a database of pre-trained models.
Highlights: high-precision (16-bit floating point) processing to provide high-quality content; dynamic data flow to exploit the sparsity of data; neuromorphic design ideas to improve efficiency; in-memory computing to reduce power consumption and latency.

7. Brainchip

Date of establishment: 2004
Financing round: Australian stock market listing
Products:
Akida (neuromorphic processor): has functions such as simultaneous processing of multi-sensor modes, on-chip one-time learning, event-based learning, and cloud independence;
MetaTF: used for neural networks Creation, training and testing to support the development of edge AI systems on Brainchip's Akida.
Highlights:
Distributed computing: each NPU has dedicated computing and memory, reducing data movement;
event-based processing: NPU is event-driven;
event-based communication: sending events through the mesh network without CPU intervention;
event-based learning: on-chip Learning algorithms.

Representative brain-like intelligence research institution in China

Tsinghua University
Brain-inspired Computing Research Center

Founded: 2014

The research center is a multidisciplinary center consisting of seven departments: computer science, microelectronics, electronics, automation, precision instruments, materials, and bioengineering. It is currently the only team in the country that conducts comprehensive research on brain-like computing systems from basic theory, chips, software environment, and applications.

Some achievements:

Three Nature papers were published consecutively in 2019, 2020 and 2021, respectively, conducting research from three aspects: architecture, software compilation and algorithm model.

The first paper talks about the design of brain-inspired chips with heterogeneous fusion architecture and how to support the integration of ANN and SNN at the architectural level;

The second article describes the architecture of the brain-inspired compilation software system and proposes the completeness of brain-inspired computing;

The third brain-like algorithm article published in Nature Communications is the first fusion algorithm that combines global learning and local learning to provide support for the training and online learning mechanism of brain-like fusion algorithm models.

Zhejiang University
Center for Frontier Sciences of Brain and Brain-Machine Integration (abbreviated as "Dual Brain Center")

Founded: 2018

The research groups include: School of Brain Science and Brain Medicine, Institute of Systems Neuroscience and Cognitive Science, Qiushi Institute for Advanced Studies, School of Biomedical Engineering and Instrument Science, School of Computer Science and Technology, School of Information and Electronic Engineering, and Department of Psychological and Behavioral Sciences.

research direction:

• Molecular and cellular biological mechanisms unique to the nervous system

• Circuit Analysis of Brain Function

• Mechanistic research and intervention measures for brain diseases

• Brain higher cognitive functions and computational models

• Brain-computer interface and hybrid intelligence

• Brain-like computing and neural computing

Some achievements:

• Darwin Brain-inspired Chip (1st & 2nd Generation)

• Billion-level neuron brain-like computer

Chinese Academy of Sciences
Brain-like Intelligence Research Center

Date of establishment: 2015

research team:

• Cognitive Brain Modeling Group: Simulating the cognitive brain at multiple scales to explore and understand how the brain works and how to develop intelligent systems inspired by the brain;

• Brain-like Information Processing Group: The current goal is to build a multimodal cognitive machine with vision, hearing, language processing, and thinking capabilities;

• Neurorobotics Group: Researching and developing next generation intelligent robotic systems by imitating and learning the mechanics of the human nervous system.

research direction:

• Multiscale brain simulation

• Cognitive Brain Modeling and Neural Computing

• Reconstructing the brain at the microscopic level

• brain-inspired computing

• Brain Knowledge Base

• Cognitive and neurorobotics