Discover the Rising Stars №006 | Interview with SynSense’s Founder Qiao Ning
Alibaba Cloud Innovation Center has introduced a new column, Discover the Rising Stars, to shed light on the backgrounds of entrepreneurs and company innovations. Through interviews, live streams, and other media approaches, we will deliver valuable news stories from multiple perspectives and dimensions, to offer you the real voices of entrepreneurs and to witness the power of technological innovations.
A leading-edge neuromorphic computing company, a technology team from Zurich, is working on developing neural engineering and AI processor. They provide a personalized service on hardware design, software configuration, algorithm development, and many more to satisfy the need for full-stack solutions in different and particular settings. In the sixth issue of Discover the Rising Stars, we talked with The CACSC & BMW Group “Internet of Vehicles” 2021 Global Finals’ New Talent Award winner, SynSense’s CEO Dr. Qiao Ning, who takes us into the mysterious world of neuromorphic technology.
Opportunities in the Chinese market is a fertile ground for a “neuromorphic dream” to grow
Neuroscience is an attractive and mysterious area of natural science; it is connected to the life of every one of us. How do we develop consciousness? Why do some people have higher IQs? Why are some people born with the autistic disorder? These are questions belonging to the field of neuroscience. Neuromorphic research is becoming the strategic focus in scientific development for many countries. Powerful players on the global stage are engaged in an increasingly competitive race.
The founder of SynSense, Qiao Ning, received his doctoral degree from the Institute of Semiconductors, CAS. He later joined the Institute of Neuroinformatics, the University of Zurich, and ETH Zurich in 2012 as an assistant professor, leading the projects on researching and developing advanced Neuromorphic chips. He was the head of just under ten neuromorphic projects for the European Union and published over 30 articles on world-class neuromorphic conferences and journals. In February 2017, Dr. Qiao Ning’s neuromorphic dream finally materialized: he started SynSense, a leading-edge neuromorphic computing company, in Zurich. As the CEO, he led the team to drive the neuromorphic technology industry to the next stage. As his company matured, Dr. Qiao Ning realized the great opportunities the Chinese market provides. He wanted to benefit from the capital, the policy supports, and the various application settings in China. He believed that directing the company to grow further in China would reap plenty of benefits. Therefore, in 2019, SynSense came to China.
Today, SynSense has offices in many regions. Zurich Advanced Research and Development Center, mainly responsible for algorithm and microchip research; Chengdu Hardware Research and Development Center, responsibilities mainly concerned with the development side; Shanghai/Nanjing IC Research and Development Center, with roles in Technology R&D. Core team members have more than 15 years of experience in neural algorithms and over 20 years of experience in microchip development. They have been working on neural technology, covering algorithms, architectures, and design with originality, unprecedented technological pathways, and professional expertise. More than 700 articles are collectively published on top-level international neuromorphic journals among them, and their works are referenced nearly 16,000 times.
Unprecedented technological pathways give birth to top-level chips
Dynamic Neurormorphic Asynchronous Processor CNN (DYNAP-CNN) is a scalable, fully configurable, digital event-driven neuromorphic processor, with 1 million Re LU spiking neurons per chip for implementing Spiking Convolutional Neural Networks (SCNN). This technology is ideal for always-on, ultra-low power, and ultra-low latency event-driven sensory processing applications. With a dedicated interface for dynamic vision sensors, it allows direct input of event cameras. It is fully event-driven with ultra-low power consumption, using uW level background current. With the support of the SCNN algorithm and 5 to 10 ms end-to-end delay, DYNAP-CNN is ten times more practical by using the sparse matrix. It also lowers power consumption 100 to 1,000 times, less than 1mW, while calculating data in real-time with no need for caches. The cost to run the system is cut in half right away. Sensor data will trigger the algorithm, and no need to switch between sleep/work mode. DYNAP-CNN makes always-on a reality.
SPECK is the world’s first intelligent dynamic vision sensor that is entirely event-driven, with ultra-low power consumption and ultra-low latency. It supports a multitude of ultra-low power consumption (<1 mW) and ultra-low latency (<10 ms) applications of intelligent vision, facial recognition, behavioral recognition, and many more. SPECK can be a diverse solution by providing always-on intelligent vision for edge computing terminals, such as smart homewares, intelligent buildings, intelligent security systems, drones/unmanned vehicles, robots. Compared to more traditional offerings, SPECK chips have many advantages. It provides a comprehensive vision solution: DVS (Dynamic voltage scaling) + DYNAP-CNN (scalable, fully configurable, event-driven neuromorphic processor) = SPECK. In comparison to traditional options, SPECK’s power consumption is 100 to 1,000 times lower (<1 mW). When operating vision recognition tasks, its latency is 5 to 10 times lower (<10 ms). It also reduces cost by ten folds. SPECK is equipped with dynamic vision recognition technology and protects user’s privacy. It is projected for SPECK to go into small-scale production by the end of 2021 and expand into large-scale production by 2022.
Pollen, the 100 uW SRNN real-time sensor processor uses dynamic sparse data, reduces the demand for hardware by ten times. Its event-driven algorithm lowers power consumption by 100 to 1,000 times. New architecture and complete parallel computing of Pollen reduces the chips’ cost by 10 times. This sensor processor can convert and respond to signals in real-time with no cache. It is compatible with various temporal information handling sensors. Pollen is projected to go into mass production starting from 2022.
SynSense’s exclusive ultra-low power consumption, ultra-low latency chips and solutions will restore the industry value chain, and provide neuromorphic intelligence for hundreds of billions of edge devices.
Tentatively exploring, we better ourselves at every step of the journey
Dr. Qiao Ning said, “This is a relatively new industry; it only started around 2014 to 2015 in China. The best industry talents are mainly from overseas. The few domestic talents who are here are primarily concerned with the academic part of things anyways and lack experience. It takes a lot of time for them to finish the initial break-in process into the industry. This is cutting-edge technology. There is no product output for capital to commit with initially, so we are constantly exploring our options at every step of the journey. We need to understand the market and its needs. We need to understand the funders and stockholders and their needs. We need to satisfy the needs and values of the society as much as we can. We make the chips a complete solution, make the process simple, make user experience convenient, and then we expand into mass production. At every turn, we always try to better ourselves.”
Impressing the audience at the CACSC and open the possibility for the future
“We have been working with BMW Innovation Lab for a while before we joined the CACSC. We got to know many enterprises operating in different industries but are relevant to us through this competition. We discussed at length about technology. The CACSC was a great learning experience. We also want to benefit from Alibaba Cloud’s platform and discover more opportunities. 5G technology is the hottest thing right now. I believe, in the future, the market will focus on end-devices by integrating cloud services on there. In the next decade, cloud services, along with robots, unmanned vehicles, drones, brain-computer, human enhancement, and computing power for behavior analysis, will see market growth and more room for development. SynSense has completed the series C funding round. The last round was in May 2020, when CTC Capital, Merck, Casstar, Ecovacs Robotics, Yunding, Yachangfu Investment, and Hygoal Capital committed close to a hundred million yuan. Pre-series B funding is in progress right now. I believe entrepreneur teams should remain confident and patient. There is no need for over-packaging if we focus on differentiating ourselves. We utilize social capital, education resources, policy supports, and industry supports to drive our company forward.” Dr. Qiao Ning expressed.
To this day, neuromorphic engineering is still at the early stage of its development. In the future, neuromorphic engineering still needs to take one step further and realize more comprehensive human-brain-level computing power. Dr. Qiao Ning led SynSense to become a pioneer in neuromorphic engineering research and development. His team and he will continue to venture into new territories and open the door of a wealth of knowledge for humanity.