Thoughts on visit to an edge AI chip design startup - Kneron

By Jeff Lee / Mike Song 2022-10-31

Kneron is a company that focuses on AI chip design, which is rare in Taiwan. The solid technological backgrounds and work experiences in world-class major IT companies of its core members bring up extraordinary mindset and points of view; the appraisal of their chip design can be on par with global major companies like Nvidia, which means Kneron had stepped up to the world stage of automotive technologies. Next, we should see the fruitful results of their cooperation with major carmakers from Germany and Japan, which include L2, smart cockpits, or the application of L4/L5 fully-autonomous driving technologies in the future. The R&D capability of Kneron has to pass strict verification within the next 2 years.

Edge computing has become the hot topic of research of automotive new technology in recent years; In comparison with cloud computing, edge computing is more suitable for various application scenarios for mobility, so the progress of the development of edge AI technology became one of the key battlefields for ICT industry to enter the automotive technology industry. The editors of Auto Future came to Kneron, which is located in Neihu, Taipei City, in the afternoon on 24th October to interview with Albert Liu, the founder and CEO of Kneron, to discuss thoroughly about the achievements of technological development since the company setup 7 years ago.

Albert Liu, the founder and CEO of Kneron (Left) had an interview with Jeff Lee (Middle) and Seven Tseng (Right), editors of Auto Future.

 

Albert established Kneron in San Diego, USA in 2015. Since then, it passed the A round and B round of financing in the past few years, and the investors include IT giants from China, USA, Hong Kong, Taiwan, etc., and their branch offices are located in Taiwan, China, and USA. The self-developed edge AI chips that they are most proud of had been listed as one of the top ten AI chips worldwide by world-famous semiconductors press EE Times in 2020, and their edge AI chips also won the IEEE Darlington Paper Award in USA last year. As being one of the ten outstanding young persons in Taiwan, Albert described the exclusive design of their chips…

 

  • Reconfigurable Streaming. Kneron utilized the Reconfigurable Streaming technology on the CNN, which is the mainstream AI model nowadays. It can be reconfigured by instruction codes from various demand of tasks just like the Lego bricks. For instance, when the task of voice AI or image AI recognition need to be executed, their chips can be reconfigured by commands. Consequently, it maintained the simplicity and flexibility of the chip structure that it can enjoy the benefits of low cost, low power consumption, and high compatibility.
  • The AIoT chips for edge computing. Edge computing has features including low latency, high privacy, and low cost of operation (doesn’t need the internet to transmit data to the cloud), so that Kneron doesn’t need to develop super computing power chips with enormous R&D resources like Nvidia and Intel, but they only need to utilize some light edge computing AIoT chips to complete most of the tasks; if an application scenario that needs higher computing powers had been encountered, then it can cope with that by bricks stacking. More importantly, the feature that edge AI chips can conduct off-line computing can ensure the driving safety of places without internet signals (ex. tunnel).
  • When implementing traditional CNN in the autonomous driving technology nowadays, it recognizes driving environments by partial characteristics of objects, which easily leads to wrong system decision; The enhanced version CNN of Kneron – Transformer, utilizes “holistic recognition”, which interprets the real-time situation of the driving environment with the full-view and time continuity that will be easier for cars to execute correct and safe responses.

 In addition, the application range of Kneron’s products includes Foxconn and MIH’s automotive system, LiteOn’s intelligent telephone pole, Delta’s intelligent factory, JVC’s intelligent rear-view cameras (it can switch interior or exterior view by one-touch), or even CS Bus Group, Gogoro electric scooters, etc., Generally, Kneron had achievements in business of car industry, safety control, edge server, and AIoT, etc., Currently, Kneron’s cooperation with carmakers includes OEM and aftermarket; In the cooperation with OEMs, Kneron is currently developing products with major carmakers in Germany and Japan.

The investors of Kneron includes Foxconn, Winbond, Macronix, LiteOn, Delta Electronics, and also world-class IT giants including AliBaba, Qualcomm. Obviously, the reason why these giant companies coincidentally have confidences on Kneron, is their exclusive concept of AI chip design that can punch about their weight to gain a foothold among the competitive IC design industry around the world. Kneron claimed to march toward the goal of intelligent driving, smart home, and smart city. In my view, just like the thoughts shared by Albert to startup friends in Taiwan… don’t be “Me Too”, but to have beautiful dreams, so that one can have enough courage and motivation to do different things that can be well received by “good judges of talent” around the world.