AI startup company Oviewo.ai released the first asynchronous convolutional neural network chip. The asynchronous architecture adopted by the chip follows completely different design principles and technical routes compared with the previous various AI chips in this field, and combines fixed-point, automatic pruning, multi-layer optimized memory access and the like. Technology can greatly reduce the power consumption required for data analysis in loT scenarios. Its function focuses on visual analysis such as object recognition, and it is suitable for wearable devices, home, self-powered monitoring and other IoT terminals that have rigid requirements for ultra-low power consumption. The first such chip has been designed and implemented by the company, which is the first time for the asynchronous architecture AI chip for the IoT field.
泓观科技 is a technology-based company founded by three Tsinghua alumni in 2015. It has been committed to deep learning through edge computing, enabling high efficiency to be embedded in devices and applications, and working with upstream and downstream partners to build an artificial intelligence industry ecosystem. . They have always been low-key stealth, and they are also using their own way to open up another product line, which is now breaking out. The Internet of Things (IoT) is the first to launch an ultra-low-power asynchronous AI chip, which has taken a different approach. The accumulation of thin hair.
The founder of Zhenguan Technology introduced: "Since we entered the Tsinghua University microelectronics major nearly 20 years ago, although the career behind us has brought more diverse knowledge background, it is considered an integrated circuit. The practitioners of this line for many years have their own understanding and understanding of the law of the huge industry of chips. Since we decided to do something else in the AI ​​chip and build a second product line, we hope to choose a unique and interesting one. The focus of the application, corresponding to a certain differentiated application areas and ecological location, and at the same time with our platform-based system chip's first product line, compatible and complementary, complement each other."
The asynchronous architecture used by the company is significantly different from the various AI chips previously introduced in this field. A common feature of previous AI chips is the design principle that follows the functional and performance priorities. The application scenarios that it targets are mainly divided into two categories: cloud-based online computing and offline computing based on front-end devices. The former such as Google TPU series, Bit Continental Sophon series, etc.; the latter such as Movidius Myriad series, HiSilicon Kirin970 containing NPU, Apple including NeuralEngine A11, etc., even as a front-end AI chip, its application environment is concentrated in mobile phones, drones Wait for high-end devices.
However, for the IoT device and its application scenarios, the architecture and features of the above AI chip are not suitable. “One reason is that loT devices usually focus on a single function. For example, devices for video capture generally do not need to run algorithms related to speech recognition and natural language processing; second, compared to high-end devices such as mobile phones and drones. The rigid demand for most loT devices lies in ultra-low power consumption, and performance is not the first consideration.†The founder of 泓观科技 said, “Like ours, we follow a completely different design principle and technical route to implement convolutional neural networks. The AI ​​chip can be said to be a bit of a singularity and even ahead of the curve. As far as the media reports are concerned, we should be the first to realize the asynchronous convolutional neural network circuit solution and realize the streaming of the startup, whether in China or worldwide."
This unique and new asynchronous convolutional neural network chip focuses on intelligent visual analysis such as object recognition, and has rigid requirements for ultra-low power consumption for wearable devices, smart homes, and self-powered (such as solar) monitoring. In the field of IoT terminals, AI applications are gradually being implemented. "The choice of this route and the point of origin comes from our thinking and judgment on technology development and market positioning." The founder of Zhenguan Technology concluded that "there will be a look back from the near future, this chip represents The traits should no longer be unique and alternative. For the broad convergence and deep integration of the two areas of AI and IoT, I believe this time means a beginning."
Fiber Optic Splice Closure,Fiber Optic Splice Case,Fiber Splice Closures,Outdoor Fiber Optic Splice Closure
Cixi Dani Plastic Products Co.,Ltd , https://www.dani-fiber-optic.com