On April 28, in Beijing, the highest temperature was 30 degrees.
In this hot early summer, on the second day of GMIC Beijing 2017, a continuous stream of technical sharing constantly stimulates the nerves of tech enthusiasts, accelerating the heartbeat of the audience and keeping the atmosphere heated. Today, as a company with technology as its core competitive advantage, Thunder Software Technology has brought the hottest cutting-edge technology sharing to all tech fans – Embedded Artificial Intelligence.
1
Artificial Intelligence Must Be Localized
Since the concept of artificial intelligence was first proposed in 1956, it has gone through more than half a century. After more than 60 years of development, artificial intelligence is on the eve of a global explosion. The development of big data, quantum computing, and super chips has been accelerating this process; technologies such as facial recognition and voice control have already “invaded” the minutiae of daily life, decisively influencing the future direction of society. The smart era is not far away.
However, if there is only cloud-based artificial intelligence, we do not consider that a comprehensive smart era. In certain specific scenarios, such as autonomous driving, smart security, and certain fields of smart homes, artificial intelligence needs to be more real-time, secure, and lightweight. The series of actions of transmitting data to the cloud, computing through servers, and returning data locally not only consumes time but also places high demands on the network. Therefore, artificial intelligence must be localized.
2
New Play of Embedded Artificial Intelligence
So, can localization of artificial intelligence be achieved?
At the GMIC New Technology Delay Show forum, Wang Fan, Technical Director of Thunder Software Technology, gave a positive answer. Moreover, we already have a set of different play methods from cloud-based artificial intelligence.
Considering that mobile chips and PC chips still have a gap in computing power, and that embedded devices have relatively small space and require lower power consumption, we must have a more economical and efficient algorithm to maximize the performance of embedded artificial intelligence. To achieve this, we can approach it from two aspects: on one hand, we need to fully utilize the performance of high-performance mobile chips; on the other hand, we must continuously optimize the algorithms of embedded systems to minimize the computational load while achieving the same effect. Only in this way can we obtain a relatively perfect embedded artificial intelligence system.
Specifically, hardware can be divided from top to bottom into: application layer, algorithm layer, operating system layer, driver layer, and chip layer. Due to the limitations of mobile chip computing power, embedded artificial intelligence not only needs to optimize the algorithm layer but also needs to focus on optimizing the underlying drivers and operating system – this requires us to be experienced in the field of operating systems and have close cooperation with chip manufacturers.
In this aspect, as a global leading provider of smart operating system and platform technology, Thunder Software Technology has been deeply engaged in operating systems for many years, accumulating rich experience; and as a deep technical partner of ARM and Qualcomm, at the 2017 CES and MWC, Thunder Software Technology demonstrated embedded artificial intelligence demos with both chip giants, achieving impressive results.
Algorithm optimization needs to consider the differences in device performance in different application scenarios to balance performance and power consumption. At this time, appropriate computing network design, proper pruning, and sharing of basic parameters can help achieve optimal algorithms.
In summary, to excel in embedded artificial intelligence, we must achieve a full integration of software and hardware.
In our view of the intelligent world, embedded artificial intelligence and cloud-based artificial intelligence complement each other. Thunder Software Technology is committed to integrating upstream and downstream resources in the industry chain to promote the development of the embedded artificial intelligence industry, ultimately accelerating the arrival of the intelligent world.
We are working hard to create a rich and colorful intelligent world,
your attention and support are our greatest encouragement.
