From the current point of view, our future is still very clear: AI can take care of us, bring entertainment, but also help us make money ~ the existing industry will be more and more efficient, such as health care and manufacturing; many new industries are Will be possible, such as AR glasses and unmanned taxi and so on.
But in the technology industry is busy building this new era of artificial intelligence, and continue to improve revenue, but suffered a large bottleneck on the speed: the computer in some specific computing needs is far less powerful and efficient. When we focus more on the use of algorithms in poker, chess and other areas to defeat the human, we have to support the future of AI to support the new computer chip is little concern.
In fact, in the face of the huge demand for more powerful chips, software and Google have been familiar with the development of their own chips in the complex tasks. They are also facing competition from other emerging start-up companies, these companies are independent research and development with AI-centric chip. Probably, Apple is still doing it quietly. Just as our future lives are being changed by artificial intelligence, these competitions will also shake the existing chip industry.
On Sunday, Microsoft has just announced its AI chip manufacturing program. At the upcoming CVPR conference in Hawaii, Dr. Shen Xiangyang, Microsoft’s global executive vice president for the Department of Artificial Intelligence and Research, showed track of the new chips and hand movements developed specifically to enhance the reality of the lens, HoloLens. The chip includes a set of custom modules that can efficiently run deep learning software. Microsoft wants us to be able to interact with the virtual items we see. They also said that there is no device on the market that can support enough software to run the machine on top of the battery-powered device.
Before Microsoft opened its plans, Google announced in 2016 its depth of learning chip research and development. Its TPU is to support the company within the cloud more efficient depth of learning operations. Google said earlier this year that with the outbreak of speech recognition technology, TPU has saved the company to build 15 new data center costs. This year, Google announced that it has developed a stronger version of the TPU, and will be the company’s cloud computing business customers rent the use of the chip.
From Microsoft’s development of a deep learning processor for HoloLens, you can see that they do not need to start developing their own server chips from scratch to compete with Google’s TPU. Microsoft has spent several years allowing field-programmable gate arrays (FPGAs) to make its own cloud computing more efficient in deep learning. This chip is designed to increase the speed of a particular software or algorithm, and can then be reconfigured. Microsoft plans to provide the chip to cloud customers next year. However, when asked recently whether Microsoft will also create a similar to Google’s consumer server chip, Microsoft FPGA technology leader Doug Burger said that does not rule out this possibility. Partial design and supply chain processes used to develop HoloLens depth learning chips can be redefined to suit the server chip.
Google and Microsoft’s two projects are clearly in order to counter the existing semiconductor manufacturers giant Intel, NVIDIA and other layout. In recent years, Apple has been developing a processor for its mobile devices. It is widely believed that Apple is also developing new chips, so that the future of the iPhone in the artificial intelligence performance more color. Many of the start-up companies are also developing their own depth of learning chips, such as the former Google TPU project engineers founded Groq. “Intel and NVIDIA are struggling to sell products that were in the past,” said Linley Gwennap, founder of Linley Group, a semiconductor industry analyst. “But we’ve seen that running faster will be the leading Cloud computing companies and start-ups because they are the ones who best understand their own data centers and the broader market needs. ”
“In recent years, graphics chip maker NVIDIA sales and profits continue to soar, it is from the chip than the traditional processor is more suitable for in-depth learning software, but they have recently chosen the existing chip design transformation and expansion, and Not developing products specifically designed for depth, “Gwennap said.
Of course, the traditional chip manufacturers will not sit still. As the world’s largest chip maker, Intel acquired a company called Nervana’s AI chip startup last summer, and is currently developing a technology specifically for depth learning based on company technology. Intel has the world’s most complex and expensive chip production business. However, a new group of rising people said they were not afraid. One reason is that they do not have to stick to the existing chip ecosystem, subject to the original development of other software for the purpose of the restrictions.
“Our mission is simpler, because we just have to do one thing, and we can start research and development from scratch.” Nigel Toon, CEO of Graphcore, a startup in the UK that specializes in AI chip development. Last week they had just announced a $ 30 million grant, including DeepMind CEO Demise Hassabis, which also had several leaders from OpenAI. You can learn more information about electronic parts from omoelec.com.
On the other hand, large cloud technology companies can apply their experience to running and developing machine learning services and technology. “One of the things that we really can benefit from Google is that we can work directly with our counterparts, such as voice recognition and street view,” says Norm Jouppi, chief engineer of the Google TPU project. “When you focus only on Customers and working with them, the development turnaround time will be greatly reduced. ”
Both Google and Microsoft have grown with innovative software, but the software is designed and developed for others. With the AI more and more important, the basis of the technology industry, “silicon chip” is changing, followed by manufacturers who have begun to bear it!
Reference: 74hc595 datesheet and ds18b20 datesheet
Electronic Components Distributor 