PC chip giant Intel wants to make a "big move" on artificial intelligence. At Intel AIDay in Beijing this week, Intel announced its overall strategy and roadmap for artificial intelligence for the first time in China.
After missing the smartphone market, Intel is eager to regain a game on artificial intelligence. In the past, Intel has made a crazy acquisition, and on this Intel AIDay, we finally saw the final ownership of the products and technologies of these acquired companies. But the question is how many opportunities does Intel leave for this market?
Artificial intelligence new architecture Nervana
For the planning of artificial intelligence, Intel's ideas are not the same as other technology giants. The whole set of strategies is “all-inclusiveâ€, which includes the hardware and software necessary for artificial intelligence, as well as industry alliances and efforts in the education sector that are intended to guide artificial intelligence to “goodnessâ€.
The hardware component of the underlying Intel strategy is centered on the Nervana platform. In the middle of this year, Intel acquired artificial intelligence chip startup Nervana for $400 million to make up for its weakness in AI chips. At Intel's AIDay, Jason Waxman, general manager of Intel's data center solutions division, announced that he will use Nervana as the Intel artificial intelligence brand name and announce the open source of the Nervana platform.
Intel's Nervana product portfolio is strong, including Intel's universal solution and the latest version of the processor Xeon Xeon, Skylake, which will increase performance by 18 times. The latest version of the Xeon Phi processor, KnightsMill, has four times the deep learning performance of the previous generation and can reduce training time by a factor of 31. The product also includes products from the acquisition of FPGA technology from Altera.
But the most notable attention may be the next generation independent processor LakeCrest on the Nervana platform. Intel said that the product will be available in the first half of 2017, and that by 2020, LakeCrest's deep learning performance will be 100 times that of the current GPU chips.
In addition to hardware, Intel has added hardware-optimized software products to the entire suite of artificial intelligence products, such as the Math Function Library (MKL) and several deep learning frameworks available in all Intel hardware architectures, from Nervana. Neon, TensorFlow with Google, and Theano, Caffe, and Torch also released the Deep Learning SDK. These SDKs are for scientists in the data center and can also be used by application developers at the end.
Intel AI product portfolio
Only products are not enough for Intel. Intel has established an artificial intelligence advisory committee with members in the AI ​​field, including Professor Bruno Olshausen from the University of California at Berkeley and Professor Jan Rabaey, who studies neural network architecture, and Professor Ron Dror from Stanford University who is working on supercomputers.
Intel also established the Nervana Institute of Artificial Intelligence, which has an Intel Developer Zone on Artificial Intelligence and an Intel Campus Developer Program for College Students, dedicated to providing the next generation with resources for researching artificial intelligence.
At the same time, Intel has formed an artificial intelligence alliance with industry-leading companies, including Google, to help enterprise IT departments build open, flexible, and secure multiple cloud infrastructures based on business needs.
Intel's opportunity
Artificial intelligence is redefining technology. With the further development of the Internet and the Internet of Things, data is showing a spurt of growth, and the ability of artificial intelligence and deep learning to process data quickly is becoming a demand of the whole society. Intel, which has always been good at processing data and using data centers as a major source of revenue, is unlikely to sit back and watch this trend.
But compared to another chip giant, Nvidia, Intel's layout in the field of artificial intelligence is not too early. The latest column from Forbes compares Nvidia to "new Intel" to show NVIDIA's success in the era of artificial intelligence. This is not an exaggeration. 3,000 artificial intelligence companies around the world are developing their own business through NVIDIA's platform. NVIDIA's GPU chips account for 70% of the total market share, and NVIDIA, which has already tasted the sweetness, will continue to develop in the future. Artificial intelligence.
Other technology giants have not slackened. At this year's Developer Conference, Google announced a processor TPU customized for the Google artificial intelligence open source platform TensorFlow. Microsoft, proud of its data-capability search, has also developed chips for its own data center. The earliest IBM in the field of artificial intelligence also launched a human brain chip called TrueNorth this year.
But all this does not mean that Intel does not have its own opportunities. The original Nervana founder and CEO AveenRao said in an interview that Naveren's processor has strong compatibility with other companies' customized versions and can accelerate multiple deep neural networks. "We have the ability to condense neural networks to a small number. The basic unit, and only a small part of the basic units of these neural networks are important."
Currently, GPU chips are the most effective way to train artificial intelligence systems, but technology giants are also looking for other types of hardware to enhance deep learning performance. But some insiders also said that the GPU chip that was originally designed for faster processing of graphics has a certain distance from the ideal deep learning chip. The head of the Silicon Valley startup incubator YCombinator and the OpenAI co-founder SamAltman, who is known for Musk, have also evaluated the relationship between GPU chips and artificial intelligence as "coincidence." At the same time, only chips are not enough for companies that research artificial intelligence, and some adapted software and ports are also required.
This may be the opportunity left for Intel and Nervana. It can be confirmed that Nervana is expected to be expected by the Intel's accumulated technical strength in the chip field, and the future market and channel resources accumulated by Intel in the chip market will not allow the Nervana platform to be underestimated.
But artificial intelligence is still just the beginning for Intel. It takes a long way to go to complete such a complex ecosystem. Intel missed the smartphone market, but there should be opportunities in artificial intelligence.
External Cavity Tunable Laser,Laser Grating Sensor,Bragg Grating Stabilized Laser,Grating Stabilized Diode Laser
AcePhotonics Co.,Ltd. , https://www.cgphotonics.com