At present, the layout of major technology giants in the AI ​​industry, we can see that hardware, open source algorithms, cloud services have become a must. This article is a AI industry research report from Goldman Sachs. The report details the technical background, application prospects and current industry competition situation and industry chain map of AI.
The future direction of artificial intelligence 100 billion market : the song of software and hardware
Artificial intelligence (AI) can be said to be a big leap in the era of scientific and technological information. It has a certain human-like logic, but also has powerful computing power and data storage capabilities. Although the industry is not in a good state of investment, AI technology is still considered to be the next big breakthrough to bring huge economic benefits and improve social productivity.
In fact, in the past two years, concepts such as AI, robotics, and autonomous driving have become the most advanced cultural and political terms. Many studies believe that we are currently at such a turning point: computing power is faster and faster, data sources are more abundant, deep learning algorithms are maturing, professional hardware (chip) and open source code are gradually emerging, more and more practical Sexual AI came into being. This article is a AI industry research report from Goldman Sachs. The report details the technical background, application prospects and current industry competition situation and industry chain map of AI.
The three main pushers behind AI: data, chips, algorithms
1. Mobile networks are popularizing data structuring or outbreaks.
Data can be said to be the key to machine learning. The ubiquitous connectivity of devices around the world, including mobile devices and the Internet of Things, has led to a proliferation of unstructured data, which means that machine learning algorithms can be used to simulate, train, and test data sources more abundantly.
For example, the Tesla Connected Cars have been used. As of now, Tesla has a total journey of 780 million miles, and the company's platform-driven extra-connected cars have also traveled 1 million miles. In terms of wireless operations, Verizon announced new transmission standards in August, making remote sensors connect to cloud software faster. At the same time, the new 5G network will also facilitate the transmission of data. IDC expects that the average annual data volume will reach 44 Zebytes (that is, 44 trillion bytes) by 2020, and the compound annual growth rate will reach 141% in the next five years. Big data technology will gradually penetrate the practical field.
At the same time, the cost of building large-scale databases and cloud processing technologies on mobile networks is also decreasing. It is expected that in less than three years, 90% of people will have unlimited free data storage supported by advertisers. This is because the cost of hard drives continues to decline, stimulating the creation of data. In fact, nearly 90% of the data was created in the past two years.
2, GPU application trend new hardware is more suitable for parallel structure
GPUs are considered to be low-cost, high-computing processing units, especially for cloud services and new neural network architectures, which increase accuracy and computational speed. The GPU-based parallel architecture allows for a faster machine learning training system that is far superior to the currently widely used CPU-based data architecture. In addition, with an additional graphics card network, the GPU system can speed up iterations for more accurate and fast training.
The chip's floating-point capability can be developed with the NVIDIA GPU (GTX 1080) as an example: the chip's performance is 9T floating-point arithmetic, worth about $700, or about 8 cents per G. Referring to the IBM 1620 in 1961, regardless of floating-point computing power, the cost per G floating point operation through series is $9 trillion.
We also emphasized the market prospects of GPUs in the 102-phase intelligent internal reference and analyzed market share. Compared to the CPU, the GPU has thousands of computing cores, and powerful and efficient parallel computing capabilities, which can achieve 10-100 times application throughput, especially for AI massive training data. Today's deep learning solutions rely almost entirely on (NVIDIA) GPUs.
3, the algorithm continues to optimize large companies to promote open source
More and more algorithm research is driving the practicality of deep learning. Berkeley, Google, and Facebook have also revealed their own source code frameworks, namely Caffe, TensorFlow and Torch. The open source code has attracted more and more software developers to try new algorithms. In less than a year, TensorFlow has formed an active repository GitHub, which is currently the largest developer cooperation website. Of course, not all AI comes from the open source framework.
Three major types of AI industry layout
From the perspective of the technology update cycle, in the past 50 years, computers have been continuously promoted by Moore's Law. In terms of system framework, computing power, storage capacity bandwidth, and programming language conversion have made great progress. See the economic boom brought about by technological changes in the 1990s, and promote the rectification of software, hardware, and network companies. The market value of public software companies in 1995 has soared from $200 million to $500 million, and it has only tended to be flat around 2000. Clearly, AI also has this trend, leading the growth of hardware, software, data and service providers. In fact, Google, Amazon, Microsoft and Salesforce have completed 17 AI-related acquisitions since 2014.
10K PCS Ready HOT OEM Ozone Generator in Car Air Purifier Mite Anti Virus Filter Sterilizer Disinfecting
Features:
O-zone & Anoin Two Mode --- This advanced air cleaner purifier with Anion and O-zone 2-in1 air sterili-zation mode to fresh air and kill harmful mold decompose smoke, odor and formaldehyde, Keep the air fresh!
Anoin Mode Powerful and Enviroment Friendly --- Releas 8million/cm3 Negtive Ion captures up the airborne particles(PM 2.5), pet odor, smoke, kill harmful mold and other harmful gases. No chemical composition and filter requied. Effectively improving the air quality and Prevent harmful substances inhaled into the lungs.
Car Air Purifier,Car Air Cleaner,Car Air Purifier Ionizer,Best Air Purifier For Car
Jiangmen soundrace electronics and technology co.,ltd. , https://www.soundracegroup.com