domeet webmaster

Imaginative reconstruction of “future AI computing” innovation.

Editor’s note: Author @ IBM Vice President, Chief Technology Officer, Greater China Dr. Xie Dong, Reprinted with permission.

Header & Photo Source | EmTech China

It seems that technology and capital are the two main driving forces of the current new business era, as well as the two major directions of our content efforts. “How future technology will develop” has always been the focus of attention. Therefore, ‘s “Science Technology · Imagine the Future” series of activities hope to provide readers and technology-loving people with a variety of views on the future direction of technology development, imagine the future with readers, and look forward to the technology The next new decade.

EmTech China is a global emerging technology summit hosted by the MIT Technology Review, which first landed in China in 2017. On December 13, 2019, EmTech China was held again in Beijing. Top Chinese and foreign experts and scholars started discussions on topics such as future reconstruction and AI + computing power. Interestingly, not long ago, the Department of Electrical Engineering and Computer Science (EECS) at MIT just split and formed three colleges such as Electronic Engineering (EE), Computer Science (CS), and Artificial Intelligence and Decision (AI + D). For the first time in the academic world, it has gained the same status as CS and EE [1].

Picture: Dr. Xie Dong at EmTech China

At this time, 57 years have passed since Purdue University established the world’s first computer science department [2]. You know, the world’s first artificial intelligence conference was held in Dartmouth College in the United States in 1956. It took 63 years to finally have its own independent faculty. It has been a research direction in computing science before. The history of IBM tells us that the mass production and application of computing machines and related sciencesThere has always been a positive interaction between industry and basic disciplines. MIT established an independent department of artificial intelligence and decision-making, which means that AI has finally become a basic discipline in a large-scale business or industrial field.

Picture: Dr. Xie Dong at EmTech China

It is worth mentioning that IBM and MIT established an AI laboratory in February 2017. IBM will invest $ 240 million over the next ten years. This is the largest academic cooperation investment in the development of IBM. One aspect validates the importance of AI to academia and industry. Undoubtedly, AI computing will become the next IT computing. The AI ​​industry will be comparable to the IT industry. The business changes brought by AI will be no less than the business changes that IT has brought. What exactly is the future of AI computing? This is the subject of future calculations.

Imagine AI calculations

2019 is a crucial year for computer science and artificial intelligence science. In this year, the researcher of IBM “Watson” Research Center , known as the father of memory Dr. Robert H. Dennard was awarded the Robert N. Noyce Award, the highest honor for the semiconductor industry in 2019 by the Semiconductor Industry Association (SIA). Dennard Scaling, conceived by Dennard, has laid the mathematical and physical foundations of Moore’s Law, and on this basis has developed the electronic computer industry that is booming today.

However, just as the concepts of computer science and artificial intelligence science actually originated and developed at the same time, the hardware and software adapted to AI algorithms were far from mature in the first 60 years of human information civilization. The supporting ERP, CRM, BI and other commercial software have established a huge business world, and have accordingly promoted computer science based on digital integrated circuits. The reason can be simply understood, that is, the gap between traditional computer science and the AI ​​computing we are heading for: one is for precise calculations, and the other is logical reasoning (that is, decision making) for inexact calculations.

WithIBM Project Debater uses an image metaphor. This intelligent machine that can debate with humans has attracted widespread attention from the global industry-university-research community in its public appearance in 2018. Project Debater, which lasted more than six years of research and development, is a follow-up to IBM ’s “Deep Blue” which defeated the human chess master in 1997 and “IBM Watson” who defeated the human champion in the puzzle game show “Dangerous Edge” in 2011. A milestone. However, the difference is that compared to the previous two tasks (play chess and open Q & A), Project Debater is faced with tasks that do not have black and white answers. In fact, even human and human debates are often difficult to separate. Win or lose.

IBM has actually reimagined hardware, software, and applications around future AI computing.

For example, the digital circuits commonly used in electronic computing are actually not suitable for AI computing. The basis of digital circuits is bits, that is, the world composed of “0” and “1”. In a bit world other than “0” or “1”, the calculation tasks completed can only be accurate and precise calculations. However, AI computing is just the opposite. In the future, AI machines may argue with humans like Project Debater, and provide a range of possible decisions under the premise of ethics. IBM was first trying to solve fuzzy problems like AI with analog circuits that were considered obsolete, because the output of analog circuits was a continuously changing digital range, which opened a variety of “0” and “1” may. The AI ​​hardware center of IBM Research Institute successfully implemented the DNN neural network with analog circuits, which significantly reduced the cost and power consumption while achieving the same level of GPU [4]. IBM is also investing in atomic storage and quantum computing, which may be hardware that is truly suitable for AI computing in the future.

In short, future AI calculations are not equal to or will replace today’s IT calculations, because future AI calculations will solve completely different problems and perform completely different tasks. The establishment of the Department of Artificial Intelligence and Decision Making independently of the Department of Computer Science at MIT is the watershed of this historical process: human civilization will finally usher in the era of AI computing.

The road to future computing