What Is the Poetry and Distance of Smart City

Let's start with a story. 50 years ago, in order to help humans land on the moon, IBM participated in all the computing tasks of the Apollo program, including launch, navigation, landing, landing and safe return to the earth.

"When Apollo 11 arrived about 120 meters above the landing site, the alarm sounded loudly in the lunar module; NASA command center also received the signal that the airborne instruments were faulty and could not continue to guide the lunar landing operation. At this time, NASA and Armstrong faced a difficult decision:" go? Or no go? " At this time, the IBM team in the control center firmly gave the "continue the landing mission" according to the results of countless simulation tests In the end, Armstrong stepped out of the lunar module and left historic footprints on the lunar surface. NASA's flight commander Gena Kranz commented afterwards: "without IBM and the systems provided by IBM, we would not be able to land on the moon."

The "one stroke decision" at the critical moment comes from the credibility of IBM computer and the accurate orbit prediction of IBM computer system based on data theory and physical principles.

Since then, IBM has an indissoluble bond with space exploration. IBM's ThinkPad boarded the space shuttle endeavour in 1993 and participated in the task of repairing the Hubble Space Telescope; In 2018, IBM cooperated with Airbus to develop the first astronaut interactive mobile partner project cimon (crew interactive mobile company) in human history to help astronauts complete many space missions like humans.

From computer to AI, IBM has made unswerving investment, and applied a steady stream of scientific and technological achievements to the industry to solve various problems of cities, people's livelihood and society, achieving the mission of "science and technology for the good".

Vision of smart city

Intelligence makes people look forward to. IBM has never stopped research and practice in the field of intelligence, and always has insight into opportunities. In November 2008, IBM took the lead in putting forward the concept of "smart earth", and prospectively predicted that the world is becoming connected and interconnected, and all things will become intelligent.

As early as 1956, before the Dartmouth conference put forward the concept of "artificial intelligence", IBM developed the first artificial intelligence program that can play international checkers in 1952; In 1997, he defeated chess champion Kasparov's dark blue, opening up a precedent of "man-machine war"; Jeopardy at the puzzle Q & a festival in 2011! Watson, who defeated two human champions; Project debater of artificial intelligence can accurately understand human natural language and debate effectively. IBM is one of the first steps in intelligence and continues to push artificial intelligence to milestones.

Entering a new era of general artificial intelligence (AGI), AI extends to cities, people's livelihood and society more and more widely and deeply. AI can improve people's lives, innovate business models, create personalized services, optimize various urban facilities and services, and so on.

As far as we can see, AI has spread to all aspects of the city, such as e-commerce, unmanned supermarket, automatic driving, intelligent building, intelligent factory, intelligent medical treatment and so on.

In transportation, IBM Watson assistant technology develops a solution to change urban driving mode to alleviate traffic congestion and provide users with personalized and optimal travel suggestions;

In medical treatment, Watson health covers many branches, such as tumor and gene, medical imaging, life science, health management and medical payment, helping doctors effectively realize "standardized medical treatment", "precision medical treatment" and "intelligent medical treatment". At the same time, help "inclusive medical care";

In manufacturing, IBM uses artificial intelligence defect detection technology to help detect the yield of the display panel production line. It can detect whether the panel has scratches or other defects within milliseconds, and the interpretation accuracy is more than 97%.

In food safety, IBM artificial intelligence sensor technology can more conveniently and quickly check the bacteria that may appear on food in real time.

The intelligent scene in the real world is much more complex than "man-machine war" or "Puzzle Challenge". With the wide use of artificial intelligence in all walks of life, "credibility" has increasingly become a serious problem.

Xie Dong, vice president of IBM and chief technology officer of Greater China

"For modern cities, the application upgrading of 'city' (urban construction) and 'industry' (industry and industry) is very important. As the saying goes, why is a city intelligent if the industry is not intelligent? And a credible AI foundation is also very important for the intelligence of 'city' and 'industry'." Xie Dong, vice president of IBM and chief technology officer of Greater China, said.

"Ai credible" torture

Artificial intelligence is a double-edged sword. On the one hand, the combination of AI and digital technology brings benefits to our life and work; On the one hand, it also has an impact on moral standards and values. AI changed its face and "Zao" exploded overnight. While becoming a phenomenal app, it also triggered public concerns and thinking about data privacy and portrait rights.

For another example, the world's first case of a driverless car hitting a dead person is that pedestrians do not take the crosswalk, and driverless cars do not have brakes; In addition, without artificial and technical intervention, artificial intelligence algorithms can even produce prejudice that can not identify ethnic minorities.

Therefore, people today have a deeper torture of "Ai credibility".

Because with the maturity of AI technology and application, how to embed human norms and moral values into AI system? Whether it is consistent and compatible with the actual norms and values; Is the behavior of AI controllable? With the landing of a large number of AI product applications, what standard application products are reliable and safe? With the higher autonomy of AI, such torture is bound to increase. Facing the real scene, is the diagnosis and treatment plan and disease prediction given by AI doctors accurate? Can the investment advice given by AI analysts be adopted?

Different from the accurate calculation of the lunar landing orbit by IBM computer, today's AI is more complex, especially the algorithm represented by deep neurons, which is inherently unexplainable.

"To correctly develop artificial intelligence, we must put forward reliable artificial intelligence. The future of artificial intelligence lies in trust. If we want to drive business success, artificial intelligence can not be hidden in the" black box ", said Xie Dong, vice president of IBM and chief technology officer of Greater China.

How to make AI more credible?

Not long ago, the national new generation artificial intelligence governance professional committee issued the new generation artificial intelligence governance principles - developing responsible artificial intelligence, and put forward eight principles: harmony and friendship, fairness and justice, inclusiveness and sharing, respect for privacy, security and controllability, shared responsibility, open cooperation and agile governance. The governance principles emphasize the need to ensure the safety, controllability and reliability of artificial intelligence, promote sustainable economic, social and ecological development, and jointly build a community with a shared future for mankind. In April 2019, the European Commission also officially issued the code of ethics for trusted artificial intelligence, announcing the launch of the trial operation of the code of ethics for artificial intelligence.

Globally, in terms of top-level design, trusted AI has become the basis for the large-scale development of artificial intelligence.

IBM not only adheres to the values of trusted AI, but also has made great progress in methodology and tools. IBM defines the dimension of trusted AI as scientific and engineering goals. It not only outlines the various methodologies to achieve these goals, but also introduces various tools to determine their integration throughout the AI application life cycle.

IBM divides enterprise AI into three dimensions: advanced AI, scalable AI and trusted AI. In brief, advanced AI makes AI more powerful; Scalable AI is to make AI more widely used in the industry; Trusted AI, that is, to ensure the fairness, interpretability, robustness and transparency of AI, is to make AI trusted by people and truly become an assistant and partner of mankind.

"IBM is vigorously carrying out research and practical work in this area. It is making AI more credible, transparent, safe, available and reliable. For interpretability, fairness, transparency and robustness, IBM also provides tools and technologies to identify and mitigate risks and violations, help understand the impact and challenges of AI models related to trust, and help developers and data scientists light up "Build a credible, safe and interpretable artificial intelligence system", Xie Dong stressed.

First of all, IBM believes that a trusted AI system needs to follow several basic principles: fairness. For example, the AI system should adopt unbiased data sets and models to avoid unfairness to specific groups; Explainability: for example, the decisions or suggestions provided by the AI system are not an unexplainable black box, and should be analyzed and understood by users and developers; Robustness, such as AI system should be safe and reliable, not tampered with, and not affected by "contaminated" training data sets; Transparency, such as AI system can be managed transparently, its development, deployment and maintenance can be tracked, and can be audited within the life cycle.

Secondly, IBM provides tools and technologies to identify and mitigate risks and violations, help understand the impact and challenges of AI models related to trust, and help developers and data scientists build trusted, secure and interpretable AI systems.

AI fairness 360 is a comprehensive open source tool set containing metrics and algorithms to check and mitigate unnecessary biases in AI and help communities build trust in AI;

AI explainability 360 is an open source set of most advanced algorithms released on August 9, 2019. It uses a series of technologies to explain AI model decisions;

In terms of AI robustness / robustness, IBM Research Institute has launched the advantageous robustness 360 toolkit, which is an open-source toolkit for detecting models and fighting attacks by IBM research team, so as to strengthen the defense of misguided AI models for developers and make AI systems more secure.

In addition, IBM is improving the transparency of AI and launching tools to let data scientists and enterprise executives find deficiencies in the model. AI openscale is part of IBM's efforts to improve the transparency of the so-called "black box approach", thus accelerating the large-scale development of AI.

In terms of AI interpretability and traceability, AI openscale technology platform can detect and solve the bias problem when various AI applications are running; No matter what environment AI applications are built or run in, AI openscale enables enterprises to realize transparent management in the whole life cycle of artificial intelligence, record each prediction, each model version, all training data and various indicators used, and help enterprises comply with European Union gdpr and other laws and regulations.

epilogue

The "poetry and distance" of a smart city is inseparable from the digital infrastructure of trusted AI. IBM will shape the values of trusted AI and create a rich trusted AI tool "unity of knowledge and practice" to ensure that cities and people benefit from AI.

Source: cloud wisdom Era