Privacy Protection in the Era of Artificial Intelligence-The interpretation of PlatON 2.0 White Paper

Author：M6

The following content is my personal opinion, it does not represent any investment advice.
If you appreciate it, please like, comment, and share!

In this era, we do not feel strange about artificial intelligence. Artificial intelligence has contributed to many breakthroughs in modern technology. As the core force of the new round of technological revolution and industrial transformation, it has promoted the overall improvement of social productivity, promoted the upgrading of traditional industries, and driven the rapid development of “unmanned economy”.

Since the birth of artificial intelligence, the theory and technology have become increasingly mature, and the field of application has continued to expand. It is conceivable that the technological products brought by artificial intelligence in the future will be the “containers” of human wisdom. Artificial intelligence is the simulation of human consciousness and thinking process. Artificial intelligence is not human intelligence, but it can think like humans and may exceed human intelligence.

During the last 5 to 10 years, the rapid growth of the Internet, mobile Internet and Internet of Things has generated enormous amounts of data. The increase in chip processing power, the popularity of cloud services and the decline in hardware prices have led to a significant increase in computing power. The broad industry and solution market has enabled the rapid development of AI technology. AI has been everywhere in human’s daily life, and AI has been applied in many industry verticals such as medical, health, finance, education, and security.

A growing number of governments and corporate organizations worldwide are gradually recognizing the economic and strategic importance of AI and are dabbling in AI from national strategies and business activities. A study by PwC on the economic impact of AI[5] on the world economy by 2030 reports that the emergence of AI will bring an additional 14% boost to global GDP by 2030, equivalent to a growth of $15.7 trillion, more than the current GDP of China and India combined. The global AI market will experience phenomenal growth in the coming years. In its 2019 Global AI Development White Paper [6], Deloitte projects that we forecast the world AI market to exceed $6 trillion in the next 2025, growing at a CAGR of 30% from 2017–2025.

According to Mind Commerce’s AGI report, the global market for general AI for enterprise applications and solutions will reach $3.83 billion by 2025, and the global market for AGI-enabled big data and predictive analytics will reach ​$1.18 billion. By 2027, 70% of enterprise and industrial organizations will deploy AI-embedded intelligent machines, more than 8% of global economic activity will be done autonomously by some kind of AI solution, compared to less than 1% today, and more than 35% of enterprise value will be directly or indirectly attributable to AGI solutions.

In the past few years, the challenge between Google AlphaGO and humans became a hot topic at the time, and it was also a landmark event of artificial intelligence. A few years later, Boston Dynamics released their new robot Spot on YouTube, and 42 million people watched the video.

lots of artificial intelligence enthusiasts regard it as an outstanding representative of the artificial intelligence industry. Spot uses artificial intelligence to visualize its surroundings. This means that the user can point the robot in any direction, and the robot can optimize each action to determine the best action to move toward the destination. Sensors detect obstacles to inform whether there are rocks to climb or walls to avoid. The movement is so precise that its four legs can bounce. Even more surprising is that if the robot accepts a challenge and causes it to fall, it can recover itself through a series of known actions.

In September 2019, Boston Dynamics officially announced that the Spot robot was put on the market for lease. In February 2020, Boston Dynamics’ four-legged robotic dog Spot officially joined the Norwegian oil company Aker, becoming the first of the oil company. With a machine with an employee number, he helped check malfunctions, detect hydrocarbon leaks, collect data and generate reports.

Not only that, Spot has also become a new miner. It enters the Kidd Creek mine which is the deepest mine in the world, and goes deep into the mining area for security checks to protect workers and keep them away from potential dangers. No matter what kind of mining terrain, it can be easily controlled.

In the field of construction, Spot is equipped with a laser scanner and can scan the route to collect high-precision 3D data about the construction progress. Spot can also remotely measure key areas of the site and compare the completion conditions with the design intent to find problems as early as possible and minimize rework.

At present, Spot has matured in terms of hardware. The most difficult thing is the algorithm. From the perspective of robot design, behind each action is an algorithm, such as walking, jumping, etc. It is necessary to combine various algorithms with software to coordinate the four legs, and the reason why Boston Dynamics’ robots can perform various shows, artificial intelligence plays a key role in it.

When artificial intelligence algorithms are further, more extensively, and more deeply integrated into the robot system, what will happen next is not just relying on a newly developed set of algorithms to teach the robot to do new actions, but to let the robot make their own set of actions to respond to changes in the situation. This will be a major change in the history of robot development.

Although Spot still has many problems to be perfected, this project is currently the most successful application of robotics technology combined with artificial intelligence. It shows us the infinite possibilities of artificial intelligence.

Machine learning technology, mainly deep learning, cannot be learned and inferred without enormous amounts of data, so enormous amounts of data becomes one of the most important resources for the development of frontier technology of artificial intelligence. Technology giants, especially those in China and the United States, have accumulated huge amounts of data through the Internet services, and as the value of data becomes increasingly prominent in the era of AI, these data will gradually evolve into an important asset and competitiveness of enterprises. According to IDC estimates, the global data volume is expected to reach 44 ZB in 2020, and China’s data volume will account for 18% of the global data volume, reaching 8060 EB (equal to 7.9 ZB) in 2020.

More than 60 years have passed since the concept of artificial intelligence was established in 1956. Why has artificial intelligence not become popular in the past few decades, and it has only become really popular in recent years.

There are two important reasons: data and algorithms.

Before the third wave of artificial intelligence, traditional artificial intelligence did not show obvious “intelligence” characteristics due to the limitation of computing power and data volume, nor did it meet the needs of practical applications. During the period, it experienced the wave of artificial intelligence in the 1950s-1960s, and also experienced the quiet period of the 1970s-1980s.

Until 2006, Professor Hinton proposed that the “Deep Learning” neural network made breakthrough progress in artificial intelligence performance. This artificial intelligence wave was significantly different from the previous two waves. Machine learning algorithms based on big data and powerful computing capabilities have made breakthrough progress in a series of fields such as computer vision, speech recognition, and natural language processing. The application of artificial intelligence technology has also begun to mature. Let artificial intelligence begin to truly move towards “intelligence” and start using it.

The more “intelligent” artificial intelligence is, the more personal information data needs to be acquired, stored and analyzed, which will inevitably involve the important ethical issue of personal privacy protection. Today, all kinds of data and information are collected all the time and everywhere, almost everyone is placed in the digital space, personal privacy is very easy to be stored, copied and spread in the form of data, such as personal identity information data, network behavior trajectory data, as well as data processing and analysis of preference information, prediction information, etc. It is foreseeable that in the near future, more and more artificial intelligence products will come into thousands of households, which will bring convenience to people’s lives while also easily accessing more data information about personal privacy.

But when we cheer for the dawn of artificial intelligence, we should also realize that the development of artificial intelligence has encountered the following problems:

Privacy leaking

A cartoon created by Peter Steiner published in 1993-“On the Internet, no one knows that you are a dog”, This cartoon reflects an understanding of the Internet and emphasizes that users can send or receive information in a way that does not disclose personal information.

However, with the rapid development of artificial intelligence technology and its wide application in finance, transportation, commerce, medical and other fields, the value contained in big data has been continuously developed, resulting in huge economic benefits and social value. Big data is gradually changing the way people live and produce. But at the same time, people are deeply dependent on big data decision-making and lose control of their own data. In addition, artificial intelligence has greatly enhanced the ability of privacy intrusion, bringing more privacy access; user data abuse and privacy leakage The problems are endless.

Each of us’s mobile phone number, e-mail address, home address and company address, mobile phone identification code, consumption records, APP usage records, Internet browsing records, face-swiping records, fingerprints, etc. they are all private informations that we are not willing to give to anyone easily, but in the era of artificial intelligence, this has become a measure for a company to train AI algorithms.

It is the numerous inconspicuous personal privacy data consist of enough training materials for AI to learn from it and gain cognitive abilities, so that AI algorithms which haven’t met us but it can know and understand us, know the preferences and motivations of us. Even know our family members and friends. Our privacy is the “potential cost ” for achieving these intelligences.

PlatON’s privacy protection is based on modern cryptographic privacy computing technologies such as MPC, homomorphism, and zero-knowledge proofs. It provides a new computing model that makes data and models available and invisible, and data can still be achieved union and value sharing under the protection of data rights.

The huge cost behind the development of artificial intelligence

While advances in hardware and software have been driving down AI training costs by 37% per year, the size of AI models is growing much faster, 10x per year. As a result, total AI training costs continue to climb. ARK[8] believe that state-of-the-art AI training model costs are likely to increase 100-fold, from roughly $1 million today to more than $100 million by 2025.

The cost of data

Artificial intelligence learning needs marked data, but the marked data will cause huge labor cost. Extensive application and in-depth learning networks require a large number of labeled data for training to achieve the expected effect. However, in the era of big data, although there are a large amount of data, but most of them are unmarked data, marking these training data needs to be carried out manually.

In addition, checking and correcting data samples is as time-consuming as generating and annotating data samples. According to the Dimensional Research report, 66% of companies often encounter deviations and errors when gathering data. Some companies choose to adopt a complete internal method (make all the marks by themselves), and some companies choose to outsource and mix internally. The second situation is to outsource most of the work, and then individual personnel are responsible for verification and cleaning.

The cost of Algorithm

The development of the industry also means the influx of industry talents, and the rapid development of artificial intelligence has also made artificial intelligence engineers become popular. “AI algorithm engineer = high salary” This is a fact recognized by the public. The cost of the algorithm is calculated based on the complexity, the difficulty of the algorithm, the height of the requirements, etc. If Google, Ali, and Huawei do this, the investment is at least tens of millions of dollars.

In September 2019, major Internet recruitment companies successively released the 2019 talent employment trend report, among which AI talents topped the list with average monthly salary.

The cost of computing power

Although the cost of computing power has declined in recent years, infrastructure is already expensive. Training an AI model may require hundreds of thousands of dollars or even more computing resources. And because the data of the Al model will change over time, retraining will also bring continuous costs. Model inference is also more complicated in calculation. Al often involves data processing such as image, audio or video, which requires higher storage resources and processing costs. For some companies, the Al model must be transferred between regional clouds, and cloud computing operations are more complicated and costly.

While AI has made tremendous progress, the benefits of AI are not widely used, AI has not yet been democratized, and there is a trend toward increasing centralization.

Most AI research is controlled by a handful of tech giants. Independent developers of AI have no readily available way to monetize their creations. Usually, their most lucrative option is to sell their technology to one of the tech giants, leading to control of the technology becoming even more concentrated.

A few tech giants have monopolized the upstream of data by providing services to consumers, gaining unprecedented access to data, training high-end AI models and incorporating them into their ecosystem, further increasing the dependence of users and other companies on the five giants. Except for a few tech giants, other market players such as small and innovative companies find it difficult to collect large-scale data, and even if they obtain data at a significant cost, they lack effective usage scenarios and are unable to exchange them, making it difficult to precisely align with relevant AI learning networks.

Someone once said: “Data is the oil of the new era.” In fact, with the development of technologies such as the Internet and big data, in an artificial intelligence era of “Internet of Everything, Everyone Online, and Everything Algorithms”, the value of data has already surpassed that of oil. Data is the foundation of artificial intelligence, and algorithms are the essence of artificial intelligence. The smarter the artificial intelligence, the more it depends on the feeding of data and the support of algorithms. The accumulation of data is increasing, and the growth rate of data is getting faster, but more and more. There is a strange phenomenon that more and more excellent algorithms have no data to train.

On the one hand, because of the increasing value of data and the frequent occurrence of data breaches, everyone’s awareness of data sovereignty and data privacy protection is increasing day by day, and companies can no longer collect and use user data unscrupulously as before.

On the other hand: enterprises also begin to regard data as their core assets. On the premise that data sovereignty and privacy cannot be protected, they are unwilling to share and utilize their own data, whether from the perspective of safeguarding their own interests or complying with laws and regulations.

The effect on the environment

Artificial intelligence is often compared to the petroleum industry. Once data is mined and refined, it can become a highly profitable commodity. According to the results of the latest paper, the energy consumption of training an AI model is as much as the total carbon emitted during the life cycle of five ordinary cars.

This result is also unexpected by many AI researchers. A computer scientist from the University of A Coruña in Spain said: “Although many of us have an abstract and vague concept of this (energy consumption), these figures show that the facts are more serious than we thought. Or other AI Researchers may not have thought that this would have such a big impact on the environment.”

In addition, the researchers pointed out that these numbers are only the basis, because the workload required to train a single model is less. In practice, most researchers will develop new models from scratch or change data materials for existing models, all of which require More time for training and adjustments, in other words, this will result in higher energy consumption.

Such high prices make it impossible for individuals, small teams, and start-ups who want to solve new problems or automate processes and decision-making, resulting in almost deadlocks.

PlatON integrates blockchain, privacy computing and artificial intelligence technology, and is committed to solving the bottleneck of artificial intelligence development. Promote the compliant circulation of data, break the data monopoly, and accumulate more data with better quality and lower cost than the technology giants through decentralization.

1.Decentralization

Any user and node can connect to the network permissionless. Any data, algorithms and computing power can be securely shared, connected and traded. Anyone can develop and use artificial intelligence applications.

2.Privacy-preserving

Modern cryptography-based privacy-preserving computation techniques provide a new computing paradigm that makes data and models available but not visible, allowing privacy to be fully protected and data rights to be safeguarded.

3.High-performance

High-performance asynchronous BFT consensus is achieved through optimization methods such as pipeline verification, parallel verification, and aggregated signatures, and its safety, liveness, and responsiveness are proven using formal verification methods.

4.Low training costs

• With blockchain and privacy-preserving computation technologies, anyone can share data and algorithms in a secure and frictionless marketplace, truly reducing marginal costs and drastically reducing training costs.
• Realized the compliant circulation of data, break the data monopoly, and accumulate more data with better quality and lower cost than the technology giants through decentralization.

5.Low development threshold

Visualize AI model development and debugging, automated machine learning (AutoML), MLOps simplifies the whole process of managing AI models from model development, training to deployment, reducing the development threshold of AI models and improving development efficiency.

6.Regulatable and auditable

All data, variables, and processes used in the AI training decision-making process have tamper-evident records that can be tracked and audited. The use of privacy-preserving technologies allows the use of data to satisfy regulatory regulations such as the right to be forgotten, the right to portability, conditional authorization, and minimal collection.

Coal mines in the steam age, electric power in the electrical age and computer technology in the science and technology age have engines to push it forward in each era. In the future, the decentralized privacy artificial intelligence network based on blockchain will become the engine to push forward the era of artificial intelligence.

Artificial intelligence is both the present and the future. It is no longer an image and concept in movies. No matter how many people understand or do not understand, AI will revolutionize the human beings who created AI; as for the future, no one knows what will happen, like the plot in Matrix? Human beings are eventually usurped by machines? We don’t know; but one thing is certain, the research and development that mankind is diligently seeking, this era will finally come.

👇👇👇

PlatONWorld — the base of PlatON ecological long-termists!

Welcome everyone to continue to pay attention to PlatONWorld!

PlatONWorld | Official website platonworld.org

PlatONWorld | Telegram

PlatONWorld | WeChat：mcqueen678