1.1 介绍

1.1.1 背景

人工智能的历史可以追溯到希腊神话的古典时代，当时机器和机械人的思想得到了很好的发展。

约翰·麦卡锡 （John McCarthy） 于 1956 年在达特茅斯会议上首次使用“人工智能”一词。第一个人工智能实验室建于 1959 年，即成立三年后。人工智能研究的时代始于这一时期。麻省理工学院实验室目前已开放，是第一个人工智能研究中心。第一台机器人被添加到“1960 年的 GM 装配线”中。

1961 年，第一个聊天机器人被开发出来。Alexa 和 Siri 现在已经司空见惯了。Eliza 是一个聊天机器人，最早创建于 1961 年。著名的 IBM Deep Blue 紧随其后。1997 年，“IBM 的深蓝”击败了国际象棋世界冠军“加里·卡斯帕罗夫”的消息传开。这被认为是人工智能的早期成功。

斯坦福赛车队的机器人汽车 Stanley 为 2005 年 DARPA 大挑战赛而制造，获得第一名。这是人工智能的又一项重大成就。Brad Rutter 和 Ken Jennings 是 Jeopardy 游戏中最伟大的两位获胜者，他们在 2011 年被 IBM 的问答机 Watson 击败。因此，人工智能的概念首先是在虚构的环境中引入的。人工智能是当今使用的最相关的技术之一。如果环顾四周，当今的许多事物都是由 AI 深度学习或机器学习提供支持的。自 1950 年代人工智能 （AI） 首次出现以来，它呈指数级增长并具有巨大的潜力。除了其他密切相关的主题外，AI 还涵盖“机器学习、深度学习、神经网络、自然语言处理、基于知识的系统和专家系统”等学科。AI 正在用于图像处理和计算机视觉。现在，如果人工智能已经存在了 50 多年，为什么它现在变得如此重要和受欢迎？

第一个原因是因为个人现在可以获得大量的计算和数据存储能力，从而能够利用 AI 的能力。人工智能的处理能力要求很高。人工智能现在在商业上是可行的，可以解决影响广泛日常活动的问题，这要归功于最近在指数级增加“计算和数据存储能力”方面的许多进展，同时成倍地降低了与这些“计算和数据存储能力”相关的成本。

第二个最重要的因素是，随着世界变得越来越数字化，人们现在可以访问随着时间的推移产生的大量数据。它一直在以深不可测的速度生产数据。它通过社交媒体、物联网 （IoT） 设备和任何其他可用的方式收集数据。因此，它需要想出一种方法或解决方案来帮助分析这些海量数据，使个人能够提取相关洞察并根据这些洞察扩展业务。例如，开发一个有用的 AI 代理，该代理能够进行智能选择，例如预测在线购物时下一步要推荐哪些产品或弄清楚如何从图片中识别对象。当在大型数据集上教授人工智能时，所有这些都是可以想象的，而大数据使这变得更加容易。

第三个原因是它现在拥有更强大的算法，可以从海量数据集中提取信息。这些强大的算法基于神经网络概念。这是深度学习的核心原则。由于更好的算法可以更精确地执行更好、更快的计算，从而带来更多更好的商业机会和更高的盈利能力，因此对 AI 的需求有所增加。

在本章中，作者的研究范围仅限于德国、加利福尼亚和印度，作者强调了人工智能何时应该获得法人资格以及何时需要获得法律地位。其次，在本章中，作者强调了 AV 卷入事故时产生的民事和刑事法律责任。最后，作者强调了 AI 的道德问题。

此外，作者将进一步讨论人工智能技术产生的挑战和问题是否可以在现有的民法和刑法（程序法和实体法）中解决，或者是否需要制定人工智能技术专门法。它是对德国、加利福尼亚和印度涉及自动驾驶汽车人工智能相关法律和道德问题的法律的并列研究，确定了设计者、开发商/程序员、制造商、生产商、用户或 AI 在自动驾驶汽车 （AV） 方面的法律责任。

作者之所以选择这个主题，是因为 AV 技术旨在通过降低事故率和提供额外的优势来彻底改变移动生态系统，例如减少交通、提高安全性和移动性、降低能源使用和产生空闲时间的潜力。各章研究的性质是教义和定性。

那么，为什么特别讨论德国呢？因为考虑到自动驾驶系统和其他领域的新发展，德国在法律上已经准备好或准备继续作为顶级汽车行业之一留在市场上。

此外，接下来讨论加利福尼亚，加利福尼亚州是一个以其技术和创新数字发明而闻名的城市，因为 Facebook、Apple、Amazon 和 Google 的总部位于加利福尼亚。考虑到这些不仅是最有价值的高科技 IT 公司，而且是最大和蓬勃发展的公司。加州为其行业提供了一个特别适应性强、充满活力和尖端的法律框架，该框架在经济和立法者之间的积极合作的帮助下付诸实施。

最后，为什么是印度？所以原因是我们国家像美国、英国和加拿大一样遵循普通法，而德国则遵循民法。虽然迄今为止还没有关于人工智能技术/系统的具体立法，但将有一个涵盖所有法律责任问题的立法框架。

因此，本书的范围仅限于专门讨论和确定自动驾驶汽车的责任，而不是一般的所有类型的 AI。

1.1.2 什么是人工智能？

“John McCarthy”于 1956 年在达特茅斯会议上首次使用“人工智能”一词。那一年，人工智能诞生了。

John McCarthy 用以下的话来解释 AI：“AI 是制造智能机器的科学和工程。换句话说，创建能够执行通常需要人类知识的任务（例如话语确认、视觉辨别、导航和语言解释）的计算机框架的关键是有知觉的思考。

因此，人工智能 （AI） 是一系列技术，让机器能够像人一样思考和行动。人工智能最近通过生产用于许多不同行业（包括医疗保健、无人机、农业等）的工具和自动化系统使这成为可能。Google 预测搜索引擎是人工智能在工作中最著名的例子之一。当用户开始输入搜索查询时，Google 会建议选项供他们选择。因此，预测搜索基于 Google 收集的有关用户的信息，包括浏览历史、位置、年龄和其他私人数据。谷歌使用人工智能来尝试推断它在寻找什么。

1.1.2.1 人工智能在不同领域的应用

Moving on, the Contract Intelligence Platform from JPMorgan Chase analyses legal documents using “ML, AI, and Image Recognition Software.” For instance, it took over 36,000 h to hand examine about 12,000 agreements, which is a long time for work that could have been done more quickly for a better company. However, once an AI system took over this task, it was finished in only a few seconds. So that’s how AI differs from natural intelligence; despite the fact that AI cannot think or reason as people do, it has much greater computational capability than humans do due to the sheer processing power of computers needed to run machine learning algorithms and robust learning. The most complex of complex issues may now be calculated by AI in only a few seconds.

‘IBM’ is one of the industry leaders who created AI software exclusively for the medical field. IBM Artificial Intelligence Product, which is essentially ‘IBM Watson’, is utilised by over 230 healthcare facilities. In 2016, ‘IBM Watson’ was able to quickly compare 20 million oncology records and properly identify a patient’s uncommon leukaemia disease. In essence, 20 million records were processed in a matter of seconds to minutes. This demonstrates how significant Artificial Intelligence has become because it has begun to affect every aspect of people’s lives. Another project involves Google collaborating with an Indian network of eye care facilities to create an ‘artificial intelligence system’ that will analyse ‘retinal images’ and be able to diagnose conditions.

Another example is the use of artificial intelligence in face recognition on social media sites like Facebook, where it uses the “principles of machine learning and deep learning” to identify both can-tag friends and include face features. People rarely realise how frequently artificial intelligence (AI) is used in their daily lives, yet the auto-tagging tool that they use in Facebook is actually powered by “machine learning, deep learning, and neural networks.” In fact, artificial intelligence plays a significant role in the daily operations of all social media platforms, including Instagram, Facebook, Twitter, and others. Another illustration of this is the AI employed by Twitter to identify any ‘terrorism-related and hate content’. Recently, Twitter found nearly 300,000 accounts connected to terrorism, and 95 per cent of those accounts were actually made by non-humans, implying that these accounts were made by artificially intelligent machines.Footnote 1

Another well-known application of AI is the virtual agents Siri and Alexa. The recently released Google Duplex virtual assistanthas astounded millions of people worldwide. It can not only answer calls and make appointments for specific people. But it may also sound incredibly lifelike by applying human filters to add a human touch to its operation. As a result, it can be difficult to tell it apart from humans when speaking on the phone.Footnote 2

Autonomous Vehicles are another well-known application of artificial intelligence. These fully automated self-driving cars use “computer vision, image detection and processing, and deep learning in conjunction with artificial intelligence to automatically detect any objects or obstacles and navigate around them.” Many Big Tech firms, including Google, Apple, Microsoft, Tesla, Facebook, and others, are essentially data-driven businesses that heavily rely on AI technology.Footnote 3

Artificial intelligence is evolving into a more human-like communication tool by responding contextually and mimicking human speech patterns. AI is evolving into a very human-like being, expressing itself through artistic creations and the swiftly developed solutions it comes up with. The degree to which AI mimics human behaviour and can fool us with persuasive content despite our knowledge that it is delusional is proof that it is evolving into a human. Will we be able to accept that AI is evolving into a human? Or will our demise result from humanising AI? The catalyst for humanity’s cataclysmic demise. However, we will undoubtedly need to shape, navigate, assess, mitigate, integrate, validate, and, most crucially, govern if AI is to become human.

1.1.2.2 Definitions of Artificial Intelligence

John McCarthy, the discipline’s founder, stated the concept in 1955 as “that of making a machine behave in ways that would be called intelligent if a human were so behaving.”Footnote 4

An onion chart illustrates the relationship between deep learning, machine learning, and artificial intelligence, showing that deep learning is a subset of machine learning, which in turn is a subset of artificial intelligence.

What is ‘Artificial’ aND ‘Intelligence’?

Artificial Intelligence is a study of how to make a system, which can think, behave, and act exactly or better than what a human being can act or react.

Black’s Law Dictionary

‘Black’s Law Dictionary’s’ explanation of artificial intelligence is “software used to make computers and robots work better than humans. The systems are rule based or neutral networks. It is used to help make new products, robotics, human language understanding, and computer vision.”

Bellman explained it as “the automation of activities that we associate with human thinking, activities such as decision-making, problem-solving, and learning.”

Kurzweil explained it as “the art of creating machines that perform functions that require intelligence when performed by people.”

Author’s Understanding of AI

AI is basically a man-made system, processed and created in such a fashion that made things autonomous in nature, either partial or full automation, in conformity with the ethical rules and deep neural network norms.

According to the author’s view, AI is an intelligent technology. However, that intelligence is artificial, which is being processed by humans through the help of algorithms (which is step-by-step) mathematical calculations, i.e., algorithms. The nature of computer programs is such that they flow data. It makes patterns out of similar data, which is very much known as machine learning, learning through the help of big data making similar patterns out of massive data, which eventually make sense out of it at the end and helping machines to adapt, act, and behave in the same way, whenever the situation demands, or act in the same way whenever the similar situation comes in.

Algorithms → Machine Learning which includes (Deep learning + Neural networks + Speech recognition + voice recognition + virtual agents = Artificial Intelligence).

Now, the question is who is programming the algorithms?

It is a human being who is putting all efforts to make the machine learn! Learn from its experiences. Ultimately, humans are mastering the machine to make it intelligent so that it functions on its own with time. However, making a machine intelligent and independent would paralyse the human soon, losing its control and giving all reliance on the machine for their respective work, so how much reliance should individuals have on AI algorithms? Suppose the system is working ideally 99 times. On the 100th attempt if it crashes or fails in the system, then can it be said that it can completely rely and depend on AI? The failure of the 100th attempt made people realise that the reliance on AI algorithms system needs a careful check and accuracy in the system with no glitch of whatsoever nature.

The majority of AI authors concur that the approach taken is just as crucial as if the problem is solved. So, another mark of intelligence is how it fails gracefully. Understanding and respecting human limits and making plausible errors are hallmarks of expertise.

Types of Artificial Intelligence: Living in the Age of Artificial Narrow Intelligence

Artificial intelligence is classified into three distinct evolutionary stages, or one could say that AI is in three stages. It naturally has narrow artificial intelligence followed by artificial general intelligence, followed by artificial super-intelligence.

Types of AI:

Artificial Narrow Intelligence

Focuses on one area and a single problem—Machine Learning

Example: Siri, Alexa, Cortana, and AVs.

Artificial General Intelligence

Refers to a computer that is completely equal in intelligence to a person—Machine Intelligence.

Artificial Super Intelligence

An intellect that, in almost every discipline, is significantly smarter than the best human minds. —Machine Consciousness.

1.1.2.3 Needs Revisiting—Personhood

Artificial Intelligence is the central factor for the disruptive shift of the 4.0 industrial revolution. Similarly, Artificial Intelligence frameworks are still at an exceptionally beginning stage. The attribution of any rights, duties, or obligations to AI frameworks is a troubling position that the Indian legal system seems unaware of. The legitimate status of an individual or substance is straightforwardly connected to its autonomy. This status is not simply conceded to people yet collaboration, companies just as associations or organisations.

However, with regards to artificial intelligence, then, at that point, it is anything but perceived at this point by any general set of laws as a lawful substance aside from “Saudi Arabia wherein a robot called Sophia,” which the state has interpreted as a falsely intelligent humanoid with rights and obligations that are the same as those of humans and as a respectable person residing within the state. Artificially intelligent robots or programs may be given legal status depending on whether it is possible to grant them specific rights and obligations that a living person is typically reliant upon.

In the House of Lords,Footnote 5 the UK constituted the AI Committee. Concerning AI’s legal definition and legal status as a person, the US government does not struggle to consider these issues.

“Section 3 of the European Bill, 2017” on AI gives the general definitions of AI as followsFootnote 6:

  • “Artificial systems capable of performing tasks without human presence (autonomous systems).”

  • “Systems that think as by analogy with the human brain and are able to pass the Turing testFootnote 7 or another comparable test by processing natural language, representing knowledge, automated reasoning, and learning.”

  • “Systems that act rationally achieve goals through perception, planning, reasoning, learning, communication, decision-making, and action.”Footnote 8

As per European Union countries, they pay very particular attention to making legal regulations for self-driving cars. The German Traffic ActFootnote 9 puts the responsibility on the owner to manage and work on an automated or semi-automated car as it contemplates only a partial involvement of the ‘Federal Ministry of Transport and the Digital Infrastructure’. As presented in the EU resolution on robotics (‘European Parliament Resolution, 2017 on Civil’), it talked about the most current, comprehensive, and conducive approach to the definition of present and potential legislation in robotics. It explains the types of AI use and ethics, covering all the liability issues, and for operators, developers, and manufacturers in robotics, it provides basic rules of conduct. These norms are based on three ‘laws of robot technology’.Footnote 10

Is it possible to give AI Legal Personality?

A collection of obligations and rights—Legal personality is a fiction created by individuals through the legal system or general laws. In this way, it has the freedom to decide what it should apply and what its composition should be. A key US decision from the eighteenth century focused on corporations’ and companies’ unique legal personalities, ‘Trustees of Dartmouth College v. Woodward,Footnote 11 Chief Justice Marshall’ communicated the idea as follows:

A corporation is an ‘artificial personality’ that only exists in the eyes of the law and is intangible and elusive. Being a straightforward creature of law, it has only the characteristics that the approval of its formation bestows onto it, whether intentionally or accidentally. The rights and obligations of the latter are consequently functionally tied to those of the former. Persons in charge of or acting on behalf of the organisation are held accountable for both the physical and mental components of any wrongdoings. In this context, it’s important to recognise that the legal personhood of corporations is not without limitations because the corporate curtain can be lifted to ignore it. A corporation may be penalised in order to punish those who oversee, manage, or possess it. Corporations and people that conceal themselves behind incorporation are subject to administrative, civil, and even criminal responsibility regimes. In the modern environment, where evolution is occurring at a faster rate than ever before, it is simple to see how constrained systems using artificial intelligence are made to do specific tasks in a timely manner.

A live person is independent or autonomous and has the capacity to make their own decisions. But an AI framework is established by persons and fills in according to the bearings of the projects that have been added to its framework to execute target jobs in a specific way. Companies and corporations are nonetheless accountable to their partners for any risk that may later develop as a result. Even if companies portray the situation as a different legal entity, they are nonetheless accountable to their partners for any risk that may later develop as a result of the trades that these businesses or organisations/corporations have made.

Challenges to AI as a Separate Entity—Not Equal to Human Being

Treating AI as a brand-new legal field that requires its own set of legal standards is a difficulty. The EU resolution does not nudge into this issue which is very much possible in the robotic application.Footnote 12

The question of granting legal status is seen chiefly as theoretical. Legal persons, according to Joanna Bryson,Footnote 13 Mihalis Diamantis, and Thomas Grant, are “fictive, divisible, and not necessarily accountable.” They claim that “legal personality is an artifice,” and that “[l]egal people need not possess all the same rights and obligations, even within the same system.”Footnote 14

A legal person serves as a tool to help natural persons more effectively pursue their objectives. What objective—and whose objective—would be served if AIs were given legal status or specific rights? No matter how ‘aware’ or ‘intelligent’ the AI is, it will never be able to acquire legal status. Intentionality, free will, and consciousness are only a few examples of human characteristics that are thought to be necessary for moral and legal responsibility but are not present in any machine. Of course, it is feasible to question this approach’s metaphysical underpinnings.

Although intentionality and free will are typically the foundations of legal accountability, ‘guilty mind’ and ‘autonomy’ frequently, they get treated as if they are real physical entities. Purely human traits, such as intention and free will, are only ‘things’ that people assign to one another in order to preserve order in their social relationships. It is a conceptual construct, not a thing that is ‘actual’, and its presence may be verified and quantified. Therefore, it cannot be a prerequisite for legal (or moral) culpability. Specific human characteristics, like consciousness and self-awareness, have been deemed unimportant from a legal standpoint, which immediately discards the limiting approach. Some people have traditionally been denied legal standing, whereas states or corporations that do not possess any human traits have been given such persons.

Therefore, it is not required to determine what computers must be able to do (in regards to imitating human characteristics or abilities) or what attributes they must possess in order to be recognised as individuals, writers, or innovators. Anyone may be given rights by the law. In the past, some legal systems have refused to acknowledge some individuals as legal persons or denied them a fundamental right, while acknowledging other entities as persons or granting them rights—even when the latter did not possess any human qualities. The legal system does not recognise something as a legal person based only on its inherent qualities since legal personhood is a fiction. Because of this, any item, including algorithms and computer systems, could be given legal personality, regardless of its level of creativity, intelligence, or consciousness. Reiterating that this grant is not based on any inherent qualities of the algorithm or machine, the legislature is entitled to introduce any legal interpretations that they deem necessary to achieve their goals. Despite the foregoing, anything other than a human cannot be given legal personality without specific arguments or ‘good reasons’. The following inquiries come up: What would be the point of giving an algorithm individual identity? What would the advantages and hazards be? It would be necessary to choose the criteria for differentiating between algorithms that require personhood and those that do not since there are neither natural nor legal requirements for personhood.

Another quality would have to be chosen because the vague notion of autonomy is inappropriate—if only because it is gradable and necessitates choosing the degree of autonomy that validates personhood. Would it have to do with how an artificial intelligence (AI) work is judged in terms of its creativity, originality, or emotional resonance? Which component of the system would be viewed as a human from a technical standpoint? Remember that an AI never exists in a vacuum and that its interests are just one component of a much larger system. How much, if any, liability would the algorithm’s designer still bear for the algorithm’s performance?

Furthermore, it is necessary to consider that algorithms do not generate themselves and that autonomous systems are not the result of the technological equivalent of a technically perfect idea.

In conclusion, giving AIs legal personhood raises a plethora of practical and technical issues. Last but not least, it’s imperative to keep in mind that the law must be logically consistent.

1.1.3 The Legitimate Status of AI: Concluding Remarks

So, according to the study, Artificial Intelligence cannot be awarded legal personhood until and unless it becomes highly and fully autonomous in nature, in terms of autonomous vehicles, and in general, when it reaches the “level of Artificial General Intelligence and Artificial Super Intelligence, where Artificial Intelligence can make decisions on its own independently.”

Otherwise, even for granting the status of electronic personality, or legal personality, it must meet the criteria of being independent in its own actions and decisions, being able to sue and be sued, and “Artificial Intelligence is still at the stage of Artificial Narrow Intelligence,” where it is not even capable of perceiving and making decisions without human intervention.