The U.S. and China are locked in an increasingly heated struggle for superpower status. Many perceived this confrontation initially only through the lenses of a trade war. However, the ZTE “saga” already indicated the issue was broader and involved a battle for supremacy over 21st century technologies and, relatedly, for international power (see 当AI开始创造AI时--人工智能与计算能力, 2018年5月7日)。

The Sino-American technological battle increasingly looks like a fight to the death, with the offensive against Huawei aiming notably to protect future 5G networks (Cassell Bryan-Low, Colin Packham, David Lague, Steve Stecklow And Jack Stubbs, “中国的挑战:5G之争", 路透社的调查, 21 May 2019). For Huawei and China, as well as for the world, consequences are far reaching, as, after Google “stopping Huawei’s Android license”, and an Intel and Qualcomm ban, the British chip designer ARM, held notably by Japanese Softbank, now stops relations with Huawei (Paul Sandle, “ARM停止供应对中国科技巨头华为造成新的打击", 路透社, 22 May 2019; “DealBook简报。华为的反击是全球性的", 纽约时报, 23 May 2019; Tom Warren, “华为的安卓和Windows替代品注定要失败", The Verge, 23 May 2019).

The highly possible coming American move against Chinese Hikvision, one of the largest world producers of video surveillance systems involving notably “artificial intelligence, speech monitoring and genetic testing” would only further confirm the American offensive (Doina Chiacu, Stella Qi, “Trump says ‘dangerous’ Huawei could be included in U.S.-China trade deal", 路透社, 2019年5月23日; Ana Swanson  Edward Wong, “特朗普政府可能将中国的监控公司海康威视列入黑名单", 纽约时报,2019年5月21日)。

China, for its part, answers to both the trade war and the technological fight with an ideologically martial mobilisation of its population along the lines of “People’s War”, “The Long March”, and changing TV scheduling to broadcast war films (Iris Zhao and Alan Weedon, “在中美贸易战中,中国电视台突然切换到反美电影的排片。“, ABC News, 20 May 2019; 迈克尔-马蒂纳大卫-劳德, "Prepare for difficult times, China’s Xi urges as trade war simmers", 路透社报道称,2019年5月22日)。这凸显了对中东国家来说有多大的利害关系,正如我们之前解释的那样(★传感器和执行器(4)。人工智能、先进机器人的长征与地缘政治).

这些举措强调了所涉及的巨大利益。事实上,新技术,从多种形式的人工智能(AI)到物联网(IoT)和通信,通过量子信息科学(QIS)和技术,参与了一个范式的变化,其中也包括治理、国际权力和可能进行的战争的方式和胜利。

内容
  1. 人类如何成为AI-代理的执行者
    1. 案例研究
      1. The example of Google DeepMind’s Go Game
      2. 乘飞机旅行
      3. 智能家居的案例
    2. 训练人类不假思索的行动
  2. 通过屈服于敌人来赢得战争:对一个乌托邦场景的反思
  3. 从连接世界到改变世界的平衡
    1. 改变世界,克服困难
      1. 数字网关
      2. 非物质化的世界
      3. 欢迎来到矩阵
    2. 世界的不可能完全非物质化和脆弱性
      1. 不存在单纯的数字世界
      2. 能源,数字化的物理隐藏部分
      3. IT企业、与气候变化有关的灾害和责任
    3. Users’ countries pay the bill, systemic threats and a strategic twist
      1. Users’ countries and systemic threats
      2. 战略性的转折

Here, we shall focus on such possible new faces of security in general and war in particular. These potential – and already operating – changes stem from the complex dynamics that have been unleashed. As we found out 之前在向先进机器人艰难迈进的过程中,加上利益相关者对获得人工智能(AI)系统的强烈兴趣,特别是 深度学习 (DL),它们的运作和盈利,导致了一个意外的后果。人类本身正越来越多地被拖入人工智能代理的生态系统中。他们实际上被变成了算法的执行者。

我们将首先审视正在发生的事情,并解释人类如何成为人工智能代理的执行者,并举例说明。然后,我们将勾勒出一个场景,解释这种演变如何导致一个乌托邦式的未来,即掌握了人工智能代理的国家行为者可以用一种新的方式赢得战争。

Third, we shall turn to the digital and the material worlds and to the bridges between them. We shall highlight that the need to see AI developing will also lead to a further dematerialisation of the world, with, as extreme, virtual reality. However, we shall explain that total dematerialisation is impossible and comes with a major hidden cost, rising energy consumption, with thus impacts on climate-change. We shall also highlight how users’ countries bear the brunt of the burden and face major systemic threats. Finally, we shall identify a way for them to preempt these systemic threats, in an interesting strategic twist.

人类如何成为AI-代理的执行者

案例研究

The example of Google DeepMind’s Go Game

First let us take again our initial example of Google DeepMind’s go game (see 在现实中插入人工智能).正如我们所解释的,游戏的设置看起来如下。

谷歌DeepMind的视频截图。突破性的AlphaGo掌握了围棋游戏 - 1:19

We pointed out that to see DeepMind’s AI-agent becoming fully operational, one had to provide a sensor to replace the lady in C and an actuator instead of the gentleman in A.

然而,要想在A中得到一个执行器,我们最好需要一个先进的机器人。正如所见,这样复杂的先进机器人还没有出现(★ Sensor and Actuator (4)…).我们离获得先进机器人的种类还有很长的路要走,我们将需要许多人工智能/DL-代理人的执行任务(同上)。

因此,正在发生的是,A在不久的将来仍然是一个人,而人工智能和值得注意的DL将继续扩大,因为利益相关者需要它们的扩张(同上)。

换句话说,促进人工智能代理及其使用的利益相关者,为了克服目前仍然缺乏的非人类执行器,将把人工智能旨在帮助的人类变成这些人工智能代理的执行器。

乘飞机旅行

作为另一个例子,让我们采取一系列旨在销售机票的人工智能代理。旅行者的最终目的是能够从家到P地。通过一系列的数字化过程和使用各种算法,最好的算法是深度学习的那种,未来的旅行者将看到一系列的目的地和航空公司的路线和机票。他/她将选择一个,然后向航空公司支付她/他的机票。

如果我们的旅客有一部智能手机,那么他/她将能够在她或他的智能手机上获得机票。如果没有,他/她将不得不打印。在机场,如果没有智能手机,他/她将不得不打印一张登机牌。

在任何情况下,他/她都必须打印行李票。根据机场现有的机器人,他/她将不得不扫描登机牌和行李票,将行李放在传送带上,检查重量,或者,将行李拖入机器人,然后检查行李票和重量。

最后,他/她要准备好通过安全检查。

对于大多数步骤,我们可以看到智能先进设备的缺失是如何由用户,即人类来弥补的。用户已经变成了航空公司的人工智能代理的执行者,同时也取代了航空公司的前雇员。此外,当一个智能设备投入使用时,消费者或用户是那些必须购买智能设备的人。因此,他们现在也承担了曾经由公司支付的部分投资。

我们在智能农业的案例中观察到了类似的情况,当时先进的农业机械还没有出现,也没有在整个人工智能驱动的过程中建立起来(见 ★人工智能、物联网和农业的未来。智能农业安全?(1)(2)).

智能家居的案例

当我们看到智能家居及其中的一些组件时,情况就不那么清楚了,例如著名的人工智能助手亚马逊Alexa,它与智能扬声器Echo或Google assistant连接,通过扬声器Google Home与从手机到平板的物联网设备相连。

We can imagine that one of these assistants could voice a suggestion such as “to reach on time the place where you must meet this client or that person, you should depart now and drive according to this itinerary.”

第一个系列的执行器将开始工作,将DL算法的结果翻译成一系列的句子,并以人类议程的方式排序。然后,其他执行器将以人类可以听到和理解的方式来表达这些建议。在其他情况下,如果没有语音功能,那么建议就可以显示在屏幕上。

可能,接受建议的人将认为人工智能助手在帮助她或他,这很可能是真的。

Yet, from the point of view of the AI-agents, the individual would also be acting on AI-agents’ suggestions. The individual would making the AI-agents output exist in the physical world.

训练人类不假思索的行动

What feels disturbing from a human being point of view is that our own “cognition to action” sequence, built over 40.000 years if we consider only the Cro-magnon (大英百科全书》(Encyclopaedia Britannica)),被打破了。简而言之,如果我们对导致我们行动的顺序做一个非常简单的评估,我们或多或少有以下模式:感知世界,分析收集的数据,根据分析做出决定,采取行动。这个模式应该利用现有的研究来完善。然而,不管有什么发现和最新的研究,有了人工智能助手,我们通常的过程被改变了,其中的一部分被移除。

在我们的案例中,人工智能代理进行分析,然后提出决策的可能性。这是为了让我们放心,让我们相信我们可以自由决定是否行动,然后采取相应行动。

However, deciding without any control over inputs and analysis, then acting upon this decision goes very much against the efforts at understanding, knowledge and education of thousands of years of history. It “feels” as if we were transformed into at best children, at worst slaves … or robots. Even though decision remains into our hands, decision without awareness of the analysis is not a real decision and the door is opened to the possibility for any manipulation or error.

Hence, here, the absolute need to develop trust, as well as the capacity to “enter into” and supervise the analysis, i.e. to overcome the AI “blackbox problem” (e.g. Will Knight, “处于人工智能核心的黑暗秘密", 麻省理工学院技术评论, 2017年4月11日)

事实上,企业部门拥有人工智能代理,因此会首先为自己的利益而使用它,其次才是为其客户的利益,这只会加剧问题。几十年的广告和营销试图操纵消费者的决策过程,只会使问题变得更糟,更不用说几百年来为公司利益而进行的游说,往往是违背公众利益的。

因此,只有政治当局作为公共利益和每个公民(无论是个人还是法律实体)安全的保证发挥非常有力的作用,才能在最后为信任创造条件,而这种信任对于看到将人类变成执行者的人工智能代理安全发展绝对是必要的。

Moreover, it will be crucial to make sure that human capabilities are not lost in the meantime. A couple of authors evoke this possibility in the case of strategic decision-making, for example (Andrew Hill, “人工智能创造了真正的战略困境", 金融时报,2019年5月20日)。

通过屈服于敌人来赢得战争:对一个乌托邦场景的反思

可以想象一个反乌托邦式的场景,以突出这种可能的现实的一些特征。

新社会被一分为二。

Wealthier citizens and companies can buy the robots that then act in their place, when these advanced robots are available. In that case, these richer people save time and resources for a certain number of tasks, completely outsourced to AI-agents and their advanced robots actuators. True enough, in the meanwhile, they also abandon part of their power, as action – as in the German 马赫特 或在英语中 权力 – is fundamentally power. Yet, a few of them, those who are wise enough to do so, use the time spared for other, more evolved tasks.

贫穷的公民和企业,大多数人,越来越多地变成了人工智能代理和他们的利益相关者的执行者。他们的意志力表面上保持着,但是,由于他们根据属于企业利益相关者的人工智能代理的建议和分析行事,他们事实上顺从于这些利益相关者的利益。

For instance, continuing with our previous example, when on his or her way to a meeting, the poorer citizen’s connected device will choose an itinerary that will come close to this or that shop. The device will then tell her or him that s/he needs to buy this very product, by chance available in that shop. On the contrary, our wealthier citizen, with his or her set of robots, will not have to go through this. S/he will find the products already delivered in his or her home.

It could appear as if the poorer people were actually better off in terms of freedom than the wealthier class. This is however questionable, because in the poorer people’s case, a habit to rely on something that tells you what to do without thinking is developed. Thus, the appearance of freedom of decision is indeed only an appearance. Then, once the habit is formed and, as a result, the capability to think before to act is progressively lost, the door is opened to any manipulation.

诚然,较富裕的人将被放在前面。 既成事实, but the very sequence leading from reflection to action will not have been broken and damaged. If – and this is a big if – the wealthier people use the time spared to educate themselves further, then they can escape another danger, which is to completely give up any mastery over some sectors of their life.

在这两种情况下,如果没有强有力的控制和保护,公民将面临失去部分人性和被转化为事物的巨大风险。他们可能会逐渐成为人工智能代理和他们的利益相关者的工具,而不会有任何反抗,因为这种转变是缓慢的,而且表面上是无害的。

现在,考虑一下,出售一系列人工智能代理的主要利益相关者是一个外国势力。或者,销售这些人工智能代理的企业可能是外国的,并且由于一系列原因,包括国家利益和国家安全,必须服从外国政治当局。

然后,该外国势力将几乎完全控制使用人工智能代理的人口。在战争的情况下,假设目标国家的军队和政治当局打算打仗,统治人工智能代理的外国行为者可以很容易地操纵使用的人口,无论是富人还是穷人,都可以根据他们被改造的方式。然后,军队可能会面临外部的攻击,特别是内部的大量敌人,因为人口可以通过各种方式来反对自己的军队。 侵略者将以最低的伤亡水平进行战斗,并可能取得胜利。

Considering the danger, political authorities – again assuming they are neither predatory nor “sold” to a stronger and more powerful actor – have an even greater interest in making sure the population they rule do not end up actually being ruled by others.

一般来说,这里的重点不是拒绝技术进步,也不是提高对人工智能的恐惧和敌意。重要的是要意识到风险,并努力确保我们采取正确的行动,以便我们最好地利用进步,同时减少意外的不利后果。

更具体地说,对于每一个政体来说,了解其中的利害关系是很重要的,以确保一个外来的、可能是负面的规则不会被强加给一个坦率的人口。即使没有侵略性的意图,建立起缓解外国统治的能力的可能性就应该敲响警钟,并引发保护性行动。

现在,正如从航空旅行的例子中可以清楚地看到,这里的关键不仅仅是人工智能代理的使用。这个问题更广泛,包括整个数字化过程,我们现在将探讨这个问题。

从连接世界到改变世界的平衡

最初,我们发现,一个AI-agent(或一系列)的传感器和执行器也是不同类型的世界或现实之间的桥梁(在现实中插入人工智能).

改变世界,克服困难

数字网关

We can have AIs that operate solely within the digital world. In that case, sensors and actuators bridge mainly different ways to understand the digital world. For example, a sensor will “read” a digital input initially intelligible to humans or to another device and make it intelligible to the AI-agent. The actuator will take the AI output and make it understandable digitally to whichever actor needs it, be it human or not.

The latest feat realised by Google DeepMind’s AI-agent AlphaStar, when it mastered Blizzard Game StarCraft II, exemplifies such digital-only environments (AlphaStar Team, “阿尔法星。掌握实时战略游戏《星际争霸II》的方法", DeepMind博客, 2019 – check their website for more photos and videos).

AlphaStar in action with the difference sequence, input and output – Check DeepMind的原始图像 用于动画

非物质化的世界

以更复杂的方式,我们有传感器和执行器,它们必须作为物理或物质世界与数字世界之间的桥梁。

面对弥合真正不同世界的困难,除了将人转化为执行者之外,一个前进的方向是将尽可能多的物理世界引入数字世界。这正是上述航空旅行的例子所描述的。

因此,我们可以预期,在未来的几年里,世界的数字化将得到更多的推动。事实上,我们之前已经看到了各利益相关方对进一步开发和制造可运营和可盈利的人工智能系统的兴趣,特别是涉及到DL的系统(见《世界日报》第三部分)。 ★传感器和执行器(4)。人工智能、先进机器人的长征与地缘政治).因此,这些行为者极有可能将人类变成执行者,同时也尽可能地减少对连接数字世界和物理世界的执行者的需求,这是一个双管齐下的战略。

世界的数字化成为一种非物质化,人类必须找到与之互动的方法。

欢迎来到矩阵

一个极端的演变是进一步发展所谓的虚拟现实,从而使人类进一步进入人工智能代理的世界。在这种情况下,执行器将被颠覆。它们将不再是作为从数字世界到物理世界的桥梁并允许人工智能代理在物理世界中输出的设备。它们将是连接物理世界和数字世界的设备,因此类似于传感器,并将人类带入人工智能代理的世界。

欢迎来到 "母体"!

The device could be external, as for example with the famous headsets (e.g. “2019年最佳VR头盔", PC杂志)或使用谷歌眼镜(最新一代 for businesses, released on 20 May 2019). They could even be implanted within human beings. They could be a mix between both, as with AlterEgo, “a non-invasive, wearable, peripheral neural interface that allows humans to converse in natural language with machines, artificial intelligence assistants, services, and other people without any voice…The feedback to the user is given through audio, via bone conduction.”…The sensors “captures peripheral neural signals when internal speech articulators are volitionally and neurologically activated” (AlterEgo网站; see also Lauren Golembiewski, “可穿戴式人工智能将如何放大人类的智慧", HBR,2019年4月30日)。

TED2019 | April 2019

世界的非物质化直至虚拟现实和执行器的反转,以及在这一阶段之前人类向执行器的转变,对武装部队有着至关重要的影响,因为其可能目标的性质发生了变化。因此,攻击和防御的目的、方式和手段也需要相应地改变。然而,正如下文所解释的,我们也不能过分夸大变化。考虑到开发新的武器系统和军备所需的时间,预测这种演变是至关重要的。

世界的不可能完全非物质化和脆弱性

现在让我们以示意图的方式看一下这些序列,但作为一个整体,作为一个系统,用下面的图来表示。

PDDP传感器执行器模型
人工智能、数字化和世界的非物质化

不存在单纯的数字世界

因为我们作为人类,生活在现实世界中,是有形的存在,那么,在某个阶段,即使是最初看起来只发生在数字世界的事情,也必须转化为有形的世界。 无论非物质化程度有多高,都必须与物理世界建立联系。.

无论非物质化程度有多高,都必须与物理世界建立联系。.

因此,实际上,从系统的角度看问题的正确方式不是设想两种类型的序列,一方面是数字-数字,另一方面是数字-物理,这两者是分开的。

我们所拥有的总是一个独特的序列,就世界或环境而言是物理-数字-数字-物理。如果数字部分等于零,那么我们就会找回经典的物理互动。但我们在任何情况下都不能去除这两个物理极端,即使我们认为我们只是在一个数字-数字序列中。

即使在广泛的虚拟现实的极端情况下,人类仍然需要看到他们的基本需求得到满足,如食物和饮料,正如电影中所描述的那样。 矩阵.他们的情感和认知过程必须保持健康,因为在 全面回顾.同时,数字系统将需要运作。

能源,数字化的物理隐藏部分

事实上,无论多么隐蔽,社会的非物质化总是伴随着与物理或物质世界的基本桥梁或联系。正如托马斯-霍默-狄克逊(Thomas Homer Dixon)所强调的那样,这种联系是对物质世界最基本的资源--能量的使用,这一点很有见地。下降的颠覆,灾难,创造力和文明的更新, 2006).

In this framework, Janine Morley, Kelly Widdicks, and Mike Hazas examine “the phenomenal growth in Internet traffic, as a trend with important implications for energy demand” (“Digitalisation, energy and data demand: The impact of Internet traffic on overall and peak electricity consumption”, 能源研究与社会科学, Volume 38, April 2018, Pages 128-137, https://doi.org/10.1016/j.erss.2018.01.018). Calling for an agenda to better understand and then mitigate “the most problematic projections of Internet energy use”, they also highlight 互联网和数字化带来的大量能源使用。,尽管考虑到不确定性,各种情况仍然是可能的。例如:

“Most estimates of ICT-related energy consumption also predict steady growth. For instance, Van Heddeghem et al. estimate that the electricity consumed by digital devices and infrastructures is growing faster (at 7% per year) than global electricity demand itself (at 3% per year), with the rate of growth of networks highest of all (at 10.4%). Andrae and Edler, also anticipating a compound rate of growth of 7% per year, calculate that the production and operation of ICT will rise to 21% of global electricity consumption by 2030: this is an absolute rise to 8000 TWh, from a base of around 2000 TWh in 2010. In a worst case scenario, this could reach as high as 50% of global electricity use by 2030, but only 8% in the best case.” 

W.Van Heddeghem, S. Lambert, B. Lannoo, et al. "2007至2012年全球ICT用电趋势",Comput.Commun., 50 (2014), pp. 64-76A and Andrae, T. Edler, "On global electricity usage of communication technology: trends to 2030", Challenges, 6 (2015), p. 117, quote by Morley, Widdicks, and Hazas.

如果存在提高能源效率的努力,到目前为止,它们还不能抵消能源使用的增长(同上)。

IT企业、与气候变化有关的灾害和责任

毋庸置疑,在气候变化方面的影响以及随后的多种相关不利影响也同样重要。

Could Microsoft’s move show not only full awareness of their heavy energy footprint but also of their participation in global climate-change related disasters?

As a further signal of this heavy energy footprint, with adverse climate-change-related impacts, “Microsoft has joined a conservative-led group that demands fossil fuel companies be granted legal immunity from attempts to claw back damages from the climate change they helped cause”… It thus “become[s] the first technology company to join the CLC [Climate Leadership Council], which includes oil giants BP, ExxonMobil, Shell, Total and ConocoPhillips among its founding members.” (Oliver Milman, “微软加入寻求扼杀历史性气候变化诉讼的团体", 卫报, 1 May 2019). Besides emphasising how much we have to take “corporate communication” with a pinch of salt, could Microsoft’s move show not only full awareness of their heavy energy footprint but also of their participation in global climate-change related disasters?

Users’ countries pay the bill, systemic threats and a strategic twist

Users’ countries and systemic threats

此外,莫利、威迪克斯和哈扎斯,强调了一个关键点。

“If accurate [the studies], this suggests that the bulk of energy consumption in Internet infrastructures takes place in the country of use.”

As a result, in the same dynamic, first the population (be it citizens or businesses) of countries using AI-systems are turned into actuators. Second, they see their world dematerialised and have to find ways to cope with it. Third, when they can, they also have to invest in expensive equipments if they want to avoid being completely “robotised”. Fourth, they also have to pay for upfront and hidden energy cost and their aftermath, through their energy bill and through their taxes.

显然,无论一个国家是非物质化和人工智能的生产者还是消费者,其对国家和人口的影响都是非常不同的。在人工智能和计算方面的领导地位和竞争,因此在影响力和市场份额方面也很重要。那些领先于竞赛和最有影响力的人对其他国家形成了巨大的和多维的力量。

对其他国家来说,只有意识到所面临挑战的强大政治当局才有希望解决这种对全体人民的系统性威胁。

因此,考虑到美国在这一问题上的巨大优势,以及中国为成为这一领域的领导者所做的巨大努力,两国的对抗就更像是一个合理的结果,甚至可以说是不可避免的。同时,其他国家,如果他们可以的话,最好尽早醒过来,做出一系列的反应,如果他们不想付出极其沉重的代价。

战略性的转折

在一个有趣的战略转折中,一方面是能源的依赖,另一方面是人类的依赖,对于整个系统来说,很可能是影响力较小的国家可能扮演的关键。

In other words, as a first type of responses, concerned political authorities could seek to educate a population not to fall prey to the worse cognitive impacts of “being turned into actuators”.

第二,行为者可以开发一系列的行动、工具和武器,旨在威胁非物质化世界的提供者和人工智能代理人的能源使用。然后,他们可以利用这些设备的存在作为一个先发制人的保险,以确保非物质化和人工智能供应商不违背他们的人口,或以一种对人口和国家有不利后果的方式行事。在需要的情况下,例如一场宣战,针对能源的行动,即数字和物理世界之间的桥梁,可以使敌人的整个数字大厦崩塌。

Finally, from a defensive and security perspective both human beings and the “energy bridge” must be secured in priority. This means also acting to make sure that climate change and its impacts, as well as energy depletion, do not finally destroy those that contributed to the spread of these existential threats for the Earth living species.


Featured image: Samsung’s Virtual Reality MWC 2016 Press Conference, by Maurizio Pesce from Milan, Italia [CC BY 2.0] 通过维基共享资源。

由Dr Helene Lavoix (MSc PhD Lond)发布

Helene Lavoix博士伦敦大学博士(国际关系) ,是Red Team Analysis Society的总裁/CEO。她专门研究国际关系、国家和国际安全问题的战略预见和早期预警。她目前的工作重点是乌克兰战争、国际秩序和中国的崛起、行星越轨行为和国际关系、战略预见和预警方法、激进化以及新技术和安全。

发表评论

您的电子邮箱地址不会被公开。 必填项已用*标注

本网站使用Akismet来减少垃圾邮件。了解您的评论数据如何被处理