We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode Feeding the machine: the hidden human labour powering AI

Feeding the machine: the hidden human labour powering AI

2024/12/4
logo of podcast LSE: Public lectures and events

LSE: Public lectures and events

AI Deep Dive AI Insights AI Chapters Transcript
People
C
Callum Cant
J
James Muldoon
K
Kirsten Sehnbruch
Topics
James Muldoon: AI 的发展严重依赖于全球范围内的隐性人力劳动,这些劳动者,例如肯尼亚的社交媒体内容审核员 Mercy,常常面临着极差的工作条件和剥削。他们工作强度大,报酬低,缺乏休息和组织的机会,甚至在遭受极度精神创伤后也被要求继续工作。这揭示了 AI 技术背后隐藏的全球生产网络的残酷现实,以及其对全球不平等的加剧作用。 本书试图揭示 AI 技术光鲜外表下隐藏的全球生产网络,从数据中心工人到数据标注员,甚至包括为 AI 数据集提供素材的艺术家和作家,都参与其中。数据标注工作,例如为自动驾驶软件标注图像或文本,性质繁重且单调,工作者面临着极度剥削。以 Sama 公司为例,其在东非中心的工作条件与公司在旧金山的总部宣传存在巨大差异,这反映了 AI 行业中普遍存在的剥削问题。 本书的研究方法是通过对数字平台工人的工作条件进行评估,并对公司进行评分,以此来推动改善工作条件。这种方法已经成功地促使许多公司做出有利于工人的改变。 Callum Cant: AI 的发展并非完全的创新,而是长期工业发展进程中的最新阶段,其背后隐藏着对劳动力的剥削,这种剥削并非新鲜事物,而是延续了历史上对劳动力的剥削模式。AI 系统如同一个“提取机器”,将劳动力、能源、水等资源转化为利润。我们试图将当代的剥削故事置于更长远的历史背景下,分析其背后的社会动力和结构性原因。 AI 系统的构建和维护依赖于全球范围内的劳动力,这些劳动者常常面临着极度剥削和不公平的待遇。AI 技术的进步并未带来劳工权利的提升,反而加剧了全球不平等。AI 发展中的垄断趋势进一步加剧了这种不平等,因为技术发展方向的决定权掌握在少数资本手中,而这些资本并不关心技术的社会影响。 为了改变这种现状,我们需要采取多方面的措施,包括支持工人组织、开展民间社会运动、推动政府监管以及发展工人合作社等。欧盟的企业尽职调查指令为改善 AI 供应链中的工作条件提供了一定的希望,但要实现根本性的改变,需要更深层次的结构性变革。 Kirsten Sehnbruch: AI 数据标注员的工作条件恶劣,工作强度大,精神压力巨大,对身心健康造成严重损害。这些工作通常缺乏自主性、意义和方向,导致工人身心俱疲,甚至出现抑郁、焦虑和自杀倾向等问题。算法管理加剧了这种剥削,因为工人必须以极高的速度完成任务,并且面临着随时被解雇的风险。 这种工作条件的恶劣性不仅体现在发展中国家,在发达国家也普遍存在。例如,亚马逊仓库的工人也面临着类似的压力和剥削。这些负面影响不仅限于工人个人,还波及社会整体,因为社会需要承担因工人身心健康受损而产生的医疗等成本。 要解决这个问题,需要从多个方面入手,包括加强政府监管、推动工人组织、改善工作条件以及改变对劳动力市场的观念等。仅仅关注工资水平是不够的,还需要关注工作质量、工作安全以及工人权益等其他方面。我们需要对劳动力市场进行根本性的改革,以确保所有工人都能够享有体面和有尊严的工作。

Deep Dive

Key Insights

What is the 'machine' metaphor in the context of AI and labor exploitation?

The 'machine' metaphor refers to the 'extraction machine,' which conceptualizes AI as part of a long process of industrial development where resources like labor, energy, and water are converted into profit through technology. It situates contemporary exploitation within a historical context of colonialism and imperialism, highlighting the structural causes of inequality in AI production.

Why are data annotation jobs considered exploitative and mentally stressful?

Data annotation jobs are exploitative due to grueling conditions, low pay, lack of breaks, and no opportunities for workplace organization. They are mentally stressful because workers are pressured to meet strict time-per-task targets, often leading to repetitive strain injuries, mental health issues, and a lack of autonomy. Workers are also exposed to distressing content, such as toxic social media posts, without adequate support.

What role do global tech companies play in maintaining poor working conditions in AI supply chains?

Global tech companies exert significant power over AI supply chains by setting low wages and poor working conditions. They outsource labor to multiple centers worldwide, creating competition among them. Middle managers in these centers enforce strict labor discipline to secure contracts, often leading to unpaid overtime and exploitative practices. Tech companies like Facebook and Meta have the power to set minimum standards but often fail to do so.

How does the global labor market contribute to the exploitation of workers in AI production?

The global labor market fosters a race to the bottom, where workers in countries like Kenya, Uganda, and the Philippines compete for low-wage jobs. This competition is exacerbated by surplus labor and precarious employment, leading to hyper-exploitation. Workers are often trapped in cycles of poverty, with limited opportunities for upward mobility or skill development.

What are the challenges in measuring and addressing the quality of jobs in AI production?

Measuring job quality is challenging due to the lack of standardized data across countries and the fragmentation of employment relationships. Traditional labor surveys often focus on quantity rather than quality, missing critical issues like mental health, safety, and worker autonomy. Additionally, companies control information about their supply chains, making it difficult to assess working conditions and hold them accountable.

What structural changes are needed to address labor exploitation in AI production?

Structural changes include supporting transnational worker solidarity, pressuring tech companies through civil society campaigns, and advocating for government regulation, such as the EU's Corporate Due Diligence Directive. Additionally, there is a need for worker-owned cooperatives and a shift away from the concentration of power in monopolistic tech companies. Overhauling global capitalism is also suggested as a long-term solution.

How does algorithmic management contribute to the deterioration of work conditions?

Algorithmic management intensifies work by constantly monitoring workers, setting unrealistic productivity targets, and enforcing precarity through temporary contracts. Workers in industries like Amazon warehouses and gig economy platforms face high stress, physical strain, and mental health issues due to the relentless pace and lack of autonomy. This system circumvents traditional labor protections, leading to a deterioration of work conditions globally.

What historical parallels exist between AI labor exploitation and past industrial practices?

AI labor exploitation parallels historical practices like the super-exploitation of workers in dependent economies, such as Latin America during the 20th century, where raw materials were extracted for Western economies without local capital accumulation. Similarly, contemporary AI production relies on cheap labor in the Global South, reproducing cycles of dependency and exploitation. The dynamics of colonialism and imperialism continue to shape these labor relations.

Shownotes Transcript

Contributor(s): Dr Callum Cant, Dr James Muldoon, Professor Kirsten Sehnbruch | Conversations around AI tend to focus on the future dangers, but what about the damage AI is inflicting on people right now? AI promises to transform everything, from work to transport to war, and to solve our problems with total ease. But hidden beneath this smooth surface lies the grim reality of a precarious global workforce of millions that labour under often appalling conditions to make AI possible. Feeding the Machine presents an urgent investigation of the intricate network of organisations that maintain this exploitative system, revealing the untold truth of AI. Authors Callum Cant and James Muldoon will be joined by Kirsten Sehnbruch to discuss the impact of AI on global inequalities, and what we need to do, individually and collectively, to fight for a more just digital future.