RexLegendi reviewed In de schaduw van ai by Madhumita Murgia
Results of data colonialism
4 stars
Code Dependent (2024) by Madhumita Murgia was the perfect book to read consecutively with Weapons of Math Destruction (2016) by Cathy O’Neil and Artificial Unintelligence (2018) by Meredith Broussard. Whereas the latter books taught me that artificial intelligence (AI) is not actually ‘intelligent’ and that the use of proxies is harmful, Murgia explores its social consequences. In ten chapters, she explains how the use of algorithms leads to a loss of agency – the capacity to act independently and make choices – and contributes to social injustice.
Exploitation data workers In the first chapter, the author examines the activities of big tech companies in developing countries. Their promise of self-learning machines is a commercial one; in reality, feeding AI with the right data is human work, often outsourced to low-wage countries. Unskilled workers are tasked with data annotation (such as labelling photos) and content moderation (e.g. removing pornographic or violent …
Code Dependent (2024) by Madhumita Murgia was the perfect book to read consecutively with Weapons of Math Destruction (2016) by Cathy O’Neil and Artificial Unintelligence (2018) by Meredith Broussard. Whereas the latter books taught me that artificial intelligence (AI) is not actually ‘intelligent’ and that the use of proxies is harmful, Murgia explores its social consequences. In ten chapters, she explains how the use of algorithms leads to a loss of agency – the capacity to act independently and make choices – and contributes to social injustice.
Exploitation data workers In the first chapter, the author examines the activities of big tech companies in developing countries. Their promise of self-learning machines is a commercial one; in reality, feeding AI with the right data is human work, often outsourced to low-wage countries. Unskilled workers are tasked with data annotation (such as labelling photos) and content moderation (e.g. removing pornographic or violent material). While working conditions are better than in sweatshops, Murgia argues that they are still unfair: complaints can result in sanctions, such as (temporary) removal from a platform, cutting workers off from their income.
Data colonialism Following Nick Couldry and Ulises Mejias (The Costs of Connection), the author uses the term ‘data colonialism’ to describe the exploitation and control of personal data by powerful entities like big tech companies, who use these data for their economic or political interests. There are some harrowing examples in developing countries, where companies, for instance, promise free healthcare to local communities. While these companies profit from vast amounts of data, the communicaties often don’t benefit from algorithms that were not designed for them. Other examples are closer to home, from deepfakes used to hyper-sexualise and intimidate women to shady algorithms mistreating Deliveroo or Uber couriers. Surveillance raises questions about discrimination and stigmas; even monitoring for ‘social’ purposes is dangerous, as the technology is usually not empathic but punitive. (On this point, I found O’Neil more convincing.)
Cast Away In the final chapter, Murgia zooms in on ChatGPT and language proficiency. By training AI to use ‘human’ language, the industry has managed to simulate intentions and emotions, further contributing to the smokescreen that obscures the black box. From this book, I learned to compare ChatGPT to the volleyball Wilson from the movie Cast Away: in the end, humans project their own emotions onto objects.
I’d highly recommend this book to anyone interested in learning more about AI and ethics. Find out for yourself why Franz Kafka’s The Trial is more relevant than ever.
