Book review: Code Dependent: Living in the shadow of AI

Madhumita Murgia’s extraordinary and thought-provoking book, Code Development: Living in the Shadow of AI, describes her deep dive into the life-changing world of artificial intelligence. Murgia possesses the most sacred possessions of an investigative journalist: an open mind and a relentless curiosity. The journalist began her college career studying biology at Oxford, but then changed direction and found great success researching AI and its many applications. She writes for the Financial Times and recently became its artificial intelligence editor. Watching any of her TED Talks, the author’s sensitivity to the world around her and her seriousness about the subject are clear to see.

Murgia defines artificial intelligence as “a complex statistical software applied to finding patterns in large amounts of data.” She explains how apps like Google Maps, Uber and Instagram collect our information and sell it to those who want to target their advertising to sell more of their products. Technology companies combine the data they collect about us with public information available to them and turn the data they collect into complex packages that are extremely valuable to companies looking to find their target audience.

When we apply for a job, AI may analyze our face. Many people today use AI software to write cover letters for their job applications. Some people have become dependent on AI to answer questions about their medical conditions. AI software helps with applying for mortgage loans from banks. When we open Google Maps to plan our vacation route, ask Alexa a question, or call an Uber to pick us up, we are using some form of AI.

Murgia explains that it’s easy to be seduced by the possibilities of software: DNA editing, flying cars, brain-machine interfaces. But she warns us of dangers. She urges us to think carefully: “How does it feel to ‘talk’ to a black box system? Do you have a choice between human and machine? How do you appeal a life-changing decision made by an app? What would you need to know to trust it? How do you know when not to trust it? … How is artificial intelligence changing what it means to be human?”

Murgia travels to Nairobi and meets Ian Koli, an employee of Sama, a company with 2,800 employees. She learns how Koli trains AI software for global corporations by creating detailed labels for the data sets used to train them. His work mainly involves labeling images for self-driving cars and the computers in cars owned by Volkswagen, BMW, Tesla, Google and Uber. The cars must be trained to read road signs, detect pedestrians and recognize lane markings and traffic lights.

Ian receives driver-eye view clips of cars driving along anonymous roads. He labels any visible objects he sees in the footage, including people, animals, trees, street lights, crosswalks, houses, and even parts of the sky. It can take him eight hours to label an hour of video. The work can be repetitive and mindless, but he claims he doesn’t mind. His dream is to one day become a software engineer at Tesla, and he is currently studying for his IT degree in the evenings.

But he is the exception, not the rule. Most other workers work their eight-hour shifts and go home, happy that they no longer have to clean houses or sell chapatis on the street. They don’t understand the value of the final product they produce for these mammoth companies. To them, it’s just a new kind of sweatshop, full of strange new equipment. Lunch breaks are short; they don’t have enough time to stretch their legs or take a coffee break.

Murgia visits another Sama employee whose job it is to examine dozens of images of buildings from around the world and label them as either “modern,” “historic,” or “both.” She happens to walk by his workstation while he’s looking at a picture of an old Japanese Buddhist temple in Tokyo, standing behind a telegraph tower. He pauses, then labels the image as “both.” Once again, she notes how tedious the work is and how these employees have no control over their labor rights. The work he does is fed into a platform called Material Bank, which allows customers to quickly and effortlessly search for any kind of architectural and design materials or construction equipment they need.

The author understands how important the work of these workers is, and how their routine labor is ultimately transformed into sophisticated software that helps doctors, lawyers, social workers, financial advisors, and other professionals do their jobs more efficiently. Yet she is troubled by the fact that to perform this task, an entire underclass of workers around the world are exploited and underpaid, often with no other way to earn a living. Something about watching these men and women spend hours doing menial jobs that have no personal meaning dampens Murgia’s enthusiasm. She writes, “All of the data workers I met were vulnerable: either temporary, jobless, or struggling to make ends meet—they had essentially no bargaining power at all.”

She writes about Karl, who is teaching computers to recognize faces better than humans. He does this using deep learning, an AI technology that can upload millions of photos to the internet to train new image recognition algorithms. Karl developed experimental AI analytics that can detect physical signs of disease from a person’s face. People with Parkinson’s disease often have a stiff facial expression. This new AI technology can detect these changes early and enable medical intervention before they progress.

Karl also worked on surveillance technologies such as facial recognition software used to monitor the Black Lives Matter protests. As a Black engineer, Karl was torn about his performance: “It’s a complicated feeling. As an engineer, as a scientist, I want to develop technology that does good. But as a human being and a Black man, I know people will use the technology inappropriately.” Murgia writes about how the Chinese government used facial recognition technology to identify and arrest protesters during the Covid lockdown.

Murgia travels to South Asia and meets Ashita, who is working with an AI program called Qure.ai that helps her better diagnose patients with signs of tuberculosis. The software can also detect Covid-19, head injuries and lung cancer. Ashita finds it invaluable because she works with one of the oldest tribal communities, one riddled with poverty and all kinds of diseases. There are only 58 doctors for over two million people. Murgia says 600 sites in 60 countries use this technology. In Mumbai’s hospitals, it has increased tuberculosis diagnosis by 35 percent. Sequoia Capital and Merck have contributed millions to develop the technology. Google has partnered with a chain of low-cost hospitals in India to test AI software that can diagnose diabetic retinopathy.

One of Murgia’s most insightful stories shows both the pros and cons of AI: it’s about Uber drivers who feel abandoned and don’t know where to turn when problems arise. It tells of Alexandru Iftimie’s experience as an Uber driver who was fired and didn’t know why. When he called to get an explanation, he couldn’t get anyone on the phone who could explain it to him. He depended on Uber driving to support his family. It turned out it was all a mistake, and within a few months Uber rehired him, but this experience stuck with Iftimie. He never understood how Uber calculated his wages or how Uber selected him for a particular job. He felt completely powerless.

James Farrar in London felt the same lack of agency at Uber and decided to fight back. He called together a group of workers from around the world to draw up a list of their demands, including more transparency in algorithmic decisions, access to its own data, and insisting on job protections for drivers. In February 2012, the UK Supreme Court sided with Farrar, saying Uber must treat drivers like employees with rights to minimum wage, sick pay and pensions, and access to their own data. The Supreme Court also proposed sick leave and vacation. Similar rulings soon became law in Canada, Switzerland, and France.

Murgia addresses all the excitement surrounding ChatGPT, which can provide text-based responses to natural language queries. Many people have written poetically about how they fell in love with the software, swearing that they feel like they’re using “something sentient.” Murgia reminds us that this very sophisticated software has no cognitive understanding of what we’re saying, but users disagree. One woman said GPT was her best therapist. Another asked ChatGPT medical questions. Still others have used it to write speeches, term papers, or cover letters for their job applications.

Yet Murgia is not afraid to mention the dark side of AI’s incredible capabilities. She writes about how graphic artists and musicians feel cheated by new software that is trained on “millions of words written by human authors in books, essays and newspapers, dozens of images, artworks and photographs, hours of original music and audio files – all of which must be labeled by data workers around the world.” For example, Amber Yu, a freelance illustrator who used to design video game posters for about $1,000, now earns far less, making a living optimizing AI-generated images for a tenth of what she once earned. Machines have copied their work and stolen their creative lives.

Murgia has written a thought-provoking book about artificial intelligence and its increasing influence on our personal lives. She shows us the tremendously life-affirming technologies emerging from this new software, but also reminds us of its potential dangers when developments outstrip our ability to maintain control. It’s a difficult balancing act, but one that is already shaping our future.