麻豆传媒

Sarah O鈥機onnor, Allen Lane (UK); (US, out 11 August)

If you are a fan of translated films, you may have noticed the subtitles on streaming platforms have changed in recent years. They aren鈥檛 wrong exactly, but they can come across as a bit, well, flat.

鈥淵ou get the meaning, but the language? It鈥檚 not as rich,鈥� Petr 膶ermoch, a translator in the Czech Republic, tells Sarah O鈥機onnor in We Are Not Machines, which explores how artificial intelligence is changing the way we work.

That lack of richness is usually because the streaming platform has used AI to translate a script, then had a professional translator like 膶ermoch finesse it. Agencies expect translators to do this work more rapidly and have slashed their pay rates accordingly.

But this new type of job is both harder and less rewarding. The translator has to look at the original source and the machine text simultaneously, meaning more effort, not less. At the same time, the joy of the work has gone. 鈥淚t鈥檚 just a tedious job 鈥� boring and bland and lifeless,鈥� says 膶ermoch.

Yet, as O鈥機onnor shows in her excellent and wide-ranging book, this is the sort of future we are increasingly being told to accept. AI replacing human translators is not a new premise, but O鈥機onnor 鈥� who is a reporter at the Financial Times and has a weekly column on work 鈥� argues that the rather lifeless translation we are getting as a result is an example of how humans are adjusting themselves to fit into an AI-led world, rather than the other way around.

O'Connor writes of how she 'has the feeling that we have somehow lost faith in ourselves'

Her central premise isn鈥檛 that AI is coming for all of our jobs, which is a well-trodden debate, but that we are already contorting ourselves to fit AI into our lives. That can range from accepting a lesser standard of a product, such as translation, to attempting to match our capabilities to that of AI at work, then chastising ourselves when we inevitably come up short. 鈥淚 have the feeling that we have somehow lost faith in ourselves鈥�, writes O鈥機onnor.

The book shows how this type of contortion is happening across many industries, to often-maddening levels. We hear about bosses who cannot hire excellent interns because they didn鈥檛 perform well in an AI-generated test, or copywriters who are seeing their online articles being down-ranked by Google because its algorithm believes they are too good and so must have been written by AI. In this case, writers are having to put their articles through a 鈥渉umanizer鈥� tool 鈥� also run by AI 鈥� which changes the copy to make it appear more 鈥渉uman鈥� and inserts 鈥済rammatical mistakes, punctuation mistakes and errors of meaning鈥�. Is this really the future we want, O鈥機onnor asks, one that is 鈥渄amaging and deforming human writing in the process鈥�?

It would have been interesting to hear some of the counter-arguments from tech executives to such points, but O鈥機onnor purposely stays away from Silicon Valley, instead focusing on humans 鈥渁t the factory gates鈥�. We hear from people around the world, from miners in Sweden to controllers of autonomous trucks in the US.

Of course, as O鈥機onnor points out, employers wanting staff to be more like machines isn鈥檛 new. Frederick Winslow Taylor, one of the world鈥檚 first management consultants, unveiled a system in 1911 that removed factory workers鈥� autonomy, telling them each day not only what task was to be done, but how it was to be done and the exact time allowed for doing it.

But AI has turbocharged this process. The book gives the example of Maria, a remote worker in Costa Rica, who is sent videos of Amazon workers putting items on shelves and has to identify anything the warehouse cameras have failed to track and itemise. Maria is expected to watch around 1200 ten-second videos during a 9-hour shift, and get an accuracy level of 99.9 per cent. At the end of the week, her bosses expect her to have made no more than three mistakes across 8000 videos. 鈥淭hey ask you to have the same accuracy as the machine and it鈥檚 not possible,鈥� says Maria.

It鈥檚 not all bad, however. O鈥機onnor visits a mine in Sweden where autonomous trucks mean miners have more productive and safer jobs. But the difference here is that the miners also had a powerful union that was able to decide how AI was used in their workplace. Staff didn鈥檛 like the idea of a real-time positioning system tracking their movements, so it is now anonymised. 鈥淎cquiescence or resistance are not the only two options available,鈥� states O鈥機onnor.

It's not all bad 鈥� she visits a Swedish mine where autonomous trucks mean miners have safer jobs

That is easier said than done, though. The subtitle to this book is: 鈥淭he fight for the future of work鈥�, but at times We Are Not Machines felt like a series of fascinating feature articles, rather than a guide to possible solutions to these issues. Perhaps the answer is in simply championing our inherent human value, says O鈥機onnor. 鈥淭he real danger is not that we successfully make machines in our image, but that we silently remake ourselves in theirs,鈥� she writes.

O鈥機onnor does offer a few practical suggestions for the fight: for example, that workers should gain a foothold over how AI is used in their industry as soon as possible. She points to Hollywood writers who negotiated how AI could be applied to their jobs, while they still had the bargaining power to do so 鈥� unlike translators, who were too late in fighting back. That process also involves working collectively, such as joining a union.

For a lighter look at AI, there is Joanna Stern鈥檚 rather different book, . Stern, a former tech columnist at the Wall Street Journal, chronicles a year of using AI and robots to help her with everything from dental work to mammograms, cleaning to cooking. It鈥檚 an engaging introduction to AI, but it is a bit uneven. Stern also inserts a joke every couple of paragraphs, which can detract from the more serious subjects she addresses, such as the climate impact of AI.

It was telling, however, that even after a year of using AI for everything possible, Stern had much the same message as O鈥機onnor at the end of her book: that we should work with AI, not for it. 鈥淭he moment you let it do most of the thinking for you, the atrophy begins, and you lose control,鈥� Stern writes. As we increasingly work alongside AI, such a mantra will be important to remember.

Tom Knowles is a technology and business journalist based in London

听

Three more great books on artificial intelligence

by Madhumita Murgia

The Financial Times鈥榮 AI editor examines how AI is infiltrating policing, welfare, justice and health, to the point where lives are being altered 鈥� and often ruined 鈥� by systems hardly any of us understand.

听

by Sebastian Mallaby

A detailed biography of the AI firm DeepMind (bought by Google in 2014) and its co-founder Demis Hassabis, this shows how AI can revolutionise scientific fields such as chemistry and biology.

听

by Karen Hao

A gripping account of OpenAI鈥檚 move from an ideological non-profit to a firm 鈥渁ggressively commercialising products鈥� like ChatGPT. Hao argues OpenAI has sparked a race in AI that is heading in an alarming direction.

When you make a purchase via the links on this page, we receive a commission.