The executive behind the great productivity AI The company quickly claimed that their products would replace a large number of workers. Human CEO Dario Amodei made headlines in May Half of entry-level white-collar jobs In the next few years. Mark Zuckerberg, CEO of the Fungal Say in April He hopes AI will write half of the company code next year. Americans believe this – Recently Pew Investigation It was found that 64% of Americans expect less work due to AI.
In this environment, it is easy to see research on killable work and start panic. When Microsoft researchers launch Report In July, the job list list, which has the least and least overlap with tasks that AI can accomplish, stimulates consternation Among those who work at the top of the chart. But digging deeper digs don’t require translators, historians and others to worry about whether AI will replace them – unless decided by human employers attracted by the hype of AI.
“I think it’s useful to focus on work rather than work,” Darrell M. West, a senior fellow at the Brookings Institutional Technology Innovation Center, told CNET. “It’s possible that there’s not as much of the whole job being eliminated. There’s definitely a lot of tasks going to be eliminated.”
Microsoft’s research will overlap with career rankings for AI, even if that’s not the key point to making headlines.
“It is easy to conclude that careers associated with activities will be automated to experience job or wage losses… This will be a mistake because our data does not include downstream business impacts of new technologies, which is difficult to predict and often counterintuitive,” the authors wrote.
Even well-known figures in generative AI, if pushed, acknowledge this uncertainty. Openai CEO Sam Altman, talks with Theo Von Recent podcast looksaid not long ago, it is hard to imagine that people could have jobs as CEOs of AI companies and PodCaster. “I think it’s hard to accurately predict what something is going on or accurately predict what future work is going on,” Altman said.
It is dangerous to think that AI can do things that really cannot be done. Here are two jobs cited by the Microsoft report because it overlaps the most with the tasks AI can do: translators and historians.
Translation is more than just finding the right word
Spanish is the official language in more than 20 countries around the world. This means there are over 20 different variations in the language – even more when you consider local and regional versions. Andy Benzo understands the importance of these differences. “I said Argentinian,” she told me. “I don’t speak ‘Spanish’. There is no “Spanish.”
As a legal translator and elected president American Translation AssociationBenzo must not only understand the basic words of Spain, but also the culture and legal culture behind it. Attorney Benzo not only changes words and sentences from one language to another—the meaning must be correct. These translations can have serious implications for the person or entity involved in the legal process and, crucially, the correct meaning is correct.
Translation is more than just transcribing and converting documents. Medical translators help people communicate with doctors and nurses to ensure they are properly cared for. These are real deaths. Financial transactions that move from one language to another need to be clear, otherwise someone’s money or livelihoods may be at risk.
Professional translators are often experts in languages, and specialists in their specific fields, Benzo said. “You pay for what we know,” she said. “We said we were going to do it accurately.”
Translation tools powered by generative AI are becoming increasingly skilled in helping someone communicate in a language they don’t understand. You can lift your phone and let you explain between you and someone who doesn’t understand any language you understand, as Apple demonstrates it with Apple iOS 26 And Google’s Gemini. But professional translators and interpreters focus on making things perfectly right. You don’t want the translation to guess well – when your money or life is online, that’s really all you get from AI. You need a translation to understand the nuances between languages. And, you want one to be responsible if it is wrong.
“If AI makes a mistake, who will be responsible for it?” Benzo said.
Language is not static either. Although the AI industry is growing rapidly, language changes faster. Every day, someone finds a new way to express an idea somewhere. For example, the Cambridge Dictionary is just Added words Just like “Skibidi” and “Broligarchy”, AI with outdated training datasets may not be understood. But those who are properly trained can keep up with those subtle adaptations.
“Language has been developing,” Benzo said. “Language belongs to the people. No one is the boss of language. The only person who can perceive the nuances of language is human.”
History is more than just telling the same ancient story
Sarah Weicksel is a historian whose research is difficult to find in books because it has nothing to do with words. She learns about clothing, not the kind of clothing you get online advertising. Her job (including Upcoming book) Study physical clothing from the American Civil War era and how they reflected the economic and political environment at that time. Studying 160-year-old clothes requires digging out some of the archives that are rarely displayed in museums. (When clothes come out for exhibitions, they often appear in a short time because they quickly decay.) It also involves looking at diaries and other historical documents and looking for references to references, rather than important, world-changing events, but to pants and shirts.
“My research process is very confusing,” Weicksel, now executive director of the American Historical Association, told me.
But can’t AI models look at museum costume collections or read all these Civil War diaries? Not exactly. The job of a historian is not to find obvious ones, but to find basic stories that are not necessarily on the surface. Weicksel looked at the clothes to consider how the tailoring of the jacket helped someone stand up more upright, or the texture of a different fabric. “AI can’t touch and feel something for me,” she said.
More importantly, Weicksel approached her research by trying to answer and understand specific questions that might not have been asked before. This is the core of the work of historians: exercising judgment and creativity to discover new interpretations of the past.
Read more: chatgpt cannot fix all issues. This is 11 times and you will regret using it
Weicksel said that research like Microsoft Research, which examines how AI handles a single task done by professional historians, does not cover the entire situation. Yes, historians maintain and edit files and provide information to people, but these tasks are “not the core of historians.”
“We are not only a set of tasks that we accomplish and produce discrete things,” Wexell said. “We are very focused on the ability to synthesize and contextualize and bring judgment, but also the creativity of these questions we ask.”
Large language models can provide you with great reports on historical events. ask chatgpt You may get a good summary of the report on the 1618 Prague revocation – unless Hallucination And confusing it with other times, people were thrown out of windows by Brague in 1419 and 1483. But hopefully, AI can do the work of historians because it can summarize or analyze historical events that will make things back. AI can sum up historical events because it stands on the shoulders of historians who dig out the facts and write down what happened.
Research on history helps our understanding of the past, but machines trained to follow past trends may not find any accidents or help us avoid repeating the same mistakes.
“Great historical works are neither predictable nor obvious,” Wexell said. “That’s what makes them transformative. This cannot be replaced by trained technologies to replicate existing models.”
As long as there are tools, automation will affect work. Artificial intelligence can improve the types of automation in certain fields by improving robotics technology, such as this robot in Beijing that can do the work of human sales personnel (or vending machines).
What kind of work can AI do?
There is a difference between tasks that AI can perform and tasks that AI can help. Large language models have proven to be good at writing software code, leading to the spread of “Vibe encoding”, where human characters come up with ideas more and troubleshoot products when AI works most of the time. AI is also increasingly used in roles such as customer service, where chatbots or something like that can handle more direct requests, leaving only more complex chatbots to humans.
recent Paper Researchers at Stanford University found that in some automation-sensitive industries, employment declines in young early-career workers also found that these declines were primarily roles that could automate tasks.
“Although we have found that employment for young workers in occupations where AI mainly automates jobs have decreased, we have found that in the most enhanced occupations in AI use, employment has grown,” they wrote, “it can enable humans to change them faster or more efficiently without changing them.” “These findings are consistent with the automatic use of AI to replace labor, while enhanced use is not the same.”
West told me that work displacement has occurred where routine tasks can be automated. There are many layoffs happening among software developers because AI can be done fairly reliably. “Most jobs are affected by AI, but not all jobs are replaced,” West said. “People should only look at specific tasks related to any job and then consider the possibility of automation.”
The impact of AI on work will be determined by people, not potentially
Most importantly, even in the years from now, no one knows what impact AI will have on the economy. Chatgpt only became a household name in 2022. The capabilities of these tools and our understanding of what they can and cannot do are constantly changing.
However, due to what it can do, the impact of this technology on the job may not necessarily occur. This will happen because of what business leaders and executives think can be done. Currently, many executives seem to be more worried about missing out on opportunities to lay off employees and save money using AI than worrying that AI will not be able to do the job. For example, Klarna said in 2024 that its AI assistant can do the job of 700 customer service agents Change your mind Earlier this year, more people were hired after not getting the expected results.
There are already doubts about the impact of company-oriented AI plans. July study MIT researchers found that 95% of AI pilots in businesses did not receive a return on investment — mainly because AI tools do not learn, grow, and develop like human employees.
“Corporate leaders may end up laying off too many employees due to optimism about AI, and they may eventually find that there is an important element,” West said. “Human judgment will be crucial.”
The human factor (whether it is judgment, creativity, or culture) may prove to be what makes AI tools unable to do their jobs, even if it seems that it may be able to do all tasks on paper.
