Short on October 6, 2025
Newsom celebrates the last legal victory over Trump, chaos, as the Prime Minister of France, resigns, and much more
Julian Pintat, a freelance translator from English to Herman, watched his 15-year-old career gradually decaying. Specializing in areas with high rates, such as medical technologies and pharmaceuticals, its experience was resold as a service in cleaning artificial intelligence. In a recent work, the translation of the operating manual for the oil installation, the AI is incorrectly broadcast “scale” – mineral accumulation – both the musical scale and the weight measurement device. According to him, the correction of such basic disadvantages, which currently make up 95% of his work, often takes more time than transfer from scratch – disappointing reality, which has charged his income and put life plans, including marriage and the beginning of the family on an indefinite order. With Google Translate, and then Deep Having entered the stage a few a few years before ChatGPT – professional translators felt the consequences of artificial intelligence longer than the majority. “I am a canary in a coal mine,” says the Pintat.
AI changes the face of work, and the pantat is one of many who look at a completely different future. While there were many talk about the technology replacing the workers of the white collar and some general directors such as Ford Jim Farly and Andy Jasi from Amazon, predicting many Corporate jobs will be destroyed In the coming years, the first wave of adoption of AI is already revising the workers in a new role and changing the contours of their work.
For some companies, AI increases efficiency. This allowed the London Law Firm A& o Shearman to effectively multiply its labor, allowing it to use projects that she would refuse, says the partner and global head of the AI Group, David Wailing. To help a large American bank to comply with European legislation, the company built a tool that scanned 20 -year licensing agreements and determined that it is necessary to amend. “Two years ago, we would have 20 lawyers in the room, maybe some parallelgles,” says Uakeling. But the tool picked up 2400 regulatory requirements to 900, halving the cost of the project, even when the time for the construction of the tool, he says. However, he freezes his optimism. He says that the basic assistant in artificial AI will probably not add great value, noting that real results often require individual or specialized solutions. “A lot of elbow lubrication is required,” says Uakeling.
Meanwhile, AIG General Director Peter Zaffino He told the time In June, the insurance company uses AI to work on underwriting faster. This is training in the system to become a “younger underwriter”, which can make most of the underwriting, allowing “more experienced practicing” everything else. “Some cultural changes are an increase, retraining in the new world, which allows them to be more productive than we were in the past,” he said.
In the MIT report, published in August, he came to the conclusion that 95% of artificial intelligence pilots could not provide for return on investments. Even the AI coding assistants, who were detained as a winning version of the AI, were doubted by a small preliminary one. July research Published by Berkeley, based in Berkeley, a research group METR. The sample of 16 experienced developers was 19% slower when using AI, despite the fact that on average she made them 20% faster. He will take advantage of excitement – and, possibly, with fear of skipping – businesses participate in the use of AI, even if they are non -optimal methods. This creates a gap between the fact that the market considers AI, and actual performance, compressing both enterprises and – in the case of translators – employees covered by the promise of superhuman efficiency and the reality of the often erroneous conclusions of the machine.
The rapid promotion of AI on the tasks of white collars can soon erase this gap. Productivity on tests developed by experienced specialists throughout banking, law and consulting almost doubled for a little over a year, according to Pre -study Published at the end of September with a company, Mercor. Paper appeared on the spots of another reportThe created Openai, which sought to measure the ability of AI to perform real tasks, comparing the machine from the performance of a person in blind tests. It was found that the best models are favorably compared with the work written by a person, almost half of the time. Nevertheless, both messages note that such tests measure performance by well-weakened tasks-quenches, which often lack a dirty real world. This means that at the moment, AI models can make a poor replacement of human workers, and the implementation remains key.
Read more: AI learns to do the work of doctors, lawyers and consultants
“Generative AI makes a really fantastic demonstration,” says Kaitlin Elliott, the head of the Morgan Stanley Generate Ai solution in Stanley, but making it more useful than it looks. The bank built its own tool for transcription and summation of the meetings, which, according to Elliott, retains clockwork hours. He also created a search tool for AI, which facilitates the staff for the surface of information. “In the early days, we thought that we could simply give all our knowledge, and it could give very accurate answers,” she says, but in practice it required well -structured data and thorough testing.
The task of implementing AI is to create a demand for new types of experience. “There is still a need for technical skills,” says Elliott, who adds that, although artificial intelligence tools have automated the work usually performed by junior employees, they are currently counting on the younger generations in the organization. “All of them are adoptive parents. They know how to use it effectively, ”she says. This demand also creates opportunities for companies selling this know-how as a service, for example Libra beforeThe most famous of its data labeling business, but which has expanded before helping enterprises, including Morgan Stanley, with artificial intelligence applications. According to Felico Su, the AI AI engineering director, when accepting artificial intelligence, it is extremely important to work in the opposite direction from problems. The use of AI for the sake of itself can have unpleasant consequences. SU gives an example of one of the customers of Scale, who built four chats for slightly different tasks, forcing the staff to constantly copy and insert between them. SU adds that identification of solutions sometimes means the use of generative AI, but often this means using traditional machine learning or software development.
Improvements come to translation. DEEPL offers functions such as the creation of individual glossary, and this year has presented a tool that asks subsequent questions to clarify the ambiguity that can help in niche domains. Translators who have successfully used AI work faster, says DEPL JAROSLAW KUTYLOWSKI Director General, although the company offers only numbers, comparing its tools with AI competitors, and not without strangers. Improved productivity can allow translators to compensate for the lower word costs by translating higher volumes of text, although Kutilovsky notes that AI tools allow some enterprises to bring a translation inside, and not to outsourcing to professionals. “I think that we are simply aligning here on this civilizational ladder,” he says, admitting that this will bring changes to how people do their work. “This is a change that we will have to go through,” he adds.
In another place, companies seek to reduce the technical bar for enterprises to accept AI, creating ready -made enterprises. For example, most of the products of virtual assemblies now have a built -in final AI collector. A&B SHEARMAN arming its lawyers using a legal tool of AI HarveyWhich, according to Wekeling, is useful for general questions, although the company now uses its own artificial intelligence tool, ContractMatrix, for especially niche requests. And the new breed of enterprises focused on enterprises receives from internal documents, weak messages and emails to respond to requests.
The Canadian art company Cohere has released North, its own such tool last month. The co -founder of the company, Nick Frost, says that he no longer sweat meets at the last minute, because he uses a tool to prepare short persons based on their entire story with a company in seconds. The North is now engaged in 90% of its total ticket tickets, although human operators are still in the know. It is not just used inside, with RBC, the largest bank of Canada, which introduces the platform. (Salesforce, where the co -chair of Time and owner Mark Benioff is the general director, is an investor in Cohere.)
While the bosses of Openai, Anpropic and Google Deepmind are considered to be the so-called artificial intelligence or the AGI system that can automate most of the human work, can be only a few years later, the FROSST prospects are relatively conservative. He does not believe that we will reach Agi, using something resembling the current technology. Nevertheless, he says that even without Agi, the influence on labor will be destructive, comparing it with the industrial revolution. “When there were huge transitions on the labor market, much of what was decided was at the government level, the level of the trade union,” he says. “This is a problem outside of any person, and we must consider it as a team.”
