Berlin Artificial intelligence will also radically change the everyday work of many people outside of California’s tech centers. This is the result of two studies that deal with the consequences of the AI revolution on the world of work.
The first study comes from the creators of ChatGPT themselves: Researchers from the start-up company OpenAI teamed up with scientists from the University of Pennsylvania to find out which jobs are most affected by ChatGPT. According to this, accountants are among the professional groups that are most affected by the possibilities of generative artificial intelligence. At least half of accounting tasks could be done much faster with this technology.
From mathematician to writer
According to the study, mathematicians, programmers, interpreters, writers and journalists should also be prepared for the fact that artificial intelligence can take over at least some of their previous tasks. Because although the AI systems currently often “hallucinate” incorrect facts in their answers, they are already delivering remarkable results in tasks such as translation, classification, creative writing and computer code generation.
The researchers from OpenAI and the University of Pennsylvania assume that most workplaces will be changed in some way by the AI language models. Around 80 percent of workers in the US are in jobs where at least one task can be completed more quickly by generative AI. But there are also professions in which AI will only play a minor role: These include cooks, car mechanics and jobs in oil and gas production, but also in forestry and agriculture.
“Generative AI” as a disruptor on the job market
In a study, a research department at the investment bank Goldman Sachs calculated what this development could mean for the labor market in concrete terms. If the so-called generative AI keeps the promised skills, this could lead to “significant disruptions in the labor market”. “Generative AI” is understood to mean computer programs that can create new ideas, content or solutions instead of just working through predefined rules or instructions.
Goldman Sachs estimates that about two-thirds of current jobs are exposed to some level of AI automation. Generative AI could replace up to a quarter of current work. “Extrapolating our estimates around the world, generative AI could expose the equivalent of 300 million full-time jobs to automation.”
Hinrich Schütze, Director of the Center for Information and Speech Processing at the Ludwig Maximilian University of Munich (LMU), sees the development of generative AI as a revolution that is technologically comparable to the Internet or smartphones. The AI systems are still a long way from a real understanding of the content of the topics: “The basic technology for the language pattern is simply to always predict the next word, very stupid, always the next word.”
The consequences are already enormous
Nevertheless, the consequences are already enormous: “There will be major changes in how we write, whenever we write texts, how we program.” This also has major consequences for day-to-day work. “A lot of jobs will disappear if it’s just a matter of writing summaries, gathering and condensing knowledge.”
However, the Munich AI expert warns against giving artificial intelligence too much radius of action when making decisions, for example in the judiciary, medicine, tax advice or asset management. The AI makes many statements with great persuasive power, although the facts are often incorrect, “People think that must be true if the model is so sure. But in reality, the model cannot assess its own safety. That’s one of the big problems we have.”
The Potsdam computer science professor Christoph Meinel sees another obstacle for the widespread breakthrough of AI in the working world, because the systems require enormous computing capacities and thus also require huge amounts of energy. “Many expectations of AI seem exaggerated to me and also unrealistic with regard to their energy consumption,” says the outgoing director of the Hasso Plattner Institute (HPI). The successful AI applications are based on so-called deep learning, i.e. training with huge amounts of data. “And they consume a lot of energy.” A broad introduction would therefore be fatal for the climate and the achievement of climate targets. “We first have to develop significantly more energy-efficient AI systems.”
Data protection as a challenge
Meinel sees a challenge not only in the high power requirement, but also in the area of data protection. “Anyone who tries out the latest artificial intelligence applications on the Internet should be careful about disclosing their own sensitive data,” advises Meinel. Despite the hype, all those responsible should be aware that the AI models of the providers can be trained and made smart with their own inquiries and data free of charge. For example, anyone who uploads internal financial data to certain platforms so that they can automatically create a presentation from it must know that this may also result in the disclosure of business secrets.