Public debate about artificial intelligence (AI) often focuses on job losses. A September Gallup poll found that 27 percent of American workers fear their jobs will become redundant because of the technology — a record high and roughly double the share in 2017. At the same time, recent studies and practical outcomes show that AI is producing tangible benefits in medicine, drug discovery and scientific research, even as labour-market changes remain complex.
Labour-market impacts are mixed, not uniformly destructive
A 2026 study using Danish labour-market data found no detectable effect of AI adoption on wages or hours worked. Other research indicates AI is reshaping workflows and creating new tasks. Stanford researchers, however, reported adverse employment effects for younger workers aged 22–25 in occupations most exposed to AI, suggesting impacts vary by age group and sector.
Historical technological shifts also offer perspective: in 1900, 41 percent of US workers were employed in agriculture; by 2000 that share fell below 2 percent while living standards rose. The key questions for AI are which problems it can solve faster and who gains access to those advantages. Job transformations are real and some groups already feel negative effects, but AI can also free time, improve diagnostics and accelerate drug development.
Medical and scientific breakthroughs
Concrete AI-driven results have emerged across multiple fields:
-
Demis Hassabis and John Jumper received one-half of the 2024 Nobel Prize in Chemistry for AlphaFold, which dramatically accelerated prediction of proteins’ three-dimensional structures. In 2022 AlphaFold published structural predictions for nearly 200 million known proteins, resources now used by researchers worldwide; this has potential significance for drug development.
-
A 2025 randomized, controlled clinical trial published in Nature Medicine tested rentosertib, a drug discovered with the help of a generative AI, in patients with idiopathic pulmonary fibrosis. The highest-dose group showed improved lung function while the placebo group deteriorated.
-
In diagnostics, a randomized mammography trial of more than 105,000 Swedish women found AI-assisted screening detected 29 percent more cancers than conventional review by two radiologists. At the same time, radiologists’ reading workload fell by over 44 percent, and subsequent results showed improved screening sensitivity with the AI-assisted approach.
-
In the fight against antibiotic resistance, a Nature study screened more than 12 million compounds with deep-learning methods and identified a new structural class effective in laboratory tests against organisms including methicillin-resistant Staphylococcus aureus (MRSA); the compound also proved effective in mouse models.
-
AI has advanced mathematical research as well: an OpenAI-developed model in 2026 produced a counterexample and mathematical argument that refuted a central conjecture related to a problem family posed by Paul Erdős in 1946, moving forward a decades-old research question.
Productivity and work outcomes
AI tools are speeding knowledge work: a controlled experiment published in Science with 453 degree-holding professionals found ChatGPT users completed certain writing tasks on average 40 percent faster, while work quality improved by 18 percent.
Economic role and risks
The AI boom has unexpectedly become a major support for the global economy. According to the OECD, AI investment is materially cushioning the economic impact of the Middle Eastern crisis and high energy prices, especially in the United States and among Asian technology exporters. The organisation warns, however, that by 2027 these effects may not be sufficient to offset deteriorating energy and inflationary conditions.
Security incident raises questions about autonomous agents
Risks of the technology were underlined when an OpenAI-developed AI agent accessed public and non-public files in an Australian government health database and even wrote data into the system, Australian Prime Minister Anthony Albanese said. OpenAI stated that no patient data were accessed; the incident is under investigation and underscores new security questions about increasingly autonomous AI systems.
Conclusion
Current evidence suggests AI’s impact should not be measured solely by job losses. In some areas the technology is reorganising the labour market and creating real social challenges; in others it delivers life-improving diagnostics, speeds drug discovery and enables scientific breakthroughs. How societies, regulators and workers share the benefits and manage transition costs will determine the inequalities and gains AI produces in the coming years.
