- New AI systems are shifting from the role of “research assistants” to automated scientists capable of self-planning, experimenting, and publishing complete research.
- The turning point emerged in late 2025 when frontier models achieved stable reasoning and planning capabilities, alongside “tool calling” to interact with external tools instead of just generating text.
- Sakana AI in Tokyo developed “The AI Scientist,” a system described as a platform for “fully automating the scientific discovery process.”
- The AI Scientist can independently scan research papers, generate hypotheses, write code, run experiments, analyze data, and produce research papers with almost no human intervention.
- A paper generated by The AI Scientist was accepted by the ICLR conference and published in the journal Nature in March 2026 after a peer-review process.
- The author of the article considers this a milestone where AI has passed a “light” version of the Turing Test in scientific research.
- The startup Analemma in Singapore demonstrated the Fully Automated Research System (Fars), which can generate 166 machine learning research papers in just about 417 hours.
- The operating cost of Fars is only about $1,100 to produce a volume of work equivalent to many weeks of human research assistant labor.
- Google Cloud AI Research also introduced PaperOrchestra, a system that transforms experiment logs and raw notes into complete papers with charts and verified citations.
- Researchers worry that AI will cause the already overloaded scientific publishing system to fall into crisis as the number of papers increases exponentially.
- The article warns that AI may cause scientific research to become “incremental” instead of creating radical breakthroughs like the eras of Galileo or Semmelweis.
- The impact is not limited to science. The AI podcast “The Epstein Files” topped the Apple Podcasts and Spotify charts in the UK with 700,000 downloads in its first week.
- The AI band “The Velvet Sundown” also reached over 1 million monthly listeners on Spotify before the platform had to implement mechanisms to protect real artists.
- US courts still refuse to grant copyrights for products created entirely by AI because the law requires the author to be human.
📌 AI has moved beyond the support stage and begun directly replacing many high-level intellectual jobs such as scientific research, content creation, and knowledge production. The fact that the AI Scientist was peer-reviewed and published in Nature is a massive milestone because, for the first time, an automated system completed almost the entire research process. However, the greatest risk is not just job loss but a fundamental shift in how society values knowledge, copyright, and professional expertise when machines can produce academic products at an industrial speed.
