{
  "id": 20526687,
  "title": "AI will revolutionise science without requiring AI geniuses",
  "url": "https://urgent.news/2026/11/18/ai-will-revolutionise-science-without-requiring-ai-geniuses",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-18T01:00:36.000Z",
  "source": {
    "name": "e27",
    "slug": "e27",
    "url": "https://e27.co/ai-will-revolutionise-science-without-requiring-ai-geniuses-20261116/"
  },
  "original_language": "en",
  "account": "For much of human history, scientific advancements have been restrained by the limited human attention available. In the past, researchers would spend hours in libraries, scanning titles and abstracts before deciding which papers to read in-depth. The rise of electronic databases and the internet revolutionized this process, making literature searchable through titles, keywords, and abstracts. However, the challenge remained that a human still had to read, evaluate, and determine the relevance of each paper to the specific problem at hand.\n\nAI has transformed this constraint by enabling the creation of substantial scientific attention within software. Large amounts of scientific data can now be discovered, assessed, interconnected across various fields, hypothesis-generating, and even experiment running. This development holds true even if AI never replicates the capabilities of an Einstein. Science has always been prone to information loss due to the overwhelming volume of knowledge. Researchers have had to aggressively filter and discard vast amounts of data, often missing valuable findings. Electronic search reduced the cost of finding relevant papers, but it did not eliminate the bottleneck of human attention. Search engines could provide thousands of results, but someone still had to select which few were worth reading.\n\nOpenScholar, for instance, scans a collection of 45 million open-access scientific papers, while systems like Kosmos combine literature search, data analysis, and hypothesis generation. In a single research run, these systems can process around 1,500 papers. Although imperfect, these systems demonstrate the potential for expanding the scale of scientific attention without a proportional increase in human researchers. Consequently, more of the existing scientific record can become usable.\n\nThe most significant gains may arise from the connections that AI can make. Human researchers specialize due to the limitations of attention, but AI does not face the same constraints. It can compare findings across thousands of papers and disciplines, even uncovering relationships that were previously unconnected. This is crucial because there may be more useful knowledge within the scientific record than currently recognized. A 2019 materials-science study revealed that machine learning applied to earlier research could identify promising materials before later scientists explored their applications. This showcases the potential for AI to uncover hidden relationships within the scientific record.\n\nScientific knowledge can also become less static with the aid of AI. Continuous assessment of old research becomes possible, with traditional peer review concentrating attention around the moment a paper is published. Afterward, new studies may support, weaken, or contradict its findings. However, there is no universal system for updating the status of every claim. AI can facilitate a different model where old findings are repeatedly checked against new evidence. Successful replications can strengthen findings, while failed replications or newly discovered mistakes can weaken them, resulting in a more dynamic scientific record. Existing tools like Scite, which analyzes how later papers cite earlier work, indicate that some aspects of checking and reassessing published research can be automated.\n\nMoreover, AI can produce new evidence. Research can be conducted entirely in software, expanding mathematical proofs, software testing, simulations, and analysis of existing datasets. While some fields require physical experiments, AI systems connected to laboratory equipment can choose experiments, run reactions, measure results, and decide on the next set of conditions. This trend is already emerging in chemistry, protein engineering, and mechanical engineering, where AI systems interact with robotic equipment to design, print, test, and select the next experiment. These advancements have the potential to significantly increase the quantity of usable scientific knowledge while simultaneously enhancing its quality through continuous monitoring and reassessment.",
  "summary": "For most of human history, scientific progress has been limited by the amount of human attention available. A few decades ago, a scientist searching the literature could spend hours in a library scanning journal titles before deciding which papers justified the time needed to read them. Electronic databases and the internet transformed that process by […] The post AI will revolutionise science…",
  "key_points": [
    "AI software can process 1,500 papers in a single research run",
    "AI uncovers hidden relationships across disciplines and centuries",
    "AI can continuously reassess and update scientific knowledge"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}