AI-rewritten: This is a summary of an article from New Scientist, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

---

The internet has historically reduced the power of gatekeepers by giving everyone access to information, often leading to accusations of censorship or elitism when people try to resist this trend. However, the rise of artificial intelligence is changing this dynamic. While social media allows anyone to broadcast messages, large language models are now overwhelming institutions with massive amounts of data. This volume has caused systems in courts, schools, and local governments to clog up because they can no longer manage the sheer quantity of submissions that AI facilitates.

This same issue is affecting science as well. Mathematicians are concerned about an AI onslaught, noting that the preprint server arXiv has seen its submissions double in just two years. Although these preprints are not formally peer-reviewed, arXiv’s small team struggles to maintain minimum standards due to the high volume of work. In response, the organization recently announced it would limit submissions to two per month. The article notes that this is still a remarkably high number, as few scientists produce twenty-four papers a year regardless of quality.

To survive this era, institutions must implement some form of gatekeeping unless they wish to let AI handle both inputs and outputs entirely. There are significant reasons not to fully automate these processes. Tools used to verify mathematical conclusions are already at risk from AI attacks, meaning human oversight remains the only way to ensure truth. This need for human control is even more critical in areas that are harder to verify than mathematics. Consequently, the article concludes that supporting gatekeepers is necessary, acknowledging that they will have a great deal of work ahead.