SUBSCRIBE NOW!

Vol. 42 No. 5 — Sep/Oct 2025

FEATURE

PRODUCTION

As you step out onto the future campus, you notice a faint hum. In the distance, you spy two small-scale nuclear generators providing campuswide power. Such reliable energy was the only option, as the campus became an intense user of local AIs and thus has its own vast data center. The campus’ publishing house doors swing wide, and you go inside and look around. Don’t worry, no one seems to notice you.

Many of the touted benefits of early AI are bearing fruit, and a half-dozen OA journals are in various states of efficient preparation. At a holographic-images workstation, you see interactive graphics, timelines, maps, and complex 3D-rotation animations being instantly made by AI, then VR-checked and embedded by human interns. You approvingly note the use of mature, ethical AI image generators, trained only on public domain images. AIs are also inserting hyperlinks and breakout boxes in articles, linked to a rolling super-glossary of field terms.

SUBMISSION

You learn that writers submitting a proposal still initially encounter their human editor, although some choose to join one of the AI-guided co-creation outreach user groups. The editor’s AI assistants can now emulate essential editorial considerations: timeliness, novelty, usefulness, likelihood of citations, the journal’s tone and standing, etc. Manuscript submission and selection are far beyond the keyword and phrase detection of yesteryear—a crude method that missed much. Once accepted, a schedule-chaser AI liaises with the contributor’s scheduling AI, while a method-chaser AI ensures that methodology setup is accurate.

PEER REVIEW

Peer reviewers remain human, but they are AI-augmented in a way that is transparent for each author. Trust is further created by automated scrutiny of the draft’s statistics, mathematics, and tables. Plagiarism detection is now so sophisticated and trusted that it runs in the background, and no one gives it a second thought. Also building trust are the new AI-assisted methods of peer-reviewer discovery and assignment. You hear one of the office team members express pride that the reviewer discovery process has such a wide and equitable global reach, alongside robust credential-checking and measures to draw from a variety of groups. Journal contributors are now happy to find themselves assigned AI-assisted peer reviewers located in far-flung places, and they have learned that they can rely on the superb voice-based auto-translation and auto-summarizing tools and constantly AI-revised, multilingual glossaries.

More peer reviewers were a necessity, since speedier production is the norm. Faster publishing encourages more submissions and more journals per year, which makes more space available to younger and/or marginalized voices. AI helps find reviewers, trains them, and keeps them in rotation. Expanded journals also means supplemental extras such as a quarterly AI-generated summary of new preprints, theses, interviews, open datasets, and tools. The journals’ online dynamic glossary map is another useful addition that draws attention online because it is constantly updated and refined.

EXPECTING THE UNEXPECTED

You overhear interns saying they love Miss Serendipity. She is the AI dealing with the serendipity problem. Early AIs always found the right thing and the right answer, so everyone had basically the same answer. To forestall this herd mentality, a new class of serendipity engines was invented. Less loved are the sliders, which can be employed to subtly rewrite an article according to the reader’s disposition. Thankfully, these are now multidimensional sliders, rather than a simple right-left switch. Sliders emerged partly as a reaction to the bland, conformist writing and generic information of the early years of mass AI.

FINANCES

You glance admiringly at the publisher’s bookshelves. Income is generated by AIs, which scour the back catalog for articles that could make a themed book by pairing these with OA chapters and articles from elsewhere. The resulting assemblage is augmented with new auto-transcriptions of events such as conference roundtables and interviews—and sometimes, a classic 20th-century reprint, newly annotated with footnotes. Yes, footnotes are fashionable in this era because early AI hallucinations meant readers demanded a way to check every reference at a glance. Once approved, the print-on-demand book is assembled, styled, and then auto-published by AI. You also notice that the publisher makes extra income by selling the rights to have an AI remotely access its wealth of back issues. OA means open, but there are now limits on AI access to journal articles and data that are enforced by bot-blockers.

Another AI handles institutional trust and ethics, discreetly examining ethics certificates and auditing practices at cloud services, local desktop AIs, campuswide AI assistants, and more. Findings are checked annually by a board of human trustees. This board also signs off on the AI-assisted budget and accounts, while publicizing an AI’s annual tabulation of the regional economic and wider civilizational value of the publisher.

LEADERSHIP

Looking across to a breakout space, you see human managers chatting with an AI. You hear them say that cash flow is stable, and they want to expand into emerging fields. They know that, ideally, they should now start using AIs to move into a position in which they are leading—rather than following—the research agenda. They are testing the new gap-seeker AI, which seeks out what is not yet known and what might yet be done in any given field of study. You smile to see that time travel is on the agenda.