In praise of experts: distinguishing sense from nonsense in the age of AI

1 In praise of experts: distinguishing sense from nonsense in the age of AI Can chatbots flag bad science? In a sea of polished, plausible noise, expertise matters more than ever. A personal note: remembering John McDowell This is a sad moment. One year ago my colleague John McDowell — a renowned expert in plant immunity and oomycete plant pathogens and one of the kindest scientists I’ve had the privilege to interact with — passed away far too young. John died on December 24, 2024 after a long battle with cancer, leaving a huge hole in our community and in the lives of everyone who knew him. In memoriam: John McDowell. Source: Virginia Tech. Many of us remember John not only for his incisive scientific mind but for the warmth and respect he showed to colleagues at every stage of their careers. His loss still feels personal — and professional. I’m grateful to have known him, learned from him, and laughed with him. Starting this piece with his memory feels right, because it reminds me why sense-making — the theme of this post — always begins and ends with people like John. 2 Stepping into big shoes John did a lot for the scientific community. Among other roles, he was also a co-editor of the Annual Review of Phytopathology with Iowa State University Gwyn Beattie. A few years ago, they recruited me to the editorial board and I enjoyed the annual board meetings, often in exotic locations. John’s passion for science — and for the field — shone through every gathering. As he said, this was his favorite day of the year. And yes — if you love science and phytopathology, it better be. About ten of us would spend the day debating advances in the field, finalizing topics and authors for the next volume. Because the board spans every angle of phytopathology, you always learn something new. The Annual Review of Phytopathology board meeting is my favorite day of the year — John McDowell. When Gwynn asked me to step in and replace John, my first reaction was simple — I really wished he were still around and that this moment didn’t exist. But saying no was never an option. Taking on this role felt like a responsibility: to John, to the journal, and to the community he cared so deeply about. A century of curated knowledge Annual Reviews will soon turn 100. That alone is remarkable. But what’s truly impressive is how this institution has stayed relevant — by sticking to a simple idea: let experts make sense of science. In a world drowning in information, Annual Reviews offers clarity. Each article is written by leading scientists and edited by board members. The goal isn’t just to summarize what’s new, but to explain why it matters, where the field is going, and what questions still need answers. It’s curated knowledge, not just content. On the importance of curation — In a world of AI, content curation is more important than ever. Source: Spencer Education. 3 And Annual Reviews has committed to open science —articles should be free to publish and free to read. That’s why I agreed to get involved. I wouldn’t be part of a publishing operation that wasn’t serious about open science. All new Annual Review of Phytopathology articles are freely available, and there’s a strong push to make science accessible — not just to specialists, but to students, teachers, and anyone with a curious mind. This shift didn’t happen by accident. Much of it is thanks to Richard Gallagher, the president and editor-in-chief. Richard is on a mission to make Annual Reviews a platform for giving back to the scientific community. He’s pushed for open access, championed new formats, and launched initiatives like Knowable Magazine, which translates science for the classroom, the newsroom, and beyond. The threats on the horizon Of course, generosity comes at a cost. Annual Reviews has committed to open access, which is the right thing to do — but it creates a paradox that all non-profit and society journals face. The journal now gives more while relying on less. Its business model still leans heavily on institutional subscriptions, especially from libraries. But why would libraries keep paying when articles are free to read? And then there’s the new elephant in the reading room: AI chatbots. Chatbots can already parse and condense Annual Reviews articles with alarming fluency. I’ve tested several of them — ask for a summary, and they deliver a decent one. Ask for figures, and you’ll get a numbered list. Ask for context or related ideas, and you’ll get those too. They don’t just summarize. they extrapolate. This raises a real question: what happens when AI systems can mine the entire Annual Reviews back catalog and serve it up, sliced and diced, on demand? One option might be to license the content — strike deals with major AI companies so that the underlying material is used ethically, and sustainably. These could be the library subscriptions of the future. But then the Trump administration’s push to weaken enforcement of copyright protections around AI training throws a bucket of cold water on that idea. If this vision becomes policy, content creators — including expert-driven publishers like Annual Reviews — could be cut out of the loop entirely. AI models would scrape and remix at will, without paying or crediting the original sources. So yes, open access and AI both promise a lot. But they also pose real, existential threats to the kind of curated, expert-driven science communication we desperately need. The nonsense problem The real challenge for AI isn’t information. It’s nonsense. The internet has always been full of it, but scientific literature is catching up fast. Bad science is everywhere: poorly designed studies, overinterpreted data, sloppy writing — and, increasingly, outright fabrication. It’s not just the fringe journals either — nonsense creeps in through the cracks of even the best ones. And the volume keeps growing. 4 Trust me, it’s bad. In my field of plant immunity, the level of bad behavior has sharply increased in recent years. And as I wrote a few months ago, I’m not just talking about flawed experiments or sloppy science. I’m talking about hallucinated data and entire research narratives that collapse under even modest scrutiny. It’s gotten so bad that many of my colleagues simply tune out entire swaths of the literature: A growing number of researchers in our field now routinely ignore large swaths of the literature — even in their own field. They’ve stopped reading certain papers on topics they themselves work on — because they’ve come to assume the results are unreliable or fabricated. Can chatbots flag bad science? The problem? Chatbots can’t tell the difference. They’re indiscriminate consumers of information. They don’t know if a paper was discredited or quietly ignored by the field. They don’t know which preprint aged poorly or which finding was quietly buried in the supplementary figures for good reason. Human experts still do this — or at least, we try. We draw on decades of accumulated experience, intuition honed by trial and error, and a lived familiarity with the nuances of our field. We’ve read thousands of papers. We know when something smells off. We know when it’s genuinely new, or just old wine in a flashier bottle. That kind of pattern recognition doesn’t come from token prediction. You don’t get it from remixing PDFs or pretraining on text. You get it by being a human, immersed in a community of practice. That’s the expert edge — especially in a complex field like plant pathology, where context and good judgment matter as much as data. And now, AI is about to flood the internet (and the scientific literature) with even more content. Some of it will be brilliant, no doubt. Much of it will be mediocre. A lot of it will be confidently wrong. It’s already getting harder to tell what’s real. The risk is that we end up in a hall of mirrors, where nonsense begets more nonsense, polished up by machines and passed off as insight. This is where human judgment matters most. Distinguishing sense from nonsense matters more than ever. Image by Nano Banana Pro. 5 Why experts matter more than ever In the coming flood, we’ll need lifeguards. Distinguishing sense from nonsense isn’t just a skill — it’s a moral duty for scientists. This is the real work: vetting claims, curating knowledge, contextualizing results. Anyone can publish a paper. Anyone can fine-tune a chatbot to sound convincing. But not everyone can tell which ideas actually hold up under scrutiny. That’s where experts come in. It’s what we do, imperfectly but persistently. We ask: Does this make sense? Is the evidence solid? Is this building on something real, or just stacking bricks on sand? Chatbots can’t do that. Not yet, anyway. They remix. They reword. They reorganize. But they don’t evaluate. They don’t weigh credibility, read between the lines, or recognize when a study is technically flawless but biologically meaningless. They don’t smell hype. We do. And as the deluge of AI-generated content accelerates, expert brands will matter more than ever. Individuals. Labs. Institutions. Names you trust. Think Annual Reviews. Think The Sainsbury Laboratory. These won’t just be content producers — they’ll be reference points. Markers of credibility in a rising sea of polished, plausible noise. Signs of hope in a noisy world Despite all the noise, I remain optimistic. The age of AI will change how we produce, access, and interact with scientific knowledge — but it won’t erase the need for experts. If anything, the value of trusted voices, curated sources, and thoughtful interpretation only grows. Experts matter more than ever. Institutions like Annual Reviews, and the wider community of scientists who take vetting and context seriously — they will become part of the essential infrastructure for navigating this new era. It’s not about stacking up dubious CNS papers that leave everyone scratching their heads. It’s not about hiding behind a glam journal editor’s decision. And it’s definitely not about one-hit wonders that collapse on replication —or worse, that no one even bothers to read, let alone replicate. Some things still need to be right. And right now, I don’t see chatbots capable of telling sense from nonsense. Keep doing the work John McDowell stood for that kind of science. Clear, careful, honest, principled, impactful. He didn’t chase hype. He made sense of things. That’s what we need more of now. So to all my fellow scientists: keep doing the work. Keep asking hard questions. Stay sharp. Stay honest. And most importantly, keep calling out nonsense when you see it. The bots are coming — and they’re impressive. But they still need us. More than ever. 6 Postscript: yes, I know experts aren’t perfect Let me be clear: I’m not saying experts are flawless. Far from it. I blogged and posted enough tweets about the #sciencecrisis to make that obvious. Science suffers from deep structural issues — perverse incentives, broken peer review, obsession with impact factors, and yes, sometimes even outright fraud. Experts can be biased. They can be wrong. They can be just as fragile and flawed as any system they work in. But many of them — many of us — are still doing this for the right reasons. Still showing up. Still trying to make sense of the world and make it better. That kind of commitment matters. Especially now. Keep doing the work. Be a scientist. Experts matter. But let’s not confuse credentials with credibility. Not all academics deserve to be called scientists. Check also: #sciencecrisis Acknowledgements I’m grateful to colleagues who helped shaped this post through discussions and social media comments. The article was written with assistance from ChatGPT 4o. This article is available on a CC-BY license via Zenodo. Cite as: Kamoun, S. (2025). In praise of experts: distinguishing sense from nonsense in the age of AI