Hating and loving AI in 2026

Human-curated science is getting more important, not less

In the 30 years that I’ve been in management, many technological advances have come along that were supposed to save money, make things more efficient, eliminate jobs, and potentially destroy industries. So far, none of them has. Here’s a trip down memory lane:

• Laptops in the classroom

• CD-ROMs for general chemistry and related classes

• Moving class registration onto the internet

• Making figures for scientific papers using Photoshop

• Massive-open online courses (MOOCs) like Coursera and EdX

• Moving scientific journals online

• Integrated HR/finance/accounting systems like Workday

• Electronic health records like Epic

I have been involved in implementing every single one of these in one way or another, either doing it myself as a faculty member or overseeing and paying for these things as an administrator. Every single time, the salespeople convincing me about this stuff told me I would be able to do more and save money because less human effort would be required. Every single time, they were half-right. I could do more, but I spent a lot of money implementing them all (all the way up to 100s of millions for Epic and Workday), and I needed even more people to make sure the systems were working properly and to take advantage of all of the new capabilities.

In some cases, the effects of these cycles have been very damaging. The University of Virginia famously fired their president because she wasn’t adopting online courses fast enough. Within a few years, MOOCs would be an add-on to higher education at best. Fortunately, the faculty saw through this and she got her job back. The move of scientific publishing online was supposed to eliminate journals, but it only made commercial publishers bigger and more profitable. And integrated systems like Workday and Epic have allowed people to do tasks they must do, but the number of jobs in university accounting and health care administration has only increased.

Here they come again

And now the salespeople are showing up with their AI wares. Tools for everything under the sun. I can’t possibly answer all the LinkedIn messages from the next person with the tool that will not only change my life, but save it from the coming apocalypse. (Also, I don’t answer these, so sorry for those I’m ghosting.)

This week, I’m out with a column about how this plays out for Science in 2026.

Like everything, AI is a two-edged sword. On the one hand, it’s leading to a degradation of the scientific literature as bad summaries and AI overviews lead people in the wrong direction, bad AI images and text show up in papers, incomplete scientific review carried out by AI leads to stuff getting through, and worse.

On the other hand, it can be used to make scientific papers better. We have long used it to look for text plagiarism, and now we can use it to catch more altered images and to make sure that adequate supporting data are posted with papers; check out this excellent blog from Valda Vinson and Lauren Kmec about how we are doing that. But all of these improvements have cost money. The software has to be licensed, the reports have to be analyzed by humans, and the problems that are turned up all have to be dealt with. Yes, these tools are making our papers better, but they are not saving money! I know because I have to ask for the resources to pay for all of this.

This time is different - (ha)

When I explain my skepticism to AI zealots, I often hear that “this time is different.” I’m conditioned never to believe this statement. Why would the whole history of technology suddenly change? I believe AI will change jobs, and I’m confident it will make many lower-paid workers’ lives worse by increasing surveillance. But I’m going to stay skeptical that knowledge work will be less expensive until I see hard evidence of that.

In the meantime, human-curated science is getting more important, not less. As AI turns the existing literature to secondary mush and generates even more primary mush, we’re going to keep our heads down and publish papers that we are willing to stake our human selves behind.