Written by Trang Nguyen / Image by Pawel Czerwinski
It’s a loaded title, I know! Some of you may even say that it’s a clickbait, it is! But it’s the end of the flu season and my body is running on a dangerous combination of caffeine and cough syrups. Stay with me… I’ve been thinking a lot about AI, maybe a little too much, and I would like to share how it’s been disrupting my research experience.
The first disruption is my productivity level. I was cynical when my colleagues first told me about Codex. “Are you telling me that instead of spending two gruesome hours knitting together 56 lines of awk and sed commands, I can just type in a generally vague request, and five Yes/No questions later, it will give me what I want exactly how I want it?”, I thought to myself. I resisted for months until one day, during a boring meeting and at the urgency of a boring task, I decided to give it a try. Oh boy, I did not look back. My productivity increased the same way your blood glucose spikes up when that sweet sip of Dunkin’ Donuts iced latte hits your stomach linings first thing in the morning. Instantly. Ecstatically.
But easy comes, easy goes. Before I knew it, my honeymoon with Codex had turned into an extended season of Lost. My code folder was an island filled with mysterious creatures and death traps. I didn’t understand it, I couldn’t debug it, and I didn’t have the skills to verify the results of the more complicated analyses. I froze. I procrastinated. There were days I couldn’t even do the most basic statistical tests that I thought I knew like the back of my hand. There were nights I spent reading articles after articles looking for a justification for why I was struggling so much while everyone else was moving full steam ahead. Ultimately, I realized that I was (and still am) faced with two options: leave the island or learn to adapt. I have chosen the latter. The island is so damn enchanting. Not because it offers quick and easy solutions but because it opens up endless possibilities – hypotheses to explore, analyses to evaluate, and fascinating stories to weave into the tropical web of science. To navigate this complex and unknown terrain, I’m intentionally sharpening two skills in addition to the technical know-hows.
The first skill is verification. I have never heard the phrase “trust but verify” so often in my life. To be clear, we always should have done it: when a colleague sent us a figure for the manuscript, when a doctor insisted that the medication was safe for long-term use, or when a friend told us a “fun” fact about lemmings’ group suicide behavior. Most of us trusted but few of us verified. Now with AI agents who are masterful people pleasers, we have to. It’s not as easy as it looks, especially for junior researchers whose “bullshit” detectors are as mature as the vision of a two-month-old panda. Cute but lethally faulty.
The second skill is not only a skill but also an attitude, a spirit, a way of life: creativity. Nature recently published an article about how AI tools “expand scientists’ impact but contract science’s focus”, not surprisingly since they are trained on existing knowledge and prioritize areas that are already of popular interest. I don’t need to explain why this is concerning for both individual scientists (who don’t want to be trapped in a forever rat race) and the collective, but it underscores the need for thinking outside of the box or even dismantling the box altogether. Again, this is challenging for PhD candidates who, ironically, have historically excelled at following academic formulae. These two skills, verification and creativity, cannot be more different from one another: one requires standards and disciplines and the other encourages going beyond what’s already established. But they are the two sides of the same coin, just like how AI has sped me up and slowed me down simultaneously.
The most personal AI disruption to me, however, is grief. Underneath all the arguments for and against AI, all the do’s and don’ts, I am grieving the loss of agency. I used to wrestle with plots, even simple ones, for hours. The wrong columns, the wrong arguments, the wrong choice of plots. Even when everything was technically correct, it was not aesthetically right. The whole process was slow, stupid, and excruciating, but I miss it. The mistakes were mine to make, and so were the victories. I don’t need to do that anymore – the plots are now immediately produced and packaged in Christmas wrappers with a butterfly-shaped ribbon on them, and a heartfelt card that makes me feel like the smartest person in the world. For a whole three seconds. Until the inner critic in me emerges and wonders what I am even doing anymore. Am I learning anything? Am I contributing anything? How do I know when I have gone too far with AI? These are not the greatest thoughts to have at 2 AM. The greatest thoughts to have at that hour are no thoughts at all.
The loss of agency may be phantom, but the perception of loss is real. As I am cycling through the five stages of grief and all my conflicting sentiments about AI, I have realized something that literally made me laugh out loud. This is it! The struggle! The core of a PhD, of science, of evolution, of life! Charles Darwin and the Buddha might have been twin flames for all we know! I told you the title was a clickbait, I can’t believe you are still reading this, but thank you. Anyway, not to give Taylor Swift more recognition than she already has received, but if haters gonna hate, then thinkers gonna think and creators gonna create. Both solutions and problems. We are just dealing with different sets of them, with or without AI. The self-doubt has always been here, the anxiety has never left, and the discomfort is more pronounced than ever.
But that, my friend, is what you and I signed up for when we looked Science in the eye and said, “I do”. We exchanged two simple vows: to struggle and to survive, with the utmost tenacity to follow our convictions and the humble willingness to be proven wrong. Each of us gets to decide for ourselves exactly how to do that. This version of agency? No matter how good AI gets, it cannot be taken away.
P. S., this blog was 100% written by a real person with fears and hopes and dreams, several expired cheese slices in the fridge and a ridiculous amount of hair ties scattered on both sides of the Atlantic Ocean!