Yesterday in class, a student asked me: if AI can already do so much, why do we still need PhD programs? Then, I came across an insightful blog by Minas Karamanis https://lnkd.in/gzpqshrz, particularly this line: “What’s great about science is its *people*. The slow, stubborn, sometimes painful process by which a confused student becomes an independent thinker. If we use these tools to bypass that process in favor of faster output, we don’t just risk taking away what’s great about science. We take away the only part of it that wasn’t replaceable in the first place.” I deeply agree with this, as it speaks to what I ultimately gained from my own PhD training. Looking back, the most valuable outcome of those six years was not the methods I learned--those can often be learned through online resources. It was not only the theories or knowledge I absorbed--those, too, can be studied independently through books and papers. It was not even the professional knowledge or network I built, which is maybe only meaningful within academia and my specific field. And not to say the degree itself--it mattered for getting the position I have now, but it would not necessarily have been essential for every possible career path that I could pursue. Before graduating, I asked myself: was this journey still seen as valuable as I had hoped it would be back in 2018, given all the tradeoffs: the years away from family and friends, the hobbies I gave up, the postponement of stability in my personal life, and the financial sacrifices. My answer was still yes. It was worthwhile because, in the end, I learned how to think and how to learn. More than any particular skill or credential, I value that training process itself. It gave me a way of approaching uncertainty, complexity, and future challenges that I will carry far even beyond academia. That is why I find the role of AI in science both promising and worrying. As a researcher who has spent years studying the social impacts of algorithmic systems, I do not see this simply as a question of individual agency or literacy. It is also a structural question: *What happens when people are placed in systems where reliance becomes hard to resist?* AI can certainly make parts of the process more efficient. But when it becomes deeply woven into training, it also becomes harder for students to resist relying on it rather than growing through the struggle themselves. And that is where the real danger lies: it often happens quietly, almost invisibly. The gains in efficiency are easy to notice; the gradual loss of agency is much harder to see. So perhaps the question is not whether AI should be part of PhD training--it already is, and it will be. The real question is about how to practice: what to preserve, and what thinking we do not want to outsource. If a PhD is still meant to cultivate independent thinkers, then this may be one of the most important questions for all of us to reflect on.

Visit the post for more.

The law will ban the video app in the United States by Jan. 19 if its owner, ByteDance, does not sell it to a non-Chinese company.