Doctoral colloquium - Marek Šuppa (28.9.2026)
Monday 28.9.2026 at 13:10, Lecture room I 9
Marek Šuppa:
AI Can Do Your PhD. So Why Are You Doing (NLP) Research?
Abstract:
No matter how you look at it, doing AI research has become weird in the past few months. Models can write code, run experiments, draft papers, argue with reviewers, review other papers to pay your dues, and even run the obligatory victory lap on social media when your work somehow gets accepted. Meanwhile, conferences drown in submissions and everyone is still told to publish more. It is tempting to blame AI but that misses the point: AI did not create the rat race, it just made running it much cheaper.
Since my last PhD colloquium, I have, somewhat ironically, added a few more papers to the pile myself. In this talk, I’ll use that progress, and our recent work on Slovak and low-resource NLP, to ask some increasingly uncomfortable questions. If AI can automate most of the publication loop, are we still training researchers for the right job (or any real job, that is)? If systems can increasingly choose experiments, learn from the results, and improve how they do research, why does an academic career still look remarkably similar? And if research itself starts becoming a recursive self-improving loop, what exactly will we need career researchers for?