Playing with AI, Cities, and Whatever Futures We Can Dream Up ✨
I tinker with AI, cities, and democracy — trying to make sure robots don’t run everything (yet).
About Me
Hi, I’m Hugo 👋 Basically a full-time nerd for AI, cities, and sustainability. Right now I’m a postdoc at theChaire UNESCO en Paysage Urbain, where I explore how AI, public spaces, and democracy can actually vibe together instead of awkwardly ignoring each other.
Before that, I did my PhD at Mila (aka Montréal’s AI playground), where I explored how teaching AIs to compete — like in image generation and adversarial games — can push the boundaries of what they create.
My thesis was titled “Adversarial games in machine learning: challenges and applications.” Translation: imagine two AIs locked in an endless video game — one fakes images, the other tries to spot the fakes. Fun to watch, a nightmare to train. My work tried to make that chaos slightly less chaotic: bringing in mathy optimization tricks, poking at why GANs love to spin in circles instead of settling down, and even inventing new adversarial games to generate sneaky attacks across different models.
In short: I research how to turn messy AI battles into something useful (and maybe even a little elegant).
Projects
AI that plays nice with urban planners. Think SimCity, but for real life.
A workshop that’s basically AI therapy for citizens. Bring snacks.
Community consultation… but make it ✨fancy✨. Social housing never looked this stylish.
Values
Hot take: AI is not neutral (shocker!). I’m into making it inclusive, sustainable, and maybe even fun. Democracy included, free of charge. If tech doesn’t make life better for actual humans (and maybe pigeons 🐦), then what are we even doing?
Beyond Research
When I’m not writing about AI, I’m usually on a bike for way too many hours 🚴♂️, yelling about climate change 🌍🔥, or building silly tools for collective brainpower 🧠✨. Basically: I get nerdy, I get sweaty, I get loud.
Publications
PhD thesis, Mila/Université de Montréal.
GANs are cool but a mess. I tried to fix them with optimization math, found out they like to spin in circles, and then invented new games that generate adversarial attacks just for fun.
Imagine SimCity, but instead of Godzilla destroying it, it’s biased algorithms. I wrote about how to stop that.
Spoiler: training giant models eats electricity like a teenager devouring pizza. Here’s what we can do about it.
It’s like group therapy, but for AI anxiety. Cards, debates, and collective brainpower included.