Wrapping up the archives - 2022

Fourth and last archival post. A video, another problematic competition, and more general thoughts on the direction of AI.

The blog was mostly inactive after March, 2021. I had a lot of articles planned, but between the teaching and several publications that we were wrapping up, plus of course the small detail of actually writing the thesis.

In March 2022, I did get back into it after the publication of our analysis of the MoNuSAC 2020 challenge result, with “Final comments on the results of the MoNuSAC challenge”.

I also published “A list of digital pathology image analysis challenges”, as I reviewed a lot of them for the thesis and thought it could serve as a useful quick reference.

In October 2022 — after the private defense of my thesis — I tried to express more general thoughts about some things that bothered me in “Is AI on the wrong path?”. To summarize: bigger, more complex models work well enough to improve results on well-curated benchmarks, but that comes at a cost, and their performance on real-world, complex datasets — such as those found in digital pathology — are not quite as convincing.

Finally, I celebrated the public defense of “The thesis” by sharing the dissertation and the slides of the presentation and, in “A new project”, I explain a bit what’s next. This new substack being, obviously, a part of it.