the bitterest lesson
Brute force computation is not enough for intelligence. Clever new methods are needed. The science of AI is far from finished. IMHOView quoted post on X
Eliezer de Souza da Silva
@sereliezer
Assistant Professor at University of Coimbra. Probabilistic AI / Bayesian ML.
4.1K Following 1.6K Followers 3.6K Posts
Latest posts 5
If we think the different venues for publications as sinks that have a certain maximum flow capacity (which is true for all venues), and the current reality of llm-generated academic content (low-friction, low-cost, low-signal-to-noise-ratio, high-volume), 1/
is it me or twitter is again interesting? are people maybe going back to open speculation, thinking and debates and not beaten down by the times??? hopefully yes
something i have been thinking about is how persistent identity has some interesting properties for emergence of cooperation (and recently been learning about things like open-source game theory, which is transparence of code, another angle of this issue)... 1/3
I got a *Gold Reviewer* award in ICML 2026. Apart from the free registration, it is the right incentives to help the reviewing process become less noisy as the scale of the work just grows.
Awards for reviewers that:View quoted post on X
- completed >= 3 reviews
- completed all reviews for papers still active at end of reviewing
- ranked in top 25% (gold) or next 25% (silver) by ACs
Free reg for ACs that didn't drop the ball, roughly (email notifications later today)
