Geoffrey Hinton talk "What is wrong with convolutional neural nets ?"
What it is
Geoffrey Hinton gives an MIT talk on what is wrong with convolutional neural networks and introduces his capsules project. The description points to the capsule network papers.
Why Hacker News recommends it
Recommended as Hinton's own explanation of capsule networks and his critique of pooling in CNNs.
- 2 people recommended it
- 2 different months
- 0 strong recommendations
- first 2017, last 2018
What HN commenters said
“Prof. Hinton has an interesting talk about his new 'capsule' model based on psychology of shape perception”
Hinton on capsule model based on shape perception psychology
bra-ket, 2017-09 · Recommendation · in “Is AI Riding a One-Trick Pony?”
“Hinton explains the concept of capsules in this video”
Hinton explains capsules in detail
dr_zoidberg, 2017-11 · Recommendation · in “A New Twist on Neural Networks”
Excerpts from Hacker News comments, linked to the originals. Summaries and labels were written by a language model from those comments.
Topics: geoffrey hinton, capsules, convolutional neural networks, deep learning, pooling
Also recommended by the same people
courseVectors | Chapter 1, Essence of linear algebra3Blue1Brown · 17 HN recommenders · 2018–2026
lectureThe spelled-out intro to neural networks and backpropagation: building microgradAndrej Karpathy · 17 HN recommenders · 2022–2025
explainerBut what is a neural network? | Deep learning chapter 13Blue1Brown · 16 HN recommenders · 2017–2025
documentaryAlphaGo - The Movie | Full award-winning documentaryGoogle DeepMind · 10 HN recommenders · 2020–2025
explainerTransformers, the tech behind LLMs | Deep Learning Chapter 53Blue1Brown · 13 HN recommenders · 2024–2026
talkUSENIX Security '18-Q: Why Do Keynote Speakers Keep Suggesting That Improving Security Is Possible?USENIX · 10 HN recommenders · 2018–2026
courseLecture 1 | Machine Learning (Stanford)Stanford · 11 HN recommenders · 2008–2020
demoMarI/O - Machine Learning for Video GamesSethBling · 10 HN recommenders · 2015–2026