Researched
Deep Learning
Deep neural networks trained on graphics cards with millions of images beat all classical methods in image recognition in 2012.
Open in the interactive tree →AlexNet (Krizhevsky, Sutskever, Hinton) cut the error in the 2012 ImageNet competition to 15.3% versus 26.2% for the runner-up, trained on two Nvidia GTX 580 cards. Deep learning then took over speech, translation, games and medicine. Hinton, LeCun and Bengio received the Turing Award for it in 2018, and Hinton and Hopfield the 2024 Nobel Prize in Physics.
Prerequisites
- Linear Algebra & Matrices1858Neural networks are layers of matrix multiplications
- The Internet1969
- Hopfield Net & Boltzmann Machine1982Hinton's Boltzmann-machine pretraining led to deep belief nets
- Backpropagation1986
- Convolutional networks (LeNet)1989AlexNet is a deeper, GPU-trained convolutional network
- Graphics Processor (GPU)1999
- ImageNet dataset2009ImageNet supplied the training data and the 2012 contest
Unlocks
- Word Embeddings (word2vec)2013
- Attention Mechanism2014
- Generative Adversarial Networks2014
- Robots & Lights-Out Factories2015–2026
- AlphaGo2016
- Transformer Architecture2017
- AI in medical diagnosis2018–2026
- AI Forecasting & Supply Twins2020-2026
- AI protein structure prediction2020
- Diffusion Models2020
- Precision agriculture2020s
- Robotaxis in Daily Service2020–2026
- AI bioacoustic monitoring2021-2026
- AI Reads Archives & Inscriptions2022-2026
- Generative Image & Video AI2022
- AI Discovers New Materials2023
- AI Reads Herculaneum Scrolls2023
- Brain-computer interfaces2023
- Animal-sound foundation models2024-2026
- Brain wiring maps2024
- Project CETI: sperm whale codas2024-2026
- AI weather models2025
- Synthetic genomes2025
- Whole Herculaneum Scroll Read2025-2026
- AI Accelerators & Data Centers2026
- Humanoid Robots2026
- Rubin and Roman sky surveys2026
- Deciphering Linear Aopen
- Deciphering the Indus Scriptopen
- Earthquake predictionopen
- Everyday Robot Dexterityopen
- Forecasting volcanic eruptionsopen
- Level 5 Autonomous Drivingopen
- Predicting traits from DNAopen
- Simulating a whole brainopen