Lesson 317 of 2116
AlexNet and the Deep Learning Revolution
In September 2012, a neural network crushed ImageNet and everything about AI changed.
Lesson map
What this lesson covers
Learning path
The main moves in order
- 1The Benchmark That Broke
- 2AlexNet
- 3ImageNet
- 4GPU training
Concept cluster
Terms to connect while reading
Section 1
The Benchmark That Broke
In September 2012, Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton submitted a convolutional neural network to the ImageNet Large Scale Visual Recognition Challenge. The benchmark had a million labeled images across a thousand categories, and the best classical computer vision systems had plateaued around 26 percent top-5 error.
AlexNet scored 15.3 percent. The runner-up was at 26.2. A ten-point leap in a benchmark that had been inching forward by fractions shocked the computer vision community.
The three ingredients that compounded
- 1ImageNet: Fei-Fei Li's team spent years crowdsourcing a million labeled images
- 2GPUs: NVIDIA's parallel hardware, built for graphics, happened to be perfect for neural networks
- 3Algorithms: deep convolutional networks with tricks like ReLU and dropout that trained reliably
Within a year, nearly every ImageNet entry was a deep network. Within three years, deep learning had reshaped speech recognition, machine translation, and drug discovery. The AI community as it existed in 2011 barely resembled the one of 2015.
“The dirty little secret is we don't understand why they work.”
Key terms in this lesson
The big idea: AlexNet did not invent deep learning, but it proved the recipe worked at scale. Everything that followed, from AlphaGo to GPT-4, traces through that 2012 submission.
End-of-lesson quiz
Check what stuck
15 questions · Score saves to your progress.
Tutor
Curious about “AlexNet and the Deep Learning Revolution”?
Ask anything about this lesson. I’ll answer using just what you’re reading — short, friendly, grounded.
Progress saved locally in this browser. Sign in to sync across devices.
Related lessons
Keep going
Creators · 45 min
Designing Your Own Eval
The eval that matters most is the one tied to your real task. Here is a step-by-step way to build one. The rubric is the product Most 'AI product' failures are actually rubric failures.
Creators · 30 min
Label Noise: When Your Ground Truth Is Wrong
Every labeled dataset has mistakes. Studies have found error rates of 3 to 6 percent in famous benchmarks like ImageNet. Noisy labels confuse models and mislead evaluations.
Creators · 55 min
The Three Ingredients: Data, Compute, Algorithms (Capstone)
Every AI breakthrough of the past decade rests on three interacting ingredients. Synthesize everything you have learned into one working model.