Artificial intelligence · 2012
Google Brain cat neuron
A network trained on YouTube stills learns to detect cats with no labels.
Andrew Ng and Jeff Dean's team trained a 1-billion-connection network on 16,000 CPU cores with 10 million unlabeled frames. A neuron emerged that responded to cats. It convinced Google to invest heavily in deep learning.
What it built on 2
Related measurements
The dashed line marks when it first worked.
Training compute of notable AI models Source: Epoch AI, Notable AI Models (CC BY 4.0)
ML hardware FLOP/s per dollar Source: Epoch AI, ML hardware (CC BY 4.0)