The central question
AI history is often told through a few famous labs and CEOs. That misses the deeper story. Modern AI depends on many researchers and engineers whose foundational work rarely gets the same attention.
Seppo Linnainmaa gave neural networks the math they needed
Backpropagation became central to deep learning, but its mathematical foundation traces back to Seppo Linnainmaa’s work on automatic differentiation. That work made it possible to compute gradients efficiently, even before it became famous inside neural networks.
Alex Krizhevsky and Ilya Sutskever turned AlexNet into a working system
The ImageNet breakthrough is often associated with Geoffrey Hinton, but AlexNet was built and trained by his students Alex Krizhevsky and Ilya Sutskever. Their use of GPUs made the experiment practical and helped launch the deep-learning wave.
Dimitri Bahdanau made attention practical
Transformers depend on attention, but attention entered neural machine translation before the transformer paper. Dimitri Bahdanau’s work gave models a way to focus on relevant parts of an input sequence instead of compressing everything into a fixed representation.
Fei-Fei Li built the dataset that made machines see
ImageNet, led by Fei-Fei Li, gave computer vision the scale and structure it needed. Without that dataset and benchmark, the AlexNet moment would not have had the same force.
NVIDIA engineers made the hardware shift usable
GPUs became central to AI because engineers turned graphics hardware into a practical training platform. Hardware alone was not enough. The software stack and researcher support mattered.
What these pioneers had in common
- They solved foundational problems before the commercial wave arrived.
- Their work became infrastructure for later breakthroughs.
- They show that AI progress comes from long chains of technical contribution, not single moments.
The practical point
The AI story is richer than the public narrative. Modern systems are built on decades of work by people who made the math, data, hardware, and architectures possible.