When people think about AI today, they tend to imagine big names like OpenAI, Google, or Nvidia — and maybe a handful of flashy CEOs. But the real story of AI is packed with lesser-known researchers who laid the essential groundwork for everything we’re doing now. These are the people whose names rarely get mentioned, but without them, we wouldn’t be talking about ChatGPT, GPT-4, or any of the massive models dominating headlines.
Let’s rewind and give credit where it’s due — to the forgotten pioneers of AI who made modern AI possible.
Everyone talks about Geoffrey Hinton when they mention backpropagation — and sure, Hinton popularized and extended it. But Seppo Linnainmaa, a Finnish master’s student, actually developed the mathematical foundations for backpropagation way back in the 1970s. And here’s the kicker: he didn’t even apply it to neural networks.
His work focused on computing derivatives efficiently — something that later became essential for training deep neural networks. But because Linnainmaa wasn’t working directly in AI, his role is barely acknowledged in the mainstream AI narrative. Without his work on automatic differentiation, though, none of the AI models we see today could learn.
So while the world calls Hinton the godfather of deep learning, it was Linnainmaa who secretly handed him the math that made it all work.
When people talk about ImageNet and the birth of modern AI, they usually mention Geoffrey Hinton, but what they often forget is that the actual engineers who built and trained AlexNet were Alex Krizhevsky and Ilya Sutskever.
Yes, the first large neural network that “saw” and understood images like a human was trained not by Hinton himself but by his students. These guys took a gamble on GPUs, which at that time were only used for gaming. And they didn’t just train a small toy model — AlexNet was 60 million parameters, enormous for the time.
And here’s a detail that almost nobody mentions: they estimated that training their model on CPUs would take months. But once they hacked together a way to run it on two GTX 580 GPUs, they brought training time down to days. That breakthrough changed AI forever — and yet most people couldn’t name Krizhevsky or Sutskever in a lineup.
Also, let’s not forget Sutskever later co-founded OpenAI and led the teams that built GPT-2 and GPT-3. So yes, the same guy who hacked AlexNet on GPUs is one of the people responsible for LLMs that run the world today.
Everyone loves to talk about transformers and “attention is all you need”, but few people know that the foundational math behind the attention mechanism was actually developed earlier by Dimitri Bahdanau.
Bahdanau was a researcher at the University of Montreal, and his work on neural machine translation introduced attention to AI — a concept that allows models to “focus” on relevant parts of an input sequence when making decisions.
Without his contribution, there would be no transformers, no GPT models, and no ChatGPT. Yet, his name is rarely mentioned when people rave about modern LLMs.
So next time you hear someone casually drop “transformers changed everything,” you might want to remind them that Bahdanau’s attention mechanism was the real spark.
If there’s one person who should be a household name in AI, it’s Fei-Fei Li, the creator of ImageNet. Before ImageNet, AI models were stumbling through blurry, tiny datasets. Nothing close to human-level visual understanding.
Fei-Fei Li’s ImageNet — 3 million labeled images across 5,000 categories — became the gold standard dataset that enabled AI models to finally “see” the world. And training on ImageNet is what led to the AlexNet breakthrough that kicked off deep learning as we know it.
Here’s the wild part: before ImageNet, the data simply didn’t exist to train large vision models. Without her vision (no pun intended), there would be no computer vision revolution. Yet, her name is often overshadowed in discussions of AI history, despite her pivotal role.
We all know that GPUs changed AI, but few realize how close AlexNet came to never happening. When Hinton’s team hit the wall trying to train their model on CPUs, it was Jensen Huang (NVIDIA’s CEO) and his engineers who stepped in.
Back then, GPUs were a gamer’s tool, not an AI tool. But when Hinton explained their problem, NVIDIA’s engineers helped rewrite the training workflow to run on GPUs, giving birth to the AI-GPU symbiosis we see today.
Without that collaboration, AlexNet might have taken months to train — or never trained at all. Yet, the engineers behind that breakthrough remain in the shadows.
So why bring up these names now? Because the AI story is incomplete without them. When we pretend that AI is just a product of a few famous labs or billion-dollar companies, we erase the decades of foundational work by people who were doing AI when nobody cared.
These pioneers didn’t have massive compute clusters or VC money. They were grinding it out in labs, piecing together algorithms and theories that now power trillion-dollar industries.
If we forget them, we miss the real story of how AI came to be. And more importantly, we risk forgetting that innovation often comes from overlooked places, from people nobody’s paying attention to — until they change the world.
Next time someone throws around AI buzzwords like “transformers” or “neural networks” at a party, remind them: there’s a whole army of hidden figures who made all this possible. AI didn’t start with ChatGPT — it started with people like Seppo Linnainmaa, Fei-Fei Li, Alex Krizhevsky, Ilya Sutskever, Dimitri Bahdanau, and a bunch of NVIDIA engineers who just wanted to make GPUs run faster.
So here’s to the ones who laid the foundations. The real pioneers of AI.
Commenting Rules: Being critical is fine, if you are being rude, we’ll delete your stuff. Please do not put your URL in the comment text and please use your PERSONAL name or initials and not your business name, as the latter comes off like spam. Have fun and thanks for your input.
Join a growing community. Every Friday I share the most recent insights from what I have been up to, directly to your inbox.
