How we got here
AI isn't new, so why did it suddenly explode? The decades of work, the AI winters, and the three things that finally came together to make this moment.
Here's a fact that surprises people: AI is not new. Researchers have worked on it since the 1950s. Inside Salesforce, "Einstein" has been our AI brand for years, and you still flip on Einstein to turn on Agentforce today. So if AI has been around for seventy years, why does it feel like it appeared out of nowhere in 2022?
Because three separate things finally lined up at the same time. None of them is magic. Each is worth understanding on its own, because the rest of this module is really just unpacking them.
The three ingredients
Enough data. Modern AI learns by example, and it needs a staggering amount of it. The internet quietly became that pile of examples: decades of text, images, and code, all in one place, all machine-readable. No internet, no training data.
Enough compute. Learning from that much data takes an enormous amount of raw mathematical horsepower. The chips that made it affordable, the massively parallel GPUs we'll meet in the next pages, only became powerful and plentiful enough fairly recently. (This is why a graphics-card company became one of the most valuable on earth. More on that soon.)
A breakthrough design. In 2017, a group of Google researchers published a new model design called the transformer. It was the piece that let models actually use all that data and compute well. Almost everything you call "AI" today is built on it.
Data was the fuel, compute was the engine, and the transformer was the design that made the engine run. Take away any one of the three and this moment doesn't happen.
So what changed in 2022?
ChatGPT. Not because anything was invented that day, but because for the first time the technology was wrapped in something anyone could just talk to. The breakthroughs had already happened quietly in research labs. ChatGPT was the moment it landed in everyone's hands at once, and the moment your customers started asking you about it.
Why this matters for us
The same wave that produced ChatGPT is exactly what makes Agentforce possible now and not five years ago. When you understand the three ingredients, you can reason about where this is genuinely going versus where it's just noise, instead of reacting to every headline. That's the whole goal here: learning how to learn AI, so each new release is a variation on something you already understand.
📝 Practice
Next time you see an AI headline, try to place it against the three ingredients: is this news about more data, more compute, or a better design? Almost every real advance is one of those three. The ones that fit none of them are usually marketing.