>>105737092AlexNet was the first time anyone proved that AI was economically viable for commercial use-cases. It's claim to fame is hotdog not-hotdog. Plenty of commercial applications in computer vision and basic classification alone.
A lot of research was being done towards fundamental optimization like transfer learning, meta-learning (hyperparameter optimization), and cloud platform lifecycle stuff.
Then transformers came along in 2017, and this was the architecture that could hypothetically scale to infinity. That's what brought the billions in.
Like with any industry, you have to get in before it's cool. That's just how it is. Sometimes following your passion is the right thing to do, and other times it isn't