Neural networks have been around for almost a century now, first being invented at MIT in 1944. It is a very old AI learning model, and only really useful nowadays if u incorporate deep learning in the hidden layer, so that there are many more layers of filtering for input.

The REAL difference is actually in regards to training data. In the past, researchers would have to manually insert their own sample data and biases. But with the advent of social media, smartphone cameras, and google, we now have trillions upon trillions of photos of every single thing you could imagine. If we didn’t have pocket cameras or the internet then none of that would be possible. This is the only reason it had taken so long to see real applications of this technology.