Regular neural networks were popular until they stopped delivering best-in-class results for a lot of problems. Recent hardware made it possible to train more

Regular neural networks were popular until they stopped delivering best-in-class results for a lot of problems. Recent hardware made it possible to train more layers and they're now getting cutting-edge results in many areas again so they're now getting more attention again.