A deep learning model achieves super-human performance at Gran Turismo Sport (https://techxplore.com/news/2020-09-deep-super-human-gran-turismo-sport.html)
The results of deep learning are always amazing but I still find it to be a somewhat lazy or uninspired approach to solving problems. In the end it's just random trial and error. Usually this is the least desired way because it yields no deeper understanding of the problem. It has only become practical through the insane speeds and "precision" of modern processors.
There are definitely places for it in engineering and programming but I hope that AI will not take a toll on human inventiveness in the long run and lead to more superficial analytical thinking.
Anyway, I wonder how long it would take that AI to master GPB :D Should be a little more complex.
Also, how would it respond in race situations with more unpredictability?
Nota big fan of deep learning in general but it has some reasonable applications. Here, it laps 10s faster than the regular approach and probably costs even less to develop and tune. So yeah, probably worth it.