One of the questions I get every now and again is whether self-driving cars are a solved problem. Is there any work left to be done in this field?
The answer is that there is so much work left to be done! It only seems like a solved problem from the outside 🙂
So I was interested to read Francois Chollet’s answer to “Is Deep Learning Overhyped?” on Quora.
Chollet is the author of Keras, which is a deep learning library we use in the Udacity Self-Driving Car Program. He explains at length why artificial intelligence generally, much like autonomous driving specifically, is not a solved problem.
Overall: deep learning has made us really good at turning large datasets of perceptual inputs (images, sounds, videos) and simple human-annotated targets (e.g. the list of objects present in a picture) into models that can automatically map the inputs to the targets. That’s great, and it has a ton a transformative practical applications. But it’s still the only thing we can do really well. Let’s not mistake this fairly narrow success in supervised learning for having “solved” machine perception, or machine intelligence in general. The things about intelligence that we don’t understand still massively outnumber the things that we do understand, and while we are standing one step closer to general AI than we did ten years ago, it’s only by a small increment.
There’s still a lot of work left to do!