Fredrik deBoer's blog post - why selection bias is the most powerful force in education - has a pretty provocative title, yet it's something people have been speculating for quite some time.
Worth a read. Not sure if I agree 100% though. But, let's just assume Fredrik is right. Then, that's one of the reasons I think Learning at Scale (eg. MOOCs) has a bright future. Not that they're perfect, but it's very promising.
For example, MOOCs provide more learning opportunities to those with the access of the internet. I don't need to be at Stanford to learn about Convolutional Neural Networks and Computer Vision. This is a huge step, because we've pushed our boundaries further.
The next challenge is to make these learning possible for people without internet access (Or provide internet access to them).