The past ten years have produced great progress in natural language processing by computers, and also significant progress in studies of human language processing based on brain imaging studies. The thesis of this talk is that we are at a unique point in time where we can now study human language comprehension by attempting to align observed neural activity when humans read a sentence, with the internal representations of artificial neural networks, such as BERT, reading the same sentence. This talk will summarize key results over the past decade in our studies of human language processing applying machine learning methods to analyze brain imaging data, including our recent analyses of human sentence processing based on BERT.