A recent research demonstrates that application of machine learning techniques applied to brain activity in a dream state do well against activity found in waking responses to visual stimuli (1). In general, visual contents of dreams seem to correlate well with those observed when the brain is awake.
The brain, albeit being a non-linear quantum computer, harbors characteristics that could be understood by traditional techniques. Machine learning techniques have been improving although most are extensions of deterministic statistical methods, engineers have been using for many decades. The observation that even such a crude technology is able to correlate brain states is encouraging and it may imply that the practice, if not the theory, of the brain is understandable. Research into biological and artificial intelligence progressed along opposite directions with little in common. The ability to build robust models from brain patterns that do well in predictions in different brain states could open a path toward collaboration among experts in these different areas.
Convergence in technologies and ideas is the most powerful concept yet. Segmented research is increasingly less productive.
(1) Neural Decoding of Visual Imagery During Sleep
T. Horikawa1,2, M. Tamaki1,*, Y. Miyawaki3,1,†, Y. Kamitani1,2,‡