Hey man, great article. I’ve studied computer science and artificial intelligence for 5 years and I work in the A.I industry. Here are my thoughts, Hidden Markov models are great for building simulated NNs and they’ve pushed us down the road for some time but…I don’t believe the answer to general intelligence will be found by studying Hidden Markov Models. We should be looking at how the human brain changes itself and more generally “how nature builds intelligence”. You might be interested in looking into the Anti-Hebbian and Hebbian rules that change the synaptic connections in the human brain and operate in nature.
At the end of the day if we wanted to simulate 30 billion neurons on a normal computer we would need more energy than we produce in a year and it would take many many many years train. My guess: we need to train NNs with local rules and they need to be physical not simulated.
It’s actually being done already! and it’s called neuromorphic computing. Check out what Knowm.org are doing with memristive devices. Good luck.