I do agree with your summary. However as brains get more complicated, such as humans, they seem to also become more accepting of changes. For instance, in some people MRI scans indicate activity in sometimes totally different locations. If this is indeed the case then replicating area X to achieve action Y is not nearly so straightforward.
Good luck with growing simulated neurons and their connections. The brain is more complicated than the known universe. The problem with this approach and all decision problems such as this is the massive amount of levels of probabilities. Suppose a probabilistic choice was made near the beginning when a different one should have been made. How will they know that?
This is the problem with science today. Projects don't get funding unless they are wildly out there in terms of concepts. Most people fail to realize though that science actually moves in small increments not wild jumps.
There is a difference between a couple of post docs or even grad students and 1.34 billion. I have a masters in CS with an emphasis in AI. AI will never be "solved" in one giant project like this. Think about trying to create an OS by building huge massively connected models which link various code snippets together. Given enough time you are guaranteed to solve it, but it might be more than the age of the universe.
Massive large projects like this almost always end in utter failure. Even the IBM cat brain project failed to accomplish much. Intelligence is much more complicated than a mere randomly connected neural network. I just hope something good comes from this and it is not a total waste.
The problem is he is pointing out systems like VARS which are woefully inadequate for vaccine related problems. There has been 2.5 billion paid out by the legal system for vaccine related injuries. http://www.hrsa.gov/vaccinecompensation/data.html I guarantee you that for every court case there is at least a 1000 others that doctors have explained away. Think about it for a second.