I think, the professor in TFA is pretty much wrong. For a reality check, please look at this Wikipedia link. It describes the current research status in the field of object recognition, a typical task in computer vision. From the article, it should become clear that there are numerous approaches to the problem, but there is not one really capable approach that solves the problem. This is independent of computing power - we just don't know yet how to make a machine recognize non-trivial objects under real-life conditions, not matter how many CPUs it has. This fact currently limits the development of autonomic cars already.

According to Quantum_decoherence, there is no superposition of quantum states. As the radioactive element interacts with the cat and the box, its decoherence time becomes very short, so the wave function collapses almost instantly. The cat is either dead or alive, but there is no extended state of superposition.

In a situation where your company outsources code generation to some external IT consultancy, it might be beneficial for that external developer not to document his code. After all, who are they going to call when they need to understand the code or extend its functionality?
This also work down the road - if they should get several offers for future coding work in an RFP, the original programmer can still price very aggressively, as he knows all the pitfalls of the solution that he wrote in the past.

"I say we take off and nuke the site from orbit. It's the only way to be sure. " - a story right out of Hollywood :)

Maybe I did not make myself clear enough - I never made the assumption or the claim that the brain resembles a computer in any way.
Recognition tasks are knowledge-driven; they are based on experience. Humans have to learn to see; a major input in the visual recognition process is experience. The brain has found a way to store years of learning in its structures. So I think it is fair to say that the brain has a vast amount of storage capacity.
The aforementioned conscious calculations are an emulation. The 7-item limit would not apply, if the calculations were executed subconsciously (by whatever algorithm).

I think this line of reasoning is not convincing.
The brain masters computations (e.g. visual recognition tasks, speech recognition etc.) that require enourmous amounts of memory when implemented in a conventional computer system. Under the assumption that those recogntition tasks are inherently memory-intensive, the brain has to have similar amounts of memory at its disposal.
So, obviously, the brain is not lacking memory to execute complex calculations, but it seems to disallow the conscious control of brain structures to execute multiplications. Each single neuron could execute such multiplications with ease.

This story tries to impress us with its "IT-is-not-for-girls" spin. But honestly, you would need a lot more data to draw such a conclusion. For instance, what is the current ratio of men leaving the IT profession in general and in the middle of their career in special? There might not even be a significant difference...
Moreover, what kind of women /people are leaving the industry? As far as I know, parts of the IT market are shrinking due to Offshoring and on-going automation, so a lot of simpler IT jobs are simply vanishing right now. So the loss of those jobs might also contribute to the observation of women leaving the industry.
Unfortunately, TFA does not provide more information to put the numbers in the right context...