I think it comes from a place of frustration. Some people don't learn well from lectures at all, but because they're considered the default so strongly, other conduits of learning aren't usually provided.
I agree, but I think the "word problems" the author was referring to are much lower quality than the ones you're thinking of. I imagined some highly-contrived exercises where all of the relevant information is already pre-processed for you, removing any need for problem decomposition. For example, "if the angle between the ground and a tree's shadow is 45 degrees and a 50 ft tall telephone pole that's 10 ft away from the tree casts a shadow..." (substitute a similar differential equation problem). As you point out, half the fun is defining a problem and breaking it down, and these kinds of word problems don't give you a chance to do that.
I've heard that this is somewhat common in nerdy communities, or at least common enough that someone felt moved to write about "geek social fallacies": http://www.plausiblydeniable.com/opinion/gsf.html
I think the author's criticisms are mostly invalid (see Chevalier's comment for a great takedown). With MOOCs, I worry more about these things instead: 1) MOOCs lack the social/deadline pressure to get people to finish their work, and 2) people learn by doing things actively, which is much harder to encourage over the internet.
For 1, as a person with ADHD I struggle to get work done. The deadlines and social pressure to meet them are sometimes the only thing forcing me to actually do work. If people formed local communities to push each other to finish their MOOCs, that would help, but this isn't built-in to MOOCs directly like it's built-in to traditional universities.
For 2, I think people only learn when actively doing something. Passively listening to a lecture does nothing; the lecture only becomes knowledge when you think about it or use it in some way. Active learning exercises are harder to build in to MOOCs because one of the best ways of evaluating learning - having someone "qualified" look at your work and provide feedback - isn't scalable. Multiple-choice questions are too easily gamed and rarely encourage deep thinking. Things like discussions, or iterative feedback on problem-solving, or essay critiques are much more useful, but again, not easily scalable.
It might be that forums where other students critique your work or help you through difficult assignments could be a good substitute. But maybe not. I'm a teaching assistant for a computer science department, and it's difficult to encourage a student in just a right way to help them get the answer for themself. It's tempting to just give out the answer, which isn't helpful for the student but is much easier for me. The collective wisdom of a forum might not be a good substitute for the subtle patience of a professor.
However, neither of these concerns mean we shouldn't pursue MOOCs. I want MOOCs to one day be a complete substitute for a university education. But to do that, concerns like these will need to be addressed.
Plastic parts are often made by pouring hot plastic into a mold. When the plastic is done drying, often a robotic arm or crane needs to pull it off to bring the plastic to the next stage of the manufacturing process. But it's hard to get the plastic off the surface once it's dried; it needs to be "popped" off of it. So manufacturers build ejector pins into the mold. When the plastic is done drying, the pins pop up and push the plastic off the mold, into the crane or whatever will take it to the next part of the operation.
I wonder if this algorithm would work if someone uploaded a rotated version of a previously uploaded icon. As it only compares row adjacent values, if the rotation was 90 or 270 degrees I'm not sure it'd detect the duplicate.