Jeff has one of the classic problems of an autodidact - gaps in his knowledge. If you followed along with the saga of SO's creation on codinghorror, you saw him run into these headlong on many occasions. He also tends to pontificate on issues he's learned recently and doesn't understand completely a bit too often for my liking. One thing he does have going for him though is an unbelievable amount of pluck.
Sounds like Jeff has a non-CS background; aka, he's one of these guys that has the capability to really understand this stuff at a deep, meaningful level, but sometimes (often?) runs into his own limits (note: this is not to say CS 'people' are incapable of this - but rather - this is a trait that is not automatic whether you are 'self-taught' or a CS grad).
The 'plunk' you are describing is his ability to just 'get it done' - a trait I think most tech founders greatly underestimate. It's an ability or mental strength that makes a person able to will the project over the goal-line. I'm amazed how hard it is to complete anything - whether it's a startup app I'm working on or a piece of music - finishing it 100% and getting it 'in the can' is no small feet. It's special gift, and one that I would highly recommend anyone searching for a founder look for in their partner(s).
I have to admit I like Jeff, from his articles and such he seems to have a pretty clear way of thinking through big tech issues, and while he may sometimes surprise me with a 'lack of knowledge', I just remind myself that we all have gaps in our knowledge and at the end of the day, he's done some incredible work, and this 'lack' hasn't hurt him much (maybe it caused him to take longer to get to where he is, but he still managed to get it done).
Ha! Good point... I was trying to get at this idea that some people are naturals and Jeff seems to be that type of person; for what he might lack in formal education - and hence may run into these 'limits' more than others might - he has other worthwhile traits that compensate or even compliment this perceived lack on his part.
For all we know it's his lack of knowledge that gives him an edge - he runs into something he doesn't get or hasn't been exposed too and dives in, eats it up because he needs to really know it. Compare that to a guy that takes a class on some esoteric aspect of coding, when he/she doesn't have any need to apply it, when they just care about getting a good grade - which person really has learned the material? Does the person that already learned it but didn't need it at the time have a really big advantage? They might work through the problem faster, so maybe that's the advantage - you save some time.
But I like the root of your point - one doesn't grow/learn without trying to hit their own limits and push past them. Good stuff man.
Theres one big difference between real life and (undergrad?) CS classes: the discovery phase. In most classes, the solution is laid out for you or implied in some way. Algorithms classes generally just consist of following a spec outlined by the book/teacher.
But when you run across a problem in real life it doesn't tell you to use a specific algorithm to solve it. Because of this, finding the solution usually requires a deeper level of knowledge.
When I graduated, that was the scariest part for me. Sure, I knew a bunch of algorithms, but I didn't have much practice in discovering when to use them and applying them to problems.
Exactly - how to find answers to new problems is huge. That's probably the question that makes or breaks interviews I give to new programmers. If they can't tell me how they'll figure something out on there own, yeah not good. In fact, even just being honest and saying you'd not consider this and weren't sure would be fine too. But you'd be surprised how many people how no real answer or scramble something out that doesn't quite cover it.
The funny thing is - esp. depending on how old you are - it is incredible how much easier it is to learn/find answers now. Pre-google it was tough - I frequented Barns & Nobles often to see what new books had arrived, or just to go find an answer in a book I knew they had.
But I digress... google/stackoverflow/etc have all made our lives so much easier.
Baby steps are the way to start, and move forward, though don't think that's what you meant by "small feet". I think it's also important to have realistic, doable goals for "completion" that aren't "100%". This is because you'll learn more by getting feedback ASAP, and... nothing can ever be 100% complete, unless its scope is rigorously perscribed, as for a theorem.
The most important thing I learned in college was accepting/understanding criticism - not algebra or philosophy or how many moons orbit jupiter (all cool/good to know too btw).
Most other skills can be learned on your own. But think about how hard it is to learn to accept and understand criticism. If you receive it from your spouse, it usually causes a fight; or from a close friend or family member - ditto, also not good.
But being hammered in a peer-review style class where it would be somewhat absurd to think 'hey I'm right, my professor with 25+ years experience, not to mention the 15 other smart peers in my class, are all wrong'.
Of course there were always one or two people that really did think they were right and the professor/classmates were all wrong. They wonder why life is always so hard, and why the entire world is against them, never considering that it might be them having the problem, not the rest of us...
He also tends to pontificate on issues he's learned recently and doesn't understand completely a bit too often for my liking.
The other part of this though is that as a blog author, he has a desire and incentive to write about things that his readership might be interested in; so I have a feeling there is a certain amount of extra explaining, citations, and feigned not-understanding going on in his posts.
However, I would point out that formally trained people also have gaps. Just different ones that everyone is used to. The gaps of autodidatics stand out because they are unusual.
Autodidacts in general also have some strengths that are rare with formally trained people. (I don't have anything particular in mind. Just they spend their time with some different focusses and are bound to pick up some things that people who all follow the same curriculum didn't.)
So, yes autodidacts have some gaps, but this shouldn't be seen as a criticism.
I admit I only ever thought the point of this technology was to succeed often enough to be annoying (why use a copyrighted clip if my video might get banned and I have to go fix it).
The problem of matching up arbitrary degraded video to source video in polynomial time seems genuinely hard to me too.
I'm guessing it's a similar problem to matching up degraded audio to song title. If you haven't tried this, the Shazam app on the iPhone does this with surprising reliability (at least from my car speakers).
I recall seeing a paper on how Shazam does their magic (ha!). Something to do with "fingerprinting" the audio in such a way that is resilient to background noise (that also can be indexed and searched very quickly). I imagine that a similar strategy to identify video would also work.
