Continuing the conversation with Daniel,
» Lectures are not where you learn, I agree. I missed as many lectures as I attended as an undergrad. Grad school is a different story as lectures tend to be more conversational and informative and the material more challenging.
» A lot of the smartest professors are terrible lecturers. Another little story: one of my grad school professor's lecture consisted of standing in silence, staring at the class in misery, writing an equation on the board, staring at it, turning back and staring at the class, and saying "there". This was the normal session.
» The best pre-recorded lecturers are likely better than your average live lecture. But most learning takes place outside the lecture.
» Being there physically is still very important. A study at Google found physical proximity is the single highest determinant of people voting the same way, which is interpreted as a good measure of information flow. I'm a big believer in telepresence, but there is no way telepresence is a substitute for physically attending the same university as someone else. At one point in undergrad my friend and I spent 3 days physically locked up in a lab creating a visual perception system, on a whim. No way that would've happened over telepresence. Eating tacos at 3am with your cohorts is an integral part of the experience. This is why Paul Graham (since we're quoting Paul ) forces the startups he funds to move to where he is.
» Chance interactions with people due to physical proximity carry as much or more information as pre-scheduled meetings. The same professor who would stare at the board in his lectures was incredibly insightful and helpful when I asked for his help as I ran into him in the hallway of the CS department. So were any number of my classmates that co-habitated that dungeon of a building.
» Good universities are filters for smarts and achievement. The level of drive and ability you'll find at Berkeley is certainly different than what you'll find at your local community college. The average level of drive and ability falls off from the top schools to the lesser schools. Would you want to spend 4+ years with the smartest, most driven people you could find, or with less smart, less driven people?
» Smart people don't want to talk to you unless you pass the filters. Daniel mentions: "… you could just go on the Web and interact with super-smart people from all over the world without having to pay Harvard-level tuitions" and that bandwidth and latency are the primary barriers. They're not. Part of what you pay for when you pay Harvard tuition is the privilege of talking to very smart people. Those same smart people wouldn't talk to you if you were some random person who hadn't passed a smarts filter because you'd likely be a waste of their time. Daniel is a professor at a well known school and as such can call on smart people as he likes; try doing the same as an undergrad in a community college.
» Expectations are important. What's expected of the average student at a name school is quite different than what's expected of a student at a community college. Again it's a gradient. I firmly believe what you expect to achieve is a key indicator of what you will achieve.
For all these reasons I think you should go the "best" school you can get into, physically be there, and live in the dorms (I commuted). This doesn't mean you can't achieve everything you want going to a "lesser" school, but you'll have to work harder at it.