Colleges don't care what you major in, as long as the dollars flow. Colleges have absolutely no incentive to provide educational experiences that are useful in the working world, since the money is already in the bank account. They have every incentive to maximize enrollments [1] and keep students enrolled as long as possible, while reducing costs via cheap graduate student labor.
Colleges also have every incentive to massage outcomes statistics to keep federal dollars flowing. When a student defaults on a loan, the college doesn't bear the brunt of the impact - it's the responsibility of the government, and therefore, the taxpayer.
"Fixing" the college system, as you suggest, would require eliminating many administrators and/or substantially reducing their salaries, structuring departments based on market demand vs. academic interest (no more liberal arts), and largely eliminating federal subsidies for academic research, student loans, and grants.
You can bet that every single individual involved in credentialing as it presently exists will fight those changes to their dying breath. Since being "strong on education" gets politicians elected, this faction will find it easy to maintain political support.
Personally, I think this sentiment sums up the alternative approach:
"You never change things by fighting the existing reality. To change something, build a new model that makes the existing model obsolete." - Buckminster Fuller
Employers don't NEED people with college educations. They need employees with economically valuable skills. They're not the same thing.
[1] http://www.businessweek.com/magazine/content/10_19/b41770642...
Government guarantees of loans are always a recipe for moral hazard and inefficiency. Look at housing.
Either the government should directly fund higher education, as in Canada, Australia and other countries, or it should end guarantees on education loans from the private sector, so that they can have a more effective market. I'd be fine with the government subsidizing loan rates (say, paying the first 3% of the interest rate.) But the overall rate should be determined by market forces and the risk that a student will not earn enough to pay back their loans.
Borrowing 100K to learn basket-weaving should not have the same implicit risk as borrowing 100K to learn software development. As long as there is, you're effectively subsidizing the basket weaving at everyone else's expense.
Also, the solution suggested by the OP doesn't scale. There aren't remotely enough available internship opportunities to handle the demand if a large percentage of college students went this route, nor would big companies want to spend their time attempting to filter what would essentially be high school graduates looking to work in the corporate space.
Sadly, most of the "value" of college today is due to the failure of high schools. If you require a basic level of literacy, numeracy, computer literacy, etc. you have to set your level at college graduates, since a HS diploma no longer guarantees such skills.
It's great to say "don't go to college" but the reality is that most employers NEED people with college educations. It's unlikely that any amount of mentoring it going to create the technical employees that our businesses are desperate for right now. Yes, a lot of kids are wasting their college education and not ending up with marketable skills.
A big reason for that is our loan system that makes it easy to get a tuition loan for a field you'll probably never make a living in, because the party making the loan doesn't have to worry about a default.
If the effective cost of an engineering degree was half that of a degree with fewer job opportunities, because there was real risk of default and that was priced into the interest rate, we'd see a much better hiring market for new college grads than we do right now. Loaning someone $100K so they can get a photography degree at NYU and make $40K a year (as per a recent NYT article) suggests that the market is not functioning properly, and it's obvious why.