How is a uniform sample generated by the computer? I know that uniform samples used on inverse cdf's can be used to generate normal samples. I was wondering how a technique would generate a uniform sample!?
 A: By uniform sample, I think you mean uniform distribution. 
The most common method is to use a pseudorandom number generator (pRNG). Alternatives include getting random bits from a physical source of noise or from a site like random.org.
Typically, the pRNGs are seeded from the clock. They are usually known to be quite close to uniform. Problems include that the numbers generated are not independent of each other, and sometimes exhibit very clear patterns in their joint distributions. Poor implementations may restart the pRNG while the clock is on the same millisecond (since computers are fast) which can result in repetitions. 

By the way, to generate normally distributed random variables I recommend the Box-Mueller transform instead of using the inverse CDF. 
A: There are a number of methods called pseudo random number generators.  The user provides a seed number and the pseudo random number generator gives a formula for calculating a random number based on the seed.  The same formula is then applied to the new random number to get the next one.  One common method is the linear congruential generator.  The generator is called pseudo random because the sequence of random numbers is actually a deterministic sequence but the numbers vary in a way that they look like independent sequence of uniform numbers.  The following link provides detailed information.
pseudo random number generators
