Bitcoin has to use something to distribute the incentive. "Satoshi" chose Hash/s, and that's pretty reasonable. If you wanted to eliminated GPUs from the equation, you could probably write a Bitcoin clone (PersonCoin) that rewards based on an updated list of cost of resource (hash, storage, bandwidth etc.) that you believe optimised for an individual, and charge a tax for updating the list of standards.
"Satoshi" chose Hash/s because he is a geek, specialist in mathematics only.
Bitcoin society needs specialists in psychology too.
Normalization by just Hash/s is WRONG. Although, it could spark, if there no ATI GPUs on the market. but they are.
If "Satoshi" were studying psychology, he normalized the algorithm, to allow 1 000 000 miners come into field, with no dependance of power of ATI GPUs.
Compare with real gold: individual gold-digger finds an instrument, that increases his individual gold mining in 100 times - from 7 Mhash/s to 700 Mhash/s - on the time interval of ONE day - time to buy ATI GPU and install phoenix-like software.
Only ONE man developed that software, say, for months & years. But ONCE the software is available for free download, each average Joe can increase his real gold mining power up to 100 times in ONE day.
Just think by head & compute numbers.
Now, I have understood, why,
difficulty is being increased on time interval FEW days! It's JUST average time for average Joe to buy ATI GPI, download free software, and start mining.
And sure, TODAY you can see that number of miners increases too. BUT, this process will continue for FEW days (weeks).
Then, mining will fall down, because of almost zero profit AND unabling to output large amounts of $.
And the bitcoin society will be frozen in some 10-20 000 (ok, let 50 000) miners & acceptors, on the time interfal FEW weeks.
This is mathematical theorem according to current difficulty algorithm!