Surely after a while security of the network stops becoming an influencing factor. There reaches a threshold where you can clearly say 'the system is secure'. From what I can gather this is at some point where 50% or more of the computational power of the network is not within the control of a single user? I don't know where this point is, I'm sure this figure is completely wrong, but equally I'm sure the point was passed ages ago when bitcoin was in its infancy, when security was still a legitimate concern.
The 50% figure is correct for certain types of attack, particularly ones involving modifying the block chain.
There are other attacks which do not rely on controlling a large portion of the network. The Finney Attack can be carried out by a single user. The basic attack is:
- have some Bitcoins on hand.
- Create a transaction to send those Bitcoins to yourself, but don't publish the transaction
- Start mining, with the above transaction in the block.
- When you find the solution, then immediately create a NEW transaction, sending the same Bitcoins to some unsuspecting victim, who acknowledges payment and sends you whatever it was you paid for
- Publish your block
If your timing is right, you have just successfully double-spent your coins. The only known defense against that attack is to keep the difficulty of creating a block containing arbitrary data as high as possible, requiring the attacker to expend a lot of effort and time. In order to make the attack worthwhile, the attacker will select high-value payments, which (presumably) the recipients will check more carefully before acknowledging.
By this I mean, the security of the system increases as more and more user mine bitcoin but the advantage to security of having 100 million unique miners is, in truth, of no real difference to having just 1 million.
I doubt we'll even get to 1 million miners. As the system matures the goal is to have the majority of users running a "light" client, which does not mine. In fact, right now the option to turn on mining is being removed from the default client - there's no point to mining unless you have a GPU setup. The gold rush has peaked - at this point, only established miners and really determined newcomers will likely set up shop and continue mining.
I'm guessing this [protein folding's centralization] doesn't work for something decentralised like bitcoin.
Ah, yes - that's another consideration. You're probably right.
I don't claim to have a solution of how this surplus processing power could be better utilised, but at the moment it seems a tiny bit crazy that all those processor cycles are doing nothing more than solving puzzles of no consequence.
Then you still have not grasped the fundamentals of Bitcoin. The puzzle has actual dollar value. Solving a puzzle will earn you 50 Bitcoins, which according to Mt. Gox you can redeem right now for about $80. That hardly seems "no consequence" to me!