Hi dacoinminster, thanks for weighing in. Some notes, below:
1) Bets can be machine readable, but this is a problematic single point of failure, especially if a miner is closing out a huge bet. The website in question could easily be down, or it could be hacked, possibly in anticipation of the bet closing. Still, that's a cool idea.
2) Yep, I read up on that thread, but it requires trusted mediators.
3) I'd like to focus on this. The major idea I wanted to test here is that for prediction markets, mediators or referees are never needed. The problem with (1) and (2) is they introduce weaknesses that Bitcoin doesn't have, and it's precisely Bitcoin's resistance to those weaknesses that make it so great. What I'd like to see is a prediction market that is as strong as Bitcoin itself, thereby expanding the Bitcoin economy from just currency and payment to pretty much of all of finance.
4) As an example of how powerful this is, if we can do a prediction market for BTC/USD with just a next generation bitcoin client, it will be as hard to shut down that market as it is to shut down bitcoin itself. Of course if you're betting on USD going up in value, you won't gain possession of any actual dollars. But that's ok, because determining the exchange rate is the hard part, the part that requires mass coordination of the market. Once we've done that, I can just meet up with some guys at a hackerspace and swap USD cash for bitbills or bitcoins. Or I could do a small bank transfer with someone on #bitcoin-otc. The exchange rate will already be set by the prediction market, and both sides can use it to hedge the exchange if they want to. It will never really be possible to shut down this kind of micro exchange between individuals, there's nobody important to go after, and no great risk if you're caught. The safest, most robust interface between meatspace and the Bitcoin economy is masses of individual people, doing small deals across the boundaries. The heavy lifting gets done either entirely in the Bitcoin economy (btc transfers, prediction markets) or entirely in meatspace (car manufacturing).
5) If this does take off, Government will be so severely crippled by the loss of finance, and finance itself will be so utterly transformed, that the meatspace side will start changing fast, too.
I am so delighted to meet somebody else who is thinking about bitcoin in exactly the same way I do!
Regarding #2, the mediators don't need to be used at all unless there is actually a problem with the bet. I.E. You and I can settle a bet without the mediator signing the transaction at all - it is only in the (presumably rare) case of a dispute that the mediator gets involved. They have no way to steal the money, so you only need to trust them to not be in collusion with one of the betters, which is a much lower bar of trust to clear.
Regarding #4, you are absolutely right. The neat thing is that any bitcoin prediction market that is also traded in real-life (like stocks, bonds, commodities, etc) will automatically stay in sync with the legitimate exchanges due to arbitrage.
Personally, I envision a future bitcoin client (or plug-in) that is a full-featured stock-trading platform which can be used to speculate on any real-life security. The interface would match existing stock trading interfaces as closely as possible (to attract real stock traders to use it), but it would be completely distributed and unstoppable like bitcoin.
