Prediction markets are a form of gambling, but there is a key difference; whilst the purpose of traditional gambling markets is entertainment, the purpose of a prediction market is information discovery.
Some prediction markets can even be subsidised in order to incentivise participants, where the house (ie the entity running the market) expects to lose money in exchange for the information that the market is revealing. It is extremely rare to find this in a traditional gambling setting, where the house usually always reserves an edge.
This fact seems to have been largely missed by the guys behind Augur though, who seem to have focused more on the "blockchain" part of "prediction markets on the blockchain" than the "prediction market" part, and this shows through their clunky UX and relatively simple platform. Prediction markets have been around for decades (30 years ago the University of Iowa ran a market on the 1988 US election and is still running them today - https://iemweb.biz.uiowa.edu) and Augur doesn't seem to offer anything new above decentralisation, which I dont think anyone was asking for.
"The IEM is operated for research and teaching purposes. All interested participants world-wide can trade in our political markets. Other markets--such as the earnings and returns markets--are open only to academic traders."
"Trading accounts can be opened for $5 to $500."
These constraints are presumably why they got their no-action letter from the CFTC. If there's demand for markets other than political markets, or for larger accounts, then there's demand for a decentralized prediction market.
You seem to be suggesting that the point of decentralisation is to avoid legislation and regulation. While this might be true for Silk Road and other criminal transactions, that does not seem to me to be a valid argument for running a prediction market on blockchain software.
Blockchain as a technology has a valid use-case where you cannot trust a third party to manage a centralised database and where you need an immutable history of events. I can just about buy into putting transactions for real-world things (eg property) onto a blockchain - after all its a lot easier to prove you own something in the case of a dispute if you can show on a distributed and trusted event log that you were the last person to purchase it than it is to dig out some old documentation and prove its not forged.
There is an argument that applies to prediction markets, which is that you might not trust the market to actually pay out your winning contracts (though believing in this relies on a distrust of the regulators and court system). Augur is not actually solving this problem though, as neither oracle contracts nor the market owner defining the settlement conditions are trustworthy, and when you have user-defined markets such as "Does God Exist?" there is huge scope for arguing about the true outcome. At least with the centralised case you have an organisation with a reputation to maintain, that is therefore motivated to police and administrate the markets properly.
You might also argue for long-running markets that the company running the market might collapse and be unable to pay out, but in the case of crypto there is simply a parallel that your currency could simply collapse and be worthless by the time your contract paid out. Given the volatility of most ETH and REP, your winnings are likely to either be dwarfed by gains in the currency or destroyed by a collapse in the currency.
Prediction markets that trade using real money are gambling, and hence are controlled by various regulators. In addition, certain markets can also catch the attention of other regulators, for example if a market involves speculation on financial instruments or commodities. There are certain ways to avoid this, such as restructuring the incentives; for example, a market using credits instead of real money and run as a competition where the most successful players receive monetary prizes does not constitute gambling. Obviously in this case the players are not risking their own money, and this relates back to my previous points about subsidising prediction markets in exchange for valuable information. The regulation issues should be dealt with in an ethical way by developing the technology and working with regulators and not by avoiding the issue via decentralised blockchains.
Actually Augur does attempt to solve the trust problem.
Regarding settlement conditions, there's an allowed outcome of "null," where the judges can decide that the outcome can't be determined. In that case all bettors get their money back.
There's no oracle. Augur's main innovation is its decentralized resolution system. Initially a small number of randomly-chosen REP holders vote on the outcome. That outcome can be disputed, which escalates to a larger number of REP holders.
This goes through several rounds. Disputers have to post bonds to make a dispute, which they lose if the dispute fails.
If they're determined, they can escalate to the ultimate resolution: Augur splits in two, one with each outcome. At this point every REP holder has voted one way or the other, and each holder only gets REP on the version he voted with. The theory is that the version of Augur which reports the truth is the one where REP will have the most value. This gives REP holders an incentive to vote for the true outcome.
I may be wrong on some details, so check their whitepaper for more, but the general idea is to make the incentive strong enough so most REP holders will vote for the truth and disputes will happen rarely, mostly when small groups of REP holders actually get it wrong. Whether it works in the real world, we'll have to see, but I think it's an interesting experiment at least.
Some prediction markets can even be subsidised in order to incentivise participants, where the house (ie the entity running the market) expects to lose money in exchange for the information that the market is revealing. It is extremely rare to find this in a traditional gambling setting, where the house usually always reserves an edge.
This fact seems to have been largely missed by the guys behind Augur though, who seem to have focused more on the "blockchain" part of "prediction markets on the blockchain" than the "prediction market" part, and this shows through their clunky UX and relatively simple platform. Prediction markets have been around for decades (30 years ago the University of Iowa ran a market on the 1988 US election and is still running them today - https://iemweb.biz.uiowa.edu) and Augur doesn't seem to offer anything new above decentralisation, which I dont think anyone was asking for.