Human civilization has done nothing but evolve in complexity over time. Why would this trend cease? It won't. Mole-men sustaining life in underground bunkers developing CERN portals to parallel dimensions is slightly more complex than President Garfield's medical treatment for bullet extraction/infection prevention just a century ago.
Smart contracts will ONLY INCREASE IN COMPLEXITY over time. Just look at history (then extrapolate the future).
The point is that a smart contract is a totally different beast than "software". A smart contract, like a normal contract, is stated once and for all, when the first two parties sign up ; in as much as it can "auto-mutate", this is part of the original contract. There's nothing *unforeseen* that you can change in a smart contract.
In fact, the closest you come to the situation of a smart contract, is firmware running on a space probe you cannot update remotely. It has to work at launch, and can never be updated any more.
Now, look at normal contracts. Normal contracts can be quite complex when you look at the paper work, but their state tree is in fact relatively simple. The hard part of normal contracts is how the legal system will interpret the state tree: this is why you use expensive business lawyers to try and verify normal contracts. A wild guess is that even a very complicated normal contract doesn't contain more than a few tens of leaves in its state tree ("if this, then so, and if so then this, but if not no, then that"). NO SINGLE NORMAL CONTRACT HAS EVER USED UNSPECIFIED LOOPS OR UNLIMITED RECURSION. So no normal contract has ever needed a Turing-complete description language. It is this absence of Turing completeness which allows to analyse the state TREE (and not a state GRAPH).
If you cannot formally make sure that the state tree is what you consider it should be, then you are using "frozen software" of which you have no formal proof that it doesn't contain bugs or exploits (if you have a formal proof, then these things cannot occur, because you have explored the full state space of your contract).
The holy grail of software development is to know how to write bug free and exploit free software in Turing complete languages, and in more than 50 years of software engineering, we never succeeded in doing so. But in non-Turing complete systems, one can, with formal state tree verification. We even know that this proof is not possible in Turing complete languages (the Halting Problem).
So, writing a contract, that cannot only never be altered, but that will run open to anyone, containing potentially a lot of money (bait for exploits), on a system where it is mathematically impossible to prove the absence of bugs or exploits, is NUTS. Especially, because that aspect which renders the system unprovable (Turing completeness) is never used in real normal contracts.
smartcoins - the bitcoin derivitives that can be executed either now (like bitcoin) or sometime in the future making wills and annuities possible. Translation - you don't have to trust a bank or third party to trade assets, or administer an annuity to your after you die, and they turn 18
But these are SIMPLE, small contracts with just a few leaves on the state tree. You do not need Turing completeness to implement that. In the same way that the statements in the "paper" version of these things also doesn't need any Turing completeness, unlimited recursion and all that.
PeerPlays = gaming smart contracts - you don't have to trust Joe Schmoe to hold the fantasy football prize money becasue the network will not release no money until the network reaches consensus on the final scores.
This is again a simple contract with just a few states.
So it's not that smart contracts are not viable, and will disappear, it's just that COMPLEX smart contract platforms require a greater level of human trust than SIMPLER smart contract platforms like bitcoin.
These are NOT "complex" smart contracts. They are very simple ones.
