I know scrypt was designed to lessen the GPU/ASIC advantage.
We now have litecoin as a real-world example of this. However, it hasn't worked out perfectly. Most coins are mined by GPUs, although the advantage is probably smaller than it would have been with SHA-256.
I wonder how the future for scrypt and newer generations of this approach looks like:
I believe scrypt is designed to be scaled to be both memory hard and cpu hard by explicit design.
As a result, most implementations of scrypt will require you to input a desired cpu difficulty as well as a desired memory difficulty. On a gpu, where each core has very little memory, it would be trivial to see that, unless gpus gain a great deal of memory in the near future, you can crank the memory difficulty up to the point where it will be barely tolerable for a cpu let alone a memory-light gpu core.
Simply, scrypt doesn't need to be improved, it just needs to be utilized in a manner that will achieve the desired results. Even if you look now, you will see that scrypt is doing a superb job at preventing a serious rise of hashing rate with the advent of the ability to use gpus as the rate is currently measured in kh/s per second rather than mh/s on the same devices.
