compute based on physical movement of water

[dragon]

kind of cool

After more than a decade of research, scientists at Stanford University have created a working computer based on the physical movement of water droplets. It’s a breakthrough in physical computing that gets at the most basic definition of a computer: any programmable device that can carry about logical (mathematical) operations. By combining cutting-edge theory in fluid dynamics with very-much-not-cutting-edge theory in computing, the team was able to create a synchronous computer based entirely on the physics of water.

As you might imagine, a computer based on the physical movement of water is much, much slower than a conventional computer based on the movement of electrons — but that’s beside the point. Nobody expects a new, super-fast liquid CPU, but by applying the principles of computing to the manipulation of matter, lead researcher Manu Prakash and his graduate students hope they can computationally revolutionize other areas of science

.

link

[Purple]

If you say so. Personally I don't really see what they were trying to achieve with this. I mean, sure it's cool that someone went and did it. Just like it was cool back when someone actually made a Turing complete computer in Dwarf Fortress using pressure plates. But I don't really see any scientific merit in this. Could someone here enlighten me?

[salm]

Apparently it can turn a workload of months into minutes for some sort of chemical testing. So while the Darf Fortress Turing computer might be cooler this seems to be more useful.

[Simon_Jester]

Basically the idea is that the water computer can do things that it is physically impossible to do with a computer based on binary logic. The same idea is behind quantum computing, really.

[Zeropoint]

Science is like sex: sometimes something useful comes out, but that is not the reason we are doing it.

[Starglider]

Basically the idea is that the water computer can do things that it is physically impossible to do with a computer based on binary logic. The same idea is behind quantum computing, really.

No, not really. Quantum computers still solve computational problems expressed in discrete logic; the implementation is different but the input, ouput and basic operations are the same as for cpnventional digital computers. In this case the proposal is to run the control program for a microscale chemical test sequence using magnetic manipulation of droplets in a labryinth rather than MEMS controlled by electronics. The main benefit is that it is much cheaper to make than a MEMS chip. The input and ouputs are drops of chemicals, with a separate sensor and analysis software needed to extract the results.

[Channel72]

Basically the idea is that the water computer can do things that it is physically impossible to do with a computer based on binary logic. The same idea is behind quantum computing, really.

No, not really. Quantum computers still solve computational problems expressed in discrete logic; the implementation is different but the input, ouput and basic operations are the same as for cpnventional digital computers.

I think Simon's statement is more or less correct - if we interpret "physically impossible" as referring to running time, rather than functionality. The superposition of qubits in a quantum computer does allow for faster algorithms - and not just "faster" in the sense of brute processing power, but actually Big O faster. Like Grover's algorithm, which searches a set of unsorted objects in faster than linear time, which is really amazing considering that a classical computer would always require linear time. So if we have an unsorted array of 1 million objects, a classical computer will always take at worst, 1 million checks to find the search key. But using Grover's algorithm, a quantum computer will take at worst sqrt(1e6) = 1,000 checks, which is pretty fucking amazing. Although, the downside is that Grover's algorithm is probabalistic, like many quantum algorithms, meaning it may need to be run more than once to have absolute confidence in the result. Still that's way better than O(N), especially consider that it gets the correct answer with high probability, so for absolute confidence it should only need to be run a few times.

Still, it is physically impossible, as far as we know, to get sub-linear running time for searching an unsorted array of objects using a classical computer, so in that sense I would say Simon's statement is correct.