Hacker News new | ask | show | jobs
by guest2143 2836 days ago
My grad school professor was at a bar with a consultant friend who was trying to find the optimal wiring paths for routing expensive electric cables.

My professor asked for the change in his pocket: they went to the hardware store and got plexiglass, dowel and drill. In the hotel room, the put together a model of the points to connect between the plexiglass sides and dunked it in the soapy hotel tub.

the pattern the soap bubbles formed was the shortest path between the electrical tower model.

2 comments

I know this as the Euclidean Steiner Tree problem: https://en.wikipedia.org/wiki/Steiner_tree_problem

There are some videos doing exactly what you described: https://youtu.be/dAyDi1aa40E?t=95

Okay, that makes sense to me—we can set up the physical system's initial conditions to mirror some problem we want solved, then let it do its thing and grab the solution back afterward.

That definitely feels better justified than the view I had of it, though it still has some weakness, I think.

From my understanding the justification for calling it computation comes down to: we can use the physical system to solve a problem that we would ordinarily approach computationally, therefore it must be computing too. Right?

> From my understanding the justification for calling it computation comes down to: we can use the physical system to solve a problem that we would ordinarily approach computationally, therefore it must be computing too. Right?

To me, it feels like the intention of "computation" as used here is to give us a convenient way to describe what the physical system _appears_ to be doing (solving something the only way we know how--computationally) as opposed to saying: the rules of our universe influenced these particles to act in such a way that manifested in something interesting to us (an efficient solution to some problem).

Very interesting to think then if we can leverage physical systems (specifically non-quantum, to make that more interesting) to solve NP problems.

It's an analog computer, just not a general purpose one. It solves one problem well.