Probably not a lot, paper is cheap. Typically a concept like this is explored by a team of 6-10 engineers over the course of 3-6 months, and they may work on multiple projects simultaneously.
Even if the project doesn't move past the concept phase, there can still be developments along the way that make it into other projects. Further, if a concept is abandoned because it is not technically feasible, identifying that it can't be done also identifies the limits of what currently can be done, which is good for determining research goals.
The cost of such a project should also be compared to the alternative: it's much cheaper to work through a design on paper and finding it's a dead end than to commit to something without that evaluation and find out 80% of the way through building it that it's a dead end.
> i wonder how much money gets wasted on these failed explorations/projects.
Failure is not a waste of money if lessons are learned. The value is in the lesson. As anyone whosed developed a reinforcement learning application knows, there is a balance between EXPLORATION and EXPLOITATION. If you don't explore, your performance will never improve. This is true even if 99% of the explored states are failures.
I suppose if one doesn't want to invest in exploration, one could always join an Amish township. They are very content and apparently happy with their perpetually static state of technological development - and there's nothing wrong with that until someone with a missile comes in to take over your land amd all you have to fend them off with is a hay fork.
This is the key statement. Even in reinforcement learning exploring intelligently is an important problem. There’s no point exploring a new state if your world model is confident about what’s going to happen there: you won’t learn much.
having worked for multiple dod contractors that did prototypes, lots.
Sad part is lots of the good ideas get culled when some head honcho doesn't like the dpt. it came from, or where the funding came from, because politics.
Lot of warfighters/ppl who are actually going to use the systems/products give us good feedback and then some O-6 comes on and doesn't like it or doesn't understand it.
