Hackers pay a reasonable fee ($10k I imagine, or less) for a year of accommodation, cooked meals and other basic luxuries. And fast internet access. They being their own computer,
And then work on whatever they like, for how many hours a week they like, with whoever they like.
The idea is programmers could use this a 'working sebatical' with no pressure to make money, or the usual living cost pressure in western countries.
They could get a start up going, or contribute to open source, or just learn lots of cool technologies, or whatever. They can escape Enterprise development, WTFs and tight deadlines and fall in love with programming again. And maybe make a bit of money or start a business in the process.
The $5m is to set up the building with accommodation, office space, place to eat etc.
For inspiration, listen to this story about the development of the birth control pill, and how it happened only because a wealthy donor made it possible. Talk about changing the world.
http://www.npr.org/blogs/health/2014/10/07/354103536/the-gre...
(I'm assuming the 5 Mill is sustainable. If not, I'd want to invest it to protect from inflation and give away an amount each year to keep it 'non-profit')
And the long shots are the ones that have a chance of being awesome.
I'd try to put together a team to tackle the incongruity of the following two facts: (1) 1/3 of healthcare dollars are spent in the last few months of life, and (2) physicians and other health care workers do not choose this level of care for themselves. Would a multi-disciplinary team be able to educate people about how bad things work out when you try too hard late in life? This is about a trillion dollar problem, even a 1% decrease would be a huge net benefit to the US population.
If I more time, I would probably explore other inexpensive, but potentially very high payoff ideas, like funding a grad student to look into practical ways of making blockchains work for auditable transactions (like smartphone based voting documentation and other government transparency).
Another option, a free time-lock cryptography server. (E.G. release public keys whenever, private keys at predetermined times. Lots of utility in this space, but also government interest - If you are really serious, I have some more ideas, just not so much time right this second...)