Ahh, others might be beating me to it, but would love to help bring low-cost computing tech to developing parts of the world. It's mind-boggling that there are $40-50 (in bulk/wholesale) Android tablets--not necessarily up to snuff with the spiffiest rich-world toys (or even the Kindle Fire), but actual computing devices nonetheless. It would would take time just to figure out what the key niches are where they could be useful (do folks need/want crop price data? weather info? news? wikipedia? Oscars coverage? something else?), what technical work has to be done to get there (connectivity, software, content, and maybe different kinds of devices, e.g., maybe the ideal device for some folks is e-ink-based and low-power like a Nook/Kindle), and all kinds of distribution/operations stuff.
And it could be a great business: being the first folks to get things right for this huge group of people is going to be a big deal as that chunk of the world moves up the development/economic ladder, one hopes.
I can't, personally, do very much of that in 5 years but had I capital and all-purpose moxie, there's the problem space I'd love to tackle.
I think there are big things to do in genomics, GWAS, medical data, etc. I don't really know what they are. I could go back to school for that; that might be the most interesting "hard tech" possibility.
No lust for this personally because it's too close to my real job, but I think too much of the effort around databases today is too focused on the lower layers of the stack--we have lots of scalable DB products but too little good software to stitch everything together (from a client-side cache to scale-out OLTP to memcache to data-warehouse-y stuff, ideally) and take the repetitive bits out of setting up a full stack and building a a minimal but modern UI.
We don't need 2013's BigTable, in other words, we need 2013's FileMaker. It's 2013, so it needs to be scalable and Web-based and not too drudgy either when you're either starting out or making a 'real,' heavily customized product. I'd want to offer code you can run on your boxen, not a service-only thing. If I had five years, even with help I could only probably attack a tiny slice; some kind of common API atop various datastores (client-side, memcache, Hadoopish, etc.), and some kind of Web form/page bindings that don't suck (allow modern UI patterns and are extensible) would be a couple interesting ones.
Re your tablet idea: I have thought it would be neat to give tablets to homeless Americans and give them a class on how to use it, not just technically but practically. Help them find and install apps. Give them free games. Outline things that can be done to improve their lives in the here and now, like freelance websites that can bring in a little extra money.
And it could be a great business: being the first folks to get things right for this huge group of people is going to be a big deal as that chunk of the world moves up the development/economic ladder, one hopes.
I can't, personally, do very much of that in 5 years but had I capital and all-purpose moxie, there's the problem space I'd love to tackle.
I think there are big things to do in genomics, GWAS, medical data, etc. I don't really know what they are. I could go back to school for that; that might be the most interesting "hard tech" possibility.
No lust for this personally because it's too close to my real job, but I think too much of the effort around databases today is too focused on the lower layers of the stack--we have lots of scalable DB products but too little good software to stitch everything together (from a client-side cache to scale-out OLTP to memcache to data-warehouse-y stuff, ideally) and take the repetitive bits out of setting up a full stack and building a a minimal but modern UI.
We don't need 2013's BigTable, in other words, we need 2013's FileMaker. It's 2013, so it needs to be scalable and Web-based and not too drudgy either when you're either starting out or making a 'real,' heavily customized product. I'd want to offer code you can run on your boxen, not a service-only thing. If I had five years, even with help I could only probably attack a tiny slice; some kind of common API atop various datastores (client-side, memcache, Hadoopish, etc.), and some kind of Web form/page bindings that don't suck (allow modern UI patterns and are extensible) would be a couple interesting ones.