Ask HN: Which are good computer science problems to work on right now?
I'm a CS undergrad and to finish my graduation I have to produce a term paper.
Considering that I have one year to work on it, it can't be something huge, but I would like to work on something which would be useful to someone else and not only a boring paper about a very specific and almost useless subject.
As I have a web development background (Ruby on Rails, Node, Redis...), anything in this area would be easier. I've tried to revise my workflow in order to find some gap but, probably due to a lack of experience (and an awesome community), everything seems to be created already.
Do you have any suggestions about problems that I could solve with the skills that I have? A very personal itch also counts!
Note: The final software must be open sourced! * User interfaces for VR helmets. * client/server web frameworks based on BigPipe-like (Facebook) architecture * turning real world problems into MMOGs in order to crowdsource solutions * peer-to-peer web infrastructure over multiple networks/communication channels (to counter surveillance, ISP filtering, natural disasters...) I like the idea about real world problems. Can you think of any examples? A well-known example is data collection about pedestrian paths, altitudes, WiFi hotspots using a game like Ingress. Those aren't for VR though, right? In my opinion, automatic summarization by abstraction. There are lots of automatic summarization algorithm that does extraction like TextTeaser (https://github.com/MojoJolo/textteaser). I think they do the job well but it doesn't have a realistic feel in it. To achieve it, I think automatic summarization by abstraction is the way to go. It almost mimics how humans summarized text. But I think there is still no algorithm that can really perform abstraction well. I'm looking to try it in the near future. With a good training data, I'm looking into using genetic algorithm eventually evolving the text into an almost human summary. I'm excited for this! Explore how you could build a web tool to do data visualization of otherwise annoying data to visualize. Some of the work by the data vis lab at Stanford (now UWash) is pretty awesome - http://vis.stanford.edu/ There is still plenty of work to do in Natural Language Processing. Also, check out Wikipedia's "List of Unsolved Problems in Computer Science": https://en.wikipedia.org/wiki/List_of_unsolved_problems_in_c... Decoupling business logic from web frameworks might be an interesting architectural problem that could use better tooling. Here's Jim Weirich going about it manually in Rails: A genuine question... Wouldn't decoupled business logic generally fall under 'workflow' systems though? But, unless you refer specifically to embedding algorithm in the code, I would assume the language / platform selection would matter. Speech Recognition, it's a hard problem to solve - but maybe with your time you can offer some interesting solutions. Text IMO is very well covered part of our field, but speech is still very much lacking. Even image recognition has more attention than speech recognition IMO. Speech recognition is an important problem, but if the problem solver does not have an angle of attack, it does not make sense to try and solve it just yet. To quote Richard Hamming from his lecture "You and your research", "It's not the consequence that makes a problem important, it is that you have a reasonable attack." It's true we should mostly focus on the easy problems... But some of our time especially IMO when we are young should be spent on the hard (unsolvable) problems. It's neat that your project is required to be open source. I've been thinking about my own too and am really interested in AI.. So with that bias in mind and your own, how about automated A/B testing using genetic algorithms? Have you thought about developing some innovative algorithms? Something in the data analysis space, text analysis tools, graphing tools ? If you are interested we could help diego at algorithmia dot com. Computer vision and machine learning (especially deep learning if you have necessary foundation).