For learning, pick one that solves a problem just for you. For research, pick an existing project and make it efficient. For building a startup, pick one that

For learning, pick one that solves a problem just for you. For research, pick an existing project and make it efficient. For building a startup, pick one that solves a problem for many people and makes money. The last one doesn't come without problems and now has just gotten even harder. It used to be 90% of startups failing. Now it most likely is going to be 98% of startups failing.