This is one of those rare books that strikes the right balance between being choked full of fascinating information, and not being over the head of a non-specialists. It has a tremendous breadth of coverage, and I would absolutely recommend it first (before anything by Sachs or Easterly, for example) for those interested in development economics. (It's not a bad read for economists either.)
For those who don't know: there's a longstanding feud between Sachs and Easterly--who sit at opposite ends of Manhattan at Columbia and NYU respectively--over, among other things, whether giving more aid to poor countries actually does any good, with Sachs arguing that it does, and Easterly basically arguing that if you don't have good "institutions"--which no one ever quite fully defines--nothing is going to help. Banerjee and and Duflo, at MIT, are trying to move the discipline beyond this old argument, and I would say largely succeed in this book. They do this by focusing on data driven results, especially experimental results, which are very rare in much of economics, but are becoming more and more the norm in development since there's a good deal of donor money and projects in poor countries can be remarkably inexpensive. So, for example, there's this really old irritating argument over whether giving away mosquito bednets, as opposed to selling them cheaply, actually leads to less usage because people don't value them. Well, someone finally actually did the experiment, and found that people who are given bednets mostly do actually use them, though they may take extras and waste them, and selling them really cheap works pretty well too. This is what development economists spend their time on.
But there are many more interesting facts to be learned from this book. For example, hunger apparently isn't a problem almost anywhere in the world, though a few specific spots in sub-Saharan Africa may be exceptions. But in most places, if you give people more money to be spent on food, they don't end up eating any more calories; they just eat nicer food. On the other hand, poor nutrition among children and pregnant women may be an issue with serious long-term costs. Community scale drinking water systems may be one of the most effective ways of preventing illness. Microfinance doesn't hurt the poor, but it doesn't seem to help all that much either. Insurance schemes for the poor may seem like a nice idea, but are very hard to implement and are often resisted by those they're intended to help.
My main quibble: This new approach to development is inherently microeconomic (as opposed to macro) in nature. You can't really do macroeconomic experiments where you transform one country's economy and not another. Which doesn't mean macro issues aren't discussed at all--there's a very long discourse on how poverty traps, essentially a macro idea, are to be understood at the micro level. But some of the big ideas in development are inherently macro in nature. One book can't do everything, but since I really do think this should be the first book non-specialists read, I would have liked the authors to summarize some of the other perspectives on the field a little more/better.