"The only sure thing about forecasts is that they are WRONG" - James P. Womack and Daniel T. Jones.
Estimates have been the bane of software development and programmers for decades. Managers/Customers want to know: When will it be done? How much will it cost? Programmers are told "We won't hold you to the estimate", and yet they often are.
Estimates in themselves are not the problem I see, but rather the dependence on following an estimate-driven approach to software development. For many years we've depended on estimates to inform our decision making process, and yet the results have often been mediocre at best. It's my contention that estimates are often not as useful as we seem to think, and even worse they misinform the decisions they are meant to support. Do we really need estimates? Is simply "getting better" at estimates worthwhile? Can we live without them? Will things be better without them?
Estimates are merely one possible approach to software development, and I suggest we must be open to discussing the possible problems, and to search for alternatives. I don't necessarily have answers for you, but I've worked with "no estimates" for over 6 years and I'm still alive. I want to explore the idea of estimates, why they are pervasive in the programming world, how they might be harmful, and see if we can grow the dialogue about finding a better way to make decisions.