In the late 1990s I figured roughly even odds of a doomsday catastrophe with nanotech. A mistake with a weapon seems much more likely than a gray-goo accident though. I also think that the risk goes up with the asymmetry of capability in nano; that is the closer to a monopoly on nano that exists, the more likely a doomsday scenario becomes. Multiple strands of development both acts as a deterrent on would be abusers and provides at least some hope of combatting an actual release.