Maybe the problem is trying to understand this in terms of confidence and not in terms of accuracy. In fact, we might say that some are underconfident, some are overconfident but almost all are inaccurate in predicting the form and/or details of progress. In that case we don't have to care about the technologists who thought the recent wave of image generation would take 10+ years AND we don't have to care about the people who think evil AI will bury us under paperclips in the next 30.