This has been done more systematically. For example, for 2009: https://www.lesswrong.com/posts/kbA6T3xpxtko36GgP/assessing-kurzweil-the-results https://www.less

This has been done more systematically. For example, for 2009: https://www.lesswrong.com/posts/kbA6T3xpxtko36GgP/assessing-kurzweil-the-results https://www.lesswrong.com/posts/kbA6T3xpxtko36GgP/assessing-... https://www.lesswrong.com/posts/kK5rabDsKWMkup7gw/kurzweil-s-predictions-good-accuracy-poor-self-calibration https://www.lesswrong.com/posts/kK5rabDsKWMkup7gw/kurzweil-s... For 2019: https://www.lesswrong.com/posts/NcGBmDEe5qXB7dFBF/assessing-kurzweil-predictions-about-2019-the-results https://www.lesswrong.com/posts/NcGBmDEe5qXB7dFBF/assessing-... My own takeaway: Kurzweil's self-assessments are misleading and borderline dishonest; Kurzweil is overoptimistic in general, and you could improve by adjusting by about a decade; there seems to be a split in accuracy to my eyes about hardware vs software - when Kurzweil talks about software such as natural language translation, he is often right or delayed by a relatively short time, but when he talks about hardware (and biology especially) he is strikingly wrong and often the prediction looks like it will never come to pass; and a lot of the software wins were driven by the deep learning revolution: even in 2015, you would have reasonably dismissed some of the 2019 predictions working out, but by the end of 2019...