a) Kurzweil has addressed the slow progress in machine intelligence effectively, by arguing that you can ignore most past AI (including his own) and just pay attention to the rate of advancement in simply reverse-engineering the human brain directly. That idea itself can be criticized, but the difficulties of non-neural-emulating AI to date are not really relevant to criticizing his prediction of oncoming sentience. This part of his singularity argument is on the firmest ground, IMHO, as it relies on predictions, not of breakthroughs to previously intractable problems, but of the continued relentless brute force improvements in size and speed of our neural reverse-engineering toolset.

b) Kurzweil’s (and others) vision of the medical fruits of a nano-tech future is the part that’s on the weakest ground. His grasp of the complexity and bite-back potential of human biochemistry is nearly nil, AFAICT, and he would quickly be shed of it if he spent a few years trying to cure most any of the many chronic diseases that have easily bested the most intense research. The discovery of DNA and the mechanics of genes has lured a great many people (including many biologists) into imagining genes act in a deterministic fashion that simply is verifiably far from how they actually act. David Moore (“The Dependent Gene”) makes a fascinating argument that this divided is rooted in the schism between geneticists and embryologists, when a large branch of science said “Aha, it’s the genes — let’s just go tinker with them in fruit flies”, while embryologists said “Aha, look at all the ways the same gene behaves differently in different situations — lets go tinker with them in amphibians” ending up with essentially two branches of science studying the same problems, reaching different conclusions, and not talking to each other. Anybody envisioning a future where we freely tinker with our genes to get predictable results just hasn’t understood what’s already known about the nuts and bolts of how genes work.