Over the past week I read Ray Kurzweil's new book "The Singularity Is Nearer". This is my review.

First off, if you've read The Singularity Is Near, which was published 19 years ago in 2005, you should be aware that the sequel book is a lot less technical. The Singularity Is Near contained a couple dozen graphs explaining the exponential improvement in computers in computer engineering terminology. This was good at getting the point across but made some of the details of the book difficult to understand if you didn't know this terminology well. The Singularity Is Nearer is much lighter on this, which is probably because less of that is needed nowadays to get the point across. The book is more accessible to non-computer nerds. Which is likely due to the fact that nowadays it's easier to convince people who are not computer nerds that the singularity is near.

I think the most important prediction that Ray Kurzweil made in The Singularity Is Nearer is that we will have Artificial General Intelligence by 2029, meaning that by 2029 we will have machines capable of doing any cognitive task that an intelligent human can do. Ray explains that he first made this prediction in 1999, and everyone thought it was crazy at the time, but he has kept consistent with this prediction, and in recent years experts have started to agree with him, with the consensus on Metacalculus having reached that AGI will happen by 2026 in May of 2022, making Ray Kurzweil now in the conservative camp.

Ray goes onto argue that AI will be used to automate the production of food, housing, medicine, transportation, and clothing, making all of our necessities very cheap. And he argues that medicine will get qualitatively better, and that we will reach Longevity Escape Velocity by the 2030s (with the more diligent / wealthy /  and well-connected among us reaching it by 2030 on the dot). He also argues that the technology needed to automate a lot of work is already here and that we're already in the cultural latency phase of waiting for technology that already exists to replace workers, that there will be an increasing shortage of jobs in the near future, and that Universal Basic Income is not just a good idea but is inevitable and that all developed countries will adopt it by the early 2030s and all developing countries will have it by the late 2030s. He argues fairly convincingly that UBI is more politically realistic than it might sound to cynics concerned about a right-wing backlash to it, because the amount of money that the USA government spends on the social safety net has consistently increased each decade regardless of whether or not left-wing or right-wing politicians are in charge.

Ray Kurzweil also paints a picture of where healthcare technology is going, pointing out that AI is already being used in the drug discovery process, being used to process medical images, being used to diagnose patients, being used to invent vaccines, being used to understand the 3D structure of proteins, and is even in the beginning stages of starting to perform surgery. Pilot studies have been demonstrated where robots have performed dental surgery and installed brain-computer interfaces. Ray argues that over the next decade or two AI will replace clinical trials with simulations, and it will replace human surgeons with robotic surgeons that have each trained on billions of surgeries, meaning that robot surgeons will be more competent than any human ever could be. He also argues that by the 2030s well will be healing and augmenting ourselves with brain-computer interfaces and nanotechnology. He is predicting that we will turn into a species of immortal cybernetic geniuses that do not compete with AI for control of civilization, but rather live in a state of symbiosis with it.

I don't have any complaints about what Ray Kurzweil has written in this book but I do have four complaints about what he has neglected to write about. These are:


  1. The second last chapter is called "Peril" where he acknowledges some existential risk from advancing technologies. He acknowledges that our world is in danger from nuclear weapons, biological weapons, grey goo, and unaligned artificial intelligence. However he doesn't seem to address the arguments being made that these are actually grave concerns. People like Eliezer Yudkowsky and Nick Bostrom have argued in favour of AI risk being a very serious problem. And Kevin Esvelt has argued in favour of bioterrorism being a very serious problem, and yet he doesn't even mention their concerns. Making predictions about where the world is going as a result of technological advancement while leaving the leading voices on securing the future conspicuously absent seems like an obvious oversight to me.
  2. I anticipate that there will be legal or regulatory issues with his healthcare predictions that are also left conspicuously unaddressed.  I don't doubt that AI-powered simulations of clinical trials and robotic surgeons will be technically feasible, but how will the AI industry convince the government to allow these things? He doesn't address this fairly obvious issue.
  3. He argues pretty convincingly that UBI will be something that society will be rich enough to afford in the near future and that it will be politically feasible, but he says very little about how he expects it to be implemented or how he thinks it should be implemented.
  4. He argues that AI will help us cure aging but does not touch on AI's role in measuring aging which seems highly relevant to knowing that we have in fact cured it. Since 2013 there has been a proliferation of "epigenetic clocks" which analyze DNA methylation or chromatin accessibility to measure a person's biological age. These epigenetic clocks should be useful in validating that anti-aging medications are effective. The first epigenetic clocks were the Horvath clock and Hannum clock in 2013. Since then we have seen other variations of this such as the PhenoAge clock in 2018, the GrimAge clock in 2019, and the ATAC-clock in 2023. In 2022, we saw the invention of the AltumAge clock by researches at Brown University, which was an epigenetic clock designed using AI. It would have made the book noticeably better if Ray had explained why epigenetic clock design using AI was a good idea, and how such AI-designed clocks might validate that Longevity Escape Velocity had, in fact, been reached.

Overall, I think the book is good. And I do recommend it. But it's incomplete. It is one smart person's ideas about where the future is headed. But other smart people are just as worth listening to.

New Comment
1 comment, sorted by Click to highlight new comments since:

Is he still predicting the singularity in 2045?