Man and Superman

Reviewing Ray Kurzweil's latest book

Ray Kurzweil has published The Singularity is Nearer, his latest book of prophecies. As I read it, he now emphasizes the path to transhumanism, in which we become immortal and super-intelligent, augmented by a brain-computer interface.

Kurzweil’s earlier books, The Age of Spiritual Machines and The Singularity is Near, included glimpses of this vision. But my main takeaways from them concerned the impact of Moore’s Law. Kurzweil saw computers shrinking, so that he saw drone warfare coming, as well as wearable computers. (In fact, he anticipated these developments proceeding notably faster than they actually have.) He saw computers becoming more powerful, so that they would take over more and more functions, such as language translation.

The advent of Large Language Models has provided a strong vindication for Kurzweil’s faith in artificial intelligence. On p. 14, Kurzweil describes his early contact with Marvin Minsky.

Minsky taught me that there are two techniques for creating automated solutions to problems: the symbolic approach and the connectionist approach.

I used different terms when I wrote Decision Rules vs. Pattern Matching. But I think I got it pretty much right.

Kurzweil explains the fatal flaw of decision rules on p. 17.

One way of looking at the complexity of rule-based systems is as a set of possible failure points.

Any given rule can go wrong. As you add rules, the number of possible ways for the system to go wrong grows exponentially.

there comes a point (where exactly this point lies varies from one situation to another) where adding one new rule to fix a problem is likely to cause more than one additional problem. This is the complexity ceiling.

The pattern-matching systems have a key advantage. On p. 18, Kurzweil writes,

they have the potential to discover subtle patterns that would never occur to human programmers trying to devise symbolic rules…it allows you to solve problems without understanding them.

…This has the potential to become a major issue because people will want to be able to see the reasoning behind high-stakes decisions…It remains to be seen how effective transparency will be as deep learning becomes more complex and more powerful.

An example with which I am familiar comes from credit scoring. Human underwriters never paid attention to the number of credit inquiries showing up on an individual’s credit report. But using pattern-matching techniques, credit scoring companies found that inquiries were predictive of defaults. In hindsight, we can explain that a lot of inquiries can be a sign that the consumer knows he is headed for trouble and is desperately searching for a loan.

Kurzweil sees pattern-matching systems as capable of solving complex scientific problems. Protein folding is an important example. On p. 238-239, he writes

Our bodies are mostly made of proteins, so understanding the relationship between their composition and function is key to developing new medicines and curing disease. Unfortunately, humans have had a fairly low accuracy rate at predicting protein folding…

This is where the pattern recognition capabilities of AI offer a profound advantage…The AI is now able to achieve nearly experimental-level accuracy for almost any protein it is given…This will accelerate the pace of biomedical discoveries.

Advances in biology and materials science are crucial to achieving Kurzweil’s vision of humans enhanced by computers and able to live forever. But is the process of scientific discovery merely a matter of applied pattern-matching? Or will AI trained on human thought from the past be unable to develop new insights?

Another key component is nanotechnology. On p. 71, Kurzweil writes,

Ultimately, brain-computer interfaces will be essentially noninvasive—which will likely entail harmless nanoscale electrodes inserted into the brain through the bloodstream.

This will enable human intelligence to improve at the rate of Moore’s Law. We will have superhuman intelligence.

Nanobots also can provide a cure for aging. On p. 245, Kurzweil writes,

A combination of AI and the nanotechnology revolution will enable us to redesign and rebuild—molecule by molecule—our bodies and brains and the worlds with which we interact.

On p. 258-259, he writes,

As I see it, the long-term goal is medical nanorobots. These will be made from diamondoid parts with onboard sensors, manipulators, computers, communicators, and possibly power supplies…

we will need huge number of nonobots, each about the size of A cell…If we augment ourselves with just one nanobot per one hundred cells, this would amount to several hundred billion nanobots…advanced nanobots could be effective even at a cell-to-nanobot ratio several of orders of magnitude greater. ..

But nanobots won’t be limited to preserving the body’s normal function. They could also be used to adjust concentrations of various substances in our blood to levels more optimal than what would normally occur in the body.

Gosh. If you thought that baseball players’ use of steroids or transgendered competition in sports sparked controversy…

Philosophical and ethical questions aside, is nanotechnology securely on a path to reach the objectives that Kurzweil lays out for it? Or will there be point where key problems go unsolved for decades, if not longer?

Even with Kurzweil’s aggressive timeline, these developments are two decades away. Meanwhile, on p. 100 he writes

During the late 2020s advanced AI will be able to create a very lifelike replicant, drawing from thousands of photos, hundreds of hours of video, millions of words of text chats, detailed data about the person’s interests and habits, and interviews with people who remember them.

I can see this being developed without any new technological breakthroughs.

Kurzweil is 76 years old. Although I read years ago that he was taking many dietary supplements in the hope of slowing the aging process, it seems to me unlikely that he will live to see whether his more extravagant predictions come true. But I certainly expect that we will have the ability to converse with a Kurzweil replicant.

[Note: “Man and Superman” is the title of a play by George Bernard Shaw. Its famous lines include “Those who can, do. Those who can’t, teach” and “Do not do unto others as you would that they should do unto you. Their tastes may not be the same.”]