In the last two years, artificial intelligence has surged into everyday life. ChatGPT was adopted faster than any previous consumer technology. Because it can instantaneously synthesize information, AI has already changed the way students must be tested and threatens to displace workers from concierges to coders. Even political candidates debate over appropriate AI policy. This technological supernova took almost everyone by surprise, as shown by the sudden jump in the prices of AI-related stocks, such as Nvidia. Yet, for Ray Kurzweil, now a Google computer scientist and the author of the new book The Singularity is Nearer, AI’s dramatic social entrance was right on schedule.

　　Kurzweil is simultaneously our most meticulous and boldest futurist. His rigorous approach stems from his past entrepreneurship. He worked for decades developing helpful aids for the blind, which required calculations about future computing power to make the most helpful aids, such as a reader that would convert print into voice, for the blind in each succeeding decade. As a result, Kurzweil has developed a careful timeline for how advances in computing power propel advances in other technologies.

　　In his previous book, The Singularity is Near, his big idea was the “law of accelerating returns.” He argues that computation gets better exponentially, and that exponential growth will be reflected in ever faster jumps of computational machine ability. To give one breathtaking example, in 1935, a computer could do .0000007 calculations per second, and in 2023 for the same price, it could do 65 billion calculations per second. As the economist Bob Lucas once said, once you start thinking about exponential growth, it is hard to think about anything else.

　　Kurzweil’s specific predictions reflect his audacity. For instance, in The Singularity is Near, he foretold that machines would be able to pass the Turing test by the end of the 2020s—that moment imagined by the mathematician Alan Turing occurs when it is impossible for humans to tell the difference between conversations with machines and conversations with fellow humans. Kurzweil also predicted that by the 2040s computation would become so advanced that it would be impossible to imagine our world. That is the technological singularity—a metaphor taken from the way astronomy describes how no light escapes from black holes.

　　When Kurzweil began making predictions, most thought his claims were outlandish. But with the rise of ChatGPT they seem less implausible—because ChatGPT is a direct result of exponential increases in computation. The “deep learning” mechanism on which it is based calculates its answers from 400 billion artificial neuron connections, approximately half the connections of the neurons in our brain. ChatGPT and similar AI programs became possible when calculating ability was able to encompass so many connections. And there is no reason to believe that the exponential progress in computing power will not continue because, as Kurzweil has demonstrated, the trend has lasted for over a hundred years and has survived even as the basis of computation has shifted radically from mechanical manipulations to vacuum tubes to silicon.

　　In his new book, Kurzweil shows that the consensus now is that the Turing test will be passed in 2029, vindicating his prediction two decades earlier. In some areas of general ability, machines are already showing capacities equal to all but the very ablest humans. For instance, just last week a machine was able to get a silver medal at the Math Olympiad, a test for the brightest high school mathematicians. The problems on the exam require substantial mathematical creativity, not rote learning. The vast majority of people, however long they prepared, wouldn’t score any points.

　　Kurzweil makes even more startling predictions, but he also advances two important claims about future human flourishing—controversial political and moral propositions that depend on more than inferences about increasing computer power. First, the only way humans can be sure of avoiding the existential threat of machines taking over is to meld with machines themselves. Second, such transhumanism is a great good, because the human experience will become far more profound than that we now enjoy.

　　Kurzweil’s Predictions

　　The most striking of his predictions is that even people on the cusp of old age might live indefinitely. The key is his claim that there will soon be the ability to reach escape velocity—a way to outrun death. Currently, for every year we age, we effectively get back a few months, because medical science has improved in ways that will statistically forestall deaths in the future. But Kurzweil argues medical progress will soon move much faster. By 2029 we will get more than one year of statistical life extension for every year of life; the longer we live, the longer we will be likely to extend our life even further due to technological advances. Of course, we will still face a risk of sudden death. To avoid that risk, computers must capture all the information in our biological brain, which Kurzweil contends will happen by the late 2040s—then our brains can be uploaded as a backup, much like a computer file.

　　As with his past book, Kurzweil provides detailed reasons for his predictions. Medicine progresses on the shoulders of computation. For instance, the rapid introduction of covid vaccines was the result of computers making a list of millions of possibilities, then trying them out computationally and deciding which was the best. That took all of 48 hours. While these vaccines were tested on humans, Kurzweil argues persuasively that computers will be able to simulate the human body so well that we will more rapidly test our medicines through computer simulation. Medicine will now accelerate exponentially along with computation.

　　Kurzweil’s struggle to depict the Singularity suffers from the same problems as those of theologians describing heaven.

　　What seemed more speculative to me was Kurzweil’s confidence in nanobots, which he sees as the next frontier of medical progress. Nanobots are robots minute enough to go inside us through blood vessels and other biological pathways. They will repair our broken, diseased, and aging bodies. Importantly for Kurzweil’s grand vision, they will also comprehensively map the brain—the prelude to the “uploading” of the consciousness that he hopes can take place.

　　One roadblock to these technological possibilities that Kurzweil does not sufficiently address is politics. Peter Thiel famously said, “We wanted flying machines and all we got was 140 characters,” referring then to Twitter, now X. Thiel blames this deceleration of progress on the heavy government regulation of things as opposed to computer bits. Medical devices and drugs are even more heavily regulated than locomotion. Moreover, the speed of general technological change that Kurzweil describes may result in political turbulence as people vote to avoid disruptions in their lives—even those that would ultimately prove very beneficial to them. Such tumult will make rational social policy even more difficult than it is today, because voters are guided more by emotion than reason.

　　AI’s Dangers

　　Despite all his optimism, Kurzweil recognizes the possible downsides of AI. Once computer scientists create an artificial general intelligence, which he sees happening within a decade, they will rapidly outperform humans at all tasks, raising the specter of unemployment. He also does not dismiss that such power will give AIs the autonomy to slip away from human control. Kurzweil’s bold solution to keep humans relevant is for us to meld with machines—augmenting our intelligence by linking to computers and thus multiplying the connections of our biological neurons with billions of artificial ones.

　　The idea may not be as outlandish as it might first appear. As Kurzweil notes, almost everyone carries around a brain extender in the form of a smartphone. Since the early 2000s, the smartphone has allowed users to access all the information of the World Wide Web and now allows us to benefit from the ever-improving cognitive abilities of generative AI. But the link is aslow and clumsy one—either through our fingers or our voice. If the connection were directly to our brain, the enrichment of intelligence would be far greater and would grow directly with the exponential increase in computing.

　　But just as the problem with achieving escape velocity may be the all too human problem of politics, the problem with protecting ourselves from AI by melding with it is the all too human problem of humans. Just as some people today use their abilities to try to gain power and exercise dominion over others, so will they use their greatly expanded powers to do the same thing in the future. The result may be much wider and more harmful conflict.

　　Kurzweil, of course, is aware of the problem of evil and the destruction it unleashes. Nonetheless, he has a Whiggish attitude towards progress. He argues at length that life has been getting better for centuries—in wealth, longevity, and even in the relative absence of war. This claim is broadly accurate, but the long-term trend has been interrupted by some descents into the abyss—the World Wars and the Holocaust, to name a couple of sad examples. Given the increasing powers Kurzweil predicts humanity may achieve, some future self-inflicted calamities may be so sudden and sharp that there will be no return. Greater intelligence without greater moral sense becomes a greater threat when integrated with the human will to power.

　　Kurzweil’s failure to fully address human evil and goodness as opposed to mere intelligence also helps explain why he cannot paint an entirely attractive picture of the singularity. Assuming his technological predictions come to pass, what will it be like to live in 2050? By his own admission, this world is difficult to describe. Indeed, his struggle to depict the Singularity suffers from the same problems as those of theologians describing heaven. When writing about paradise in the Divine Comedy, Dante, one of our most eloquent poets, said he was reduced to the “babbling of a baby.”

　　Kurzweil says progress will give us a deeper appreciation for the connections of things. For instance, we will be able to make and understand more beautiful music and visualize more than three-dimensional matrices. I would add that, just as in heaven, we are likely to be much more equal in our talents and perceptions as our added computation will swamp individual biological differences in brainpower.

　　But is more knowledge enough for true human flourishing without more generosity of spirit? Will his beings of surpassing intelligence have great souls or just great ability to indulge whatever turns them on? Kurzweil has nothing to say about this. Dante—for all his inability to precisely describe the rapture of heaven—put at its center “the love that moves the sun and other stars.” The void at the center of Kurzweil’s vision makes it no less fascinating, but even more unsettling.