Since the launch of version 3.5 of ChatGPT at the end of 2022, the discourse on the future impact that artificial intelligence (AI) will have on our societies has become central, generating very polarized reactions. From some like Bill Gateswho considers it one of “the two technological demonstrations that seemed revolutionary” in his entire life, to others like Nick Bostrom, who defends that AI is one of the greatest challenges facing the human species been confronted.
The debate essentially revolves around whether the problems that AI can solve, such as disease or climate change, will compensate for the possible risks to which we will be exposednotably the delicate problem of man-machine alignment.
One of the most optimistic versions of our future with AI has been championed by Ray Kurzweil, current director of Google Research, for more than two decades. The publications of his two books, The era of spiritual machines (published in 1999) and The singularity is near (published in 2005), were greeted with a certain amount of skepticism, since Kurzweil presented several of his predictions about the evolution of AI which were then considered excessively optimistic. Among his most famous visions, two stand out: the first, according to which before 2030, machines would pass the Turing test, that is to say that a human judge would be unable to distinguish the responses of a human from those of a machine, and therefore we could consider that the latter has reached the level of intelligence of humans; and the second, that by 2050 we will have reached “the singularity”, the moment in history where machines will be able to design and improve at an exponential rate.
In his latest book, whose title is a nod to that of 2005, The singularity is closer: when we merge with AI, Kurzweil updates his predictions and also shows the current state of AI in various fields. For those who are familiar with Kurzweil’s work, they will find this latest book somewhat repetitive (which, in addition, is a little less technical than his others like How to Create a Mind, which I reviewed for this same column last year). more than three years ago), especially its final part, although readers interested in the field of AI will find the first chapters very provocative and full of important ideas. However, how have Kurzweil’s predictions aged since then? Perhaps most notable is the previously mentioned claim that machines will pass the Turing Test before the end of this decade. Recently, this date has become the most likely possibility in Metaculus, one of the largest prediction portals in the world. On said portal, and as I write these lines, the average of predictions places said event for the year 2029, which adds credibility to the possibility that his prediction is closer to coming true than many believe. had planned at the time.
Kurzweil’s long-term forecasts have always been based on what he calls the Law of Accelerating Returns (LOAR). This law states that, unlike many other technologies (such as transportation-related technologies, which have not progressed much in recent decades), information technologies (such as printing, computers or data sequencing) evolve exponentially over time, as each advancement builds on previous ones and, in turn, facilitates the next stage of technological evolution.
As Kurzweil explains: “One dollar, adjusted for inflation, can buy about 11,200 times more computing power than when The Singularity is Near was published. Many readers will find this law very similar to that stated by Gordon Moore, co-founder of Intel. “, in 1965, when he predicted that the number of transistors in a circuit would double approximately every eighteen months, thus leading to improvements in chip efficiency and reduced costs. For Kurzweil, Moore’s Law is simply another paradigm in the long arc of the history of the improvement of our computing systems and, when this paradigm is exhausted (as happened before with relays and vacuum tubes), it will be replaced by another which will allow us to continue to exponentially increase our computing capacity.
We are currently experiencing, according to Kurzweil, a crucial moment, since this exponential progress has reached a stage where it will allow us to radically catalyze further advances in three other key areas: our programming capacity (which is already a reality thanks to ‘AI like AlphaCode), our understanding of biology and our ability to design and manufacture materials at smaller and smaller scales. These developments will allow us to exceed the natural biological limit of human life, extending it beyond 120 years and, ultimately, achieving the possibility of transferring our consciousness to the cloud, where it can have a more substrate. durable and permanent.
All of these developments, as well as many others that I have not mentioned but are covered in the book, may sound like science fiction to many of our readers. To be sure, many of Kurzweil’s predictions will remain mere speculation. This is why I wanted to start this article by emphasizing that Kurzweil is not a typical science fiction writer, but rather a researcher very close to the development of AI and who has a pretty decent track record when it comes to predictions bold. Quite simply, we should think long and hard about the implications this might have for our future. Throughout the book, Kurzweil presents countless graphs and figures to support his view that technological progress has always been good. From increasing global per capita income and literacy over the last century to decreasing poverty rates, violence or increasing life expectancy, just to name a few- one, all these indicators do indeed indicate indisputable improvements. However, the nature of AI advancements is fundamentally different from any previous technology.
Not only because AI can lead to the extinction of our species if not properly developed, but because it will change our concept of what human intelligence means and our place in the universe. massive disappearance of jobs that will be done by machines (while we lie on our sofa enjoying our universal basic income, which will then be very widespread, according to Kurzweil) and the rest of the cornucopias make us doubt that All this sounds like a utopia or dystopia, don’t worry, you’re not alone; The ethical and social implications of these advances are far from explored in depth in the book. And this is perhaps the weakest part of the job.
Most of the issues Kurzweil mentions throughout The singularity is closer They are resolved by the author in passing and superficially. Several of these (e.g. job automation) are seen as having negative impacts in the short term, but are forgiven in the long term (e.g. all technologies resulted in job reallocation in the short term, but at long term ended up creating more jobs than were destroyed). For those looking for a more in-depth analysis of these issues, Nick Bostrom’s latest book, Deep Utopia, offers some interesting alternative perspectives. Readers will nevertheless find it interesting The singularity is closerboth as an accessible introduction to crucial debates and because of the inevitable and growing importance that AI will have in our lives.
‘Technical sheet’
Title: “The Singularity is Closer: When We Merge with AI”
Author: Ray Kurzweil.
Publisher: Bodley Head, 2024, pp.432, hardcover.
