Clear and Present Dangers of AI 

What to make of recent developments in artificial intelligence (AI), in which field there seems to be something noteworthy happening on almost a daily basis? Some intelligent critics, focusing on the big picture, have little trouble presenting not completely implausible if low probability scenarios in which the whole enterprise eventually culminates in a disaster for humanity. Daily news coverage of AI-facilitated individual bad behavior and tragic outcomes alone suggests good reasons for concern even in the short term. Hanging over all is the unease about AI proving to be an economy-threatening bubble. 

On the other hand, even in the face of such loud criticism, adoption of AI tools in business, government, the professions, and the academy has spread with remarkable speed. (I am known in my social circles as something of an AI skeptic, and people seem to delight in telling me the in-their-view innocent and useful things they have just done with AI.) We are assured by both disinterested and highly interested parties that, while AI is going to cause some near-term disruptions in the manner of past revolutionary technologies, that’s just how technological progress always works. The printing press was highly disruptive, but imagine a world without readily available books and periodicals! Luddites never win and shouldn’t win, because their victory would mean sacrificing the possibility of a future with AI that will probably be very bright indeed.  

It is genuinely difficult to know how to begin to navigate all of this chatter and adopt a more thoughtful stance on the issues facing us. But Sherry Turkle’s latest book, Artificial Intimacy: Who We Become When We Talk to Machines, is a good place to begin. From her perch at MIT, Turkle has been an insider critic of developments in robotics and artificial intelligence for some time now. Details of her professional biography and incidents she recounts in this book suggest that her critical stance has not harmed, and may even have aided, her professional standing, which speaks well of her and her MIT colleagues. On the other hand, I think it is fair to say that so far she has been on the losing side of most of the issues she has taken up, despite having been (in my view) on the right side on things like the internet and social loneliness, or the consequences of developing robotic “companions.” It’s relatively easy to make space for a voice in the wilderness when you can be pretty sure it is going to remain in the wilderness. 

Only time will tell whether the same fate awaits her analysis of the deeply problematic consequences of our increasing reliance on AI. Artificial Intimacy has many virtues. The prose is clear and compelling. Turkle writes with passion yet avoids any trace of sensationalism; this is not a “we’re all going to die” look at AI. She is thoughtful and can appreciate more than one point of view. As we will see shortly, Turkle identifies some of the most crucial, yet at times underappreciated, immediate challenges that AI is creating. 

Turkle, whose training is in both sociology and psychology, structures the bulk of her chapters around the impact of AI on different phases of “the arc of life”: childhood, adolescence, identity, love, work, therapy, mourning. These chapters are built around a series of carefully conducted interviews with people about their experiences with AI. She seems to have posed thoughtful questions and plainly received some thoughtful and telling responses in return. An opening chapter lays out the essential situation as she sees it (“A State of Emergency”); a closing chapter is another overview of AI in a post-truth world and contains some modest proposals for “reclaiming the real.” 

Those who are most likely to pick up this book are probably going to be familiar with a host of the problems related to AI that Turkle covers: “hallucinated” responses, toxic advice and “relationships,” students and mature professionals alike using it to cheat and cut corners. Turkle’s skill is weaving all this material together into a coherent and compelling critique not of future horribles, but of the world AI is already creating. She provides insights into the reasons people are behaving these ways, both in terms of specific motivations of misusers and in terms of the broader context created by this new technology and out of which it arises. 

The latter is particularly important to the extent that this technology, like nearly all new technologies, comes along with announced beneficent intentions and projected results, the developers doubtless being motivated (for the most part) by some combination of regard for human well-being, in accord with their own visions of what constitutes “progress,” and the desire for commercial success and personal profit. Attuned to the possibility of criticism, skepticism, and cynicism, AI developers assure us that they are working hard to maximize benefits, which they see as being potentially world changing, and minimize risks, which they see as manageable. This technology, like all technology, is “just a tool,” and can of course be misused, but there is nothing new about that and we can just move on.  Indeed, AI may help us to develop new and more effective ways of dealing with the risks it is going to pose.  

One specific line of argument that Turkle deals with particularly effectively is that by always keeping a human being “in the loop,” we can aim at using AI in ways that enhance human capacity rather than replace it. There are two layers of difficulty with this kind of assurance, Turkle suggests. Merely having a human being in the loop does not guarantee that the human will be effective in that role; according to her interviews there are already reasons to be concerned that people, acting within the preexisting constraints of bureaucratic organizations, find reasons not to come forward and call AI errors into question. When any bureaucracy, business, or government has decided as a matter of policy to “integrate” AI into its core functions, then management is not necessarily going to want to hear about problems with or limits to that aspiration. The problem may only be exacerbated when “integration” involves reducing the very workforce that is most likely to have the specialized knowledge to catch issues with how AI is functioning.  

This point brings us to a yet deeper problem with “having a human being in the loop” that is perhaps most clearly evident in what Turkle tells us about AI’s impact on higher education. It is already pretty clear that the more AI is used in this setting, licitly or illicitly, the less people learn. As I have had occasion to remark elsewhere, using AI in education and thinking you will be educated is like using a robot to exercise for you at a gym and thinking that you will become fit. The substantial risk already evident today is that people will have fewer and fewer intellectual skills and a smaller and smaller knowledge base, such that they will have very limited capacity to serve as a “check” in any AI-dependent decision-making loop. They will lack the foundations on which to act prudently. 

This is one of the deepest and most important implications of Turkle’s book. For most of us most of the time, developing and maintaining the good things we want out of life (family, friends, satisfying work, physical and mental well-being) requires just the kind of effort that AI promises to reduce. It is true that the long-term trend of modern technology has been, by design, to reduce the effort required for many of the once physically arduous requirements of living, and long before AI there were inventions that lowered the bar for mental effort as well. But what will the consequences of yet greater abdication of mental effort look like?  

We can imagine various possibilities. We see in the United States how those with sufficient income, discipline, and leisure put avocational effort into careful programs of exercise and diet to avoid or mitigate the various “lifestyle” diseases consequent upon reduced physical activity and easier access to calories. Analogously, in a world where AI takes over activities requiring mental resources, skills, attention, and effort, might we see the intellectual skills once required from boardrooms to battlefields “gamified” into niche and varied competitive displays of mental prowess (and complex wagering modalities pertaining to them)? 

At least one AI advocate has a different vision of how AI could yet enhance human capacity and not replace it. He imagines AI tutors with godlike awareness of and patience with the individual needs of each student, so that all learners can reach their highest potential. They will enter the “workforce” ready to “be in the loop” and cooperate with AI to make the world an unimaginably better place. Or perhaps because AI will, according to some of its advocates, increasingly free us of the curse of work, this highly individualized instruction will be designed to be a “liberal” education in an older sense: that is to say, an education appropriate to those free of the necessity of labor. 

The last of these three possibilities, while perhaps most appealing to some, seems least likely to me. Turkle introduces the argument why, but she might have done more with it. In all the talk of AI’s revolutionary potential, it is easy to miss how its promises and perils alike arise in the context of preexisting conditions. Again, the aspiration to perform less work with greater productivity has been around for a long time. But as far as perils go, AI turns out to be an enabler of much that is problematic about twenty-first-century American society. AI companions look attractive to some not because AI created the problems of loneliness and social isolation; both long predate it. AI “therapists” arise in a context where there is already a shortage of mental health workers. AI can only look appealing to students in higher education because so many colleges and universities have lost any serious sense of what higher education ought to mean, and because primary and secondary education is generally speaking not doing a great job of preparing students for higher education. Where AI proponents see it as being able to solve these problems, the more thoughtful critics see it as exacerbating them.  

My own judgment is that with respect to our various social, political, and cultural ills, AI will continue to be enabling rather than disruptive. I am old enough now to have seen waves of reform in education, in the mental health “system,” and in “community building,” that have not prevented us from being where we are (and some of them have probably contributed to our current ills). I don’t see why AI will be any different, not only because it is likely to underachieve what is promised for it, but because so many of those promises point to a world that hardly seems conducive to human well-being. I am among those who have been writing for some time now about our increasingly “thin” understanding of what it means to be a human being. The current AI debate takes place in this depleted context, as is seen by the fact that we are so easily distracted by arguments over an “AI apocalypse.” It is certainly worth paying some attention to safety at this level, but for the most part, attention directed here is a distraction from the troubling consequences of AI that are right in front of us. But dealing with such consequences requires us to go well beyond what AI can or cannot do, to get beyond even the question of the value of the mere life of humanity, to questions of what makes for a good human life—just the question our time is so ill prepared to take seriously.  

Turkle tries to end her book on a hopeful note, and I would like to be able to do the same with this review. Given the potential for devasting economic consequences, it would be perverse to hope that AI proves to be a bubble. But if it is not, it has too much momentum behind it to stop or even slow down much. But that does not mean that individuals and associations of the like-minded have to give in to it. Again predating AI, there are schools and religious communities in a position to provide some antidote for those who take advantage of their existence. There are already signs that classical schools, for example, are looking even more appealing to concerned parents under our present circumstances. In some corners of higher education, bans on technology in the classroom are being implemented. If these institutions do their jobs well, there will be a cohort of citizens with the moral and intellectual wherewithal to value human intelligence over artificial intelligence, to aspire to “reclaim the real,” and to structure their lives accordingly. Perhaps they will be an example to us all. 

 Image licensed via Adobe Stock.