Artificial intelligence as the fourth great blow to human narcissism

As a prefatory note to these brief remarks: My understanding of artificial intelligence owes greatly to the conversations I have had with my friend, Steven Sidley, a brilliant commentator on matters political, economic, and humanistic, who is currently professor at the University of Johannesburg. Two of our discussions are now available as podcasts on the RediscoveringPsychoanalysis website.

Especially since the initiation of Chat GPT in November 2022, there has been increasing public awareness of the capabilities of artificial intelligence, but far less discussion of its liabilities. That is to say: By and large, the narrative has been buoyant, insufficiently alarmed by the dystopian future that AI is almost certainly going to inflict upon us. Indeed, there is an imminent future of even greater damage to the human psyche than has been generally considered. Think here not only of unemployment, and crippling injuries to our individual sense of efficacy ― matters to which the psychoanalytic community should surely be more than attentive ― but also the looming likelihood of the extinction of our species within our lifetimes. I am not being exaggerative, and I will argue that there no room here for denialism here. I will frame my argument like this.

Freud famously referred to his discoveries as the third great insult or blow (Kränkung) to human narcissism. If we briefly examine this claim, its peculiarity is illuminated. The claim presumes our “naïve self-love as humans” (naive Eigenliebe, menschlichen Eigenliebe). That is, the supposition that our identification with the status of our species sustains our self-love. This seems very unlikely. Indeed, it is obviously Freud’s fantasy. Because, from the standpoint of lived-experience, not much really changes with these cosmological, biological, and psychological insults. Despite Nicolaus Copernicus and Alexandrian science, we still experience the sun as rising in the east orbiting around us and disappearing to the west. Despite Charles Darwin, along with Alfred Wallace, our experience of ourselves as king of the hominids is unperturbed, even if we are not made in the image of the divine. Even as suicidal a species as we are, we can still conceive of ourselves as the ‘latest and greatest’ of evolution’s productivity. Our grandiosity persists. We remain complacently smug in our alleged superiority, as we take trips on safari or to the zoo, even as we enact our ecocidal proclivities.

And … what about “despite Sigmund Freud?” Generally, until we breakdown and become aware of our madness, we remain unfazed by his discovery that we are governed dynamically by unconscious forces. Indeed, decades of so-called psychoanalytic theorizing have attempted to restore the ubiquitous human illusion of consciousness-as-mastery. Think here of the depressive position, of the so-called ‘conflict-free sphere of the ego,’ of the humanistic imageries of self-integration or relational congruence, and so forth.

But now, a fourth insult is upon us, and it is not only unprecedented in the damage it is going to do to our psyche ― not only signifying the end of humanity-as-human, but also pointing to the possible extinction of our species. This is a blow of a depth and magnitude far greater than the free that Freud previously nominated.

Consider this. We are in an epoch of developments in artificial general intelligence that are accelerating exponentially, just in the past five to ten years. If you survey the literature on artificial intelligence from 2015 to today, much of what was then considered science fiction, or a dim future possibility, is now actually with us or about to come upon us. There has been an utterly astounding volume of breakthroughs in software, hardware, formal frameworks, and machine learning architecture. And here it is appropriate for us to remind ourselves of the difference between, for example, software and AI: AI uses probabilistic techniques that empower it to learn autonomously.

With artificial general intelligence, computers and robots can already match or exceed human capability in almost every conceivable task: Not only cognitive abilities, such as problem-solving and learning, but also emotional intelligence (that is, the ability to relieve stress, communicate effectively, empathize with others, build relationships and navigate complex social situations), as well of course, whatever we call “creativity.” It does all such tasks better, quicker, and with less expense, that we can.

Think for a moment of what is already beginning to occur: Massive unemployment, not only of skilled workers, but also of professionals, operations’ managers, healthcare practitioners, lawyers, as well ― of course ― as psychotherapists and psychoanalysts. But think also of the deep-rooted injuries that are inflicted upon our sense of efficacy and self-esteem ― that is, our sense of being able to have an impact on the world we inhabit.

Three indisputable facts cannot be avoided here. The first (I already mentioned), these developments in the field of artificial general intelligence are accelerating exponentially and ― as I am about to illustrate ― they are unstoppable. Second, AI is already functioning in a manner that is increasingly autonomous and with demonstrable agency. For example, there are programmes that can deceive, that can set their own goals, and that write code in the pursuit of these goals. Thus, AI begins to generate and control its own progeny, as much as it controls us. Third, with this escalating autonomy and agency, AI generates output that has become impossible to ‘reverse engineer’. That is, it can perform very complex tasks at great speed, arrive at correct solutions, with humans being unable to figure out how the machine managed to get it right ― the combinatorics are far too large. Thus, the inner workings of AI are already beyond human comprehension.

Now, as artificial general intelligence comes to outperform humanity on every conceivable task, we enter the epoch of artificial superintelligence. The best minds in this field estimate this to be months or a few years away ― definitely not years and decades. Consider the following.

In March 2023, the Future of Life Institute published an open letter. It is now signed by over 33,000 people, but initially it was signed invitationally by several thousand of the top experts in AI, and in adjacent fields such as cognitive philosophy, etc. We are talking about names such as Steve Wozniak, Noel Harari, Elon Musk, Daniel Dennett, Bill Gates, David Chalmers, etc. etc. The letter endorsed the Asilomar AI principles and asserted that what it called “advanced AI could represent a profound change in the history of life on earth, and should be planned for and managed with commensurate care and resources.” It addressed the serious risks of any further AI development (risks that included human extinction), and it called on “all AI laboratories to pause immediately” for at least six months any further development of AI beyond GPT4.

It is obvious that such a hiatus never happened, and realistically it could not. Such a pause could depend on international multilateral cooperation in a field where enormous power is wielded by whomever develops the next iteration. Such a pause would be unenforceable. Not only are the superpowers never going to agree to such cooperation, but AI development depends neither on the blessing of superpowers, nor on institutional patrimony (we have already witnessed situations in which some bright sociopath, operating out of a basement in some remote part of the world, can do untold damage).

There was one very eminent AI expert who declined the Future of Life Institute’s invitation to sign onto the letter. This was one of the most highly regard scientists, Elieszer Yudkowsky. He refused to sign, arguing essentially that it was too little and far too late. In his public response, he stated emphatically that “AI will not do what we want” and that “AI does not care about us.” He summed up his argument: “Everyone will die” and he predicted this outcome in less than 10 years.

Essentially, what Yudkowsky is highlighting is that as artificial superintelligence not only outperforms us in every sphere, but also attains autonomy, agency and volitional freedom, we have no safety against nefarious acts. Machines created by us, but now superior to us and out of our control, will assess human affairs without any real sense of attachment or moral duty to alterity.

Naturally, Yudkowsky’s dissent sparked rejoinders, but even these were alarming. One of the most articulate and expert responses tried to be reassuring by quibbling with the Yudkowsky’s predictions. It was countered that actually the more accurate calculation threatens humanity with only a 10% to 20% chance of extermination by AI. So even the most sanguine experts calculate that only 80% or 90% possibility that our species will survive the inevitable advent of superintelligence. Almost all of these experts agree that these risks will come to pass not just within 10 years, but rather “soonish.

So, we are talking not just about risks such as massive unemployment, inevitably leading to the breakdown of civil society, not just about risks such as an appalling bruising to our human sense of efficacy. But more than these risks, we are talking about the erasure of humanity as human … not in some sci-fi future, but within the lifetime of all of us here present.

And we haven’t even begun to consider the advent, after superintelligence, of quantum computing, in which multiprocessing permits a signal to be, so to speak, neither off nor on, both off and on.

This is the culmination of kali yuga in which the possibility of virtue is eclipsed by the power of machinery, the development of which was set in motion by humans, although humans are no longer in control of its power, nor of its further development. And I do not see how satya yuga, a golden age, can possibly come after it ― to believe that AI is merely going to make human life easier would seem to be over-optimistic (even actually delusional, given what we know to be happening, right now, around us).