The Death of Thinking

The zeal for AI large language models is naive.

That great British humanist-scientist, Humphry Davy, once said in a lecture delivered in 1810:

“Nothing is so fatal to the progress of the human mind as to suppose our views of science are ultimate; that there are no mysteries in nature; that our triumphs are complete; and that there are no new worlds to conquer.”

Were he to be exhumed to deliver a similarly themed lecture in 2023, he might well find himself saying that nothing is so fatal to the progress of the human mind as to suppose that it is no longer required. Such has been my sentiment at the recent explosion of exuberance over the artificial intelligence [AI] large language model, ChatGPT. In response, The Spectator  declared that ‘AI is the end of writing’; I think it may be the death of thinking.

That these evolutions in our technology have some utility is clear, and I don’t intend to dispute that. But we appear, even after all the toxic effects of BigTech for the past two decades, to continue to think that any new technology is a free lunch. That these are the benevolent gifts of the omnipresent, omnipotent Tech Gods, bestowed upon us for our ubiquitous benefit. That we have paid the Piper and consequently we think we call the tune, but in our intoxicated trance we’re oblivious to the fact that the Piper knew what we wanted to hear already and played us for fools.

Rather than question the utility of any technological tool to which we can outsource thinking, I want to think about the wider context of how tech culture is situated in our societies, and contrast this moment with its clearest historical analogy; the dialectic between the rationalist Enlightenment scientific revolution and its reaction in the Romantic revolution in literature and art. In consideration of this context, the utility of any technology like ChatGPT becomes somewhat secondary to the more pertinent question of whether, with tech and AI, we are the product or the consumer.

This latter question is at the core of how BigTech has led us on a merry dance, portraying itself as a benign mediator, draped in the buzzwords of “connection” and “experience”. As Shoshana Zuboff’s ‘The Age of Surveillance Capitalism’ revealed in painstaking detail, this is the BigLie at the core of BigTech; the overall illusion that we are the consumers and that tech advances for our benefit, when in reality we are the product, the unit of extraction for our personal data, our beliefs, habits, and preferences. We’ve already given BigTech a monopoly on our personal data; we’re now going to give it a monopoly on turning our knowledge into deepfake replicas.

The pertinent question posed recently by Ted Chiang in The New Yorker, is whether “statistical regularities in text actually do correspond to genuine knowledge of the real world?” Large language models, upon which ChatGPT is built, interpolate from texts to create, in Chiang’s analogy, blurry photocopies of photocopies to the point where the image just gets blurrier, and worse. Knowledge is now a simulation.

My anxiety at this prospect does not exist in a vacuum. These tensions have historical precedent in the emergence of the Romantic revolution as the reactionary intellectual force to the rationalist and scientific Enlightenment. The tension in this dialectic was captured by Caspar David Freidrich:

“...beware of the superficial knowledge of cold facts, beware of sinful ratiocination, for it kills the heart and when the heart and mind have died in a man, there art cannot dwell.”

The Romantic revolution evolved as a centrifugal counterpart to the Enlightenment, a literary and artistic movement which found expression in emphasis of the sublimity of nature, the primacy of the inner world of feeling, of the soul and subconscious, and a rejection of pure rationalism, which they saw as dismantling the world, its beauty, and the individual conscious, and rendering it into a meaningless collection of component parts. To paraphrase Richard Wagner’s critique at the time, it was that science would analyse everything and explain nothing. To quote Tim Blanning from his wonderful book, ‘The Romantic Revolution’:

“Nothing roused the romantics to greater indignation than the notion that nature was inert matter, to be understood by dissection, experiment, and analysis.”

Given this context, however, it could be tempting to assume that the Enlightenment-Romantic dialectic tension was a dichotomy of science and reason against nature and feeling. But this is not the case; the two movements fed into each other. The natural sciences at the time provided the very same of sublimity and awe of the natural world as Percy Bysshe Shelley’s poem, Mont Blanc. This interrelationship between science and the natural world of the Romanticists was expressed by Alexander Pope  in 1730:

“Nature and nature's laws lay hid in night,

God said 'Let Newton be!' and all was light.”

Indeed the very profession that we now call “scientists” were, at the time, referred to as “natural philosophers”, often with backgrounds in the humanities and Classics before their forays into experimentation and natural sciences. Humphrey Davy epitomised this overlap between the Scientific and Romantic Revolutions; he was simultaneously a distinguished poet as well as one of the most prolific contributors to the science of chemistry.

Yet our contemporary technological advancements, particularly related to AI, have no such intellectual counterpart. The Enlightenment-Romantic dialectic served to enrich both the sciences and the arts in an ongoing process of reciprocal invigoration. Nothing even remotely comparable exists today as a counterpoint to AI, the relationship between artificial intelligence and genuine knowledge, whether mere information can explain our world and ourselves without the component of human experience, and whether the boundless world of the human imagination can truly be replaced by predictive algorithms. And no such dialectic exists because, even before more ubiquity of AI, social media platforms have already reduced us to societies who don’t want to think, and don’t know how to think; a collection of bots convinced of their agency.

Large language models, to use Chiang’s analogy, are like a zip file of the internet, all compressed down to be opened up and used to assemble a blurry simulation of knowledge based on statistically derived commonalities of available language. In this way, it is a development that further unmoors us from language in this postmodern “desert of the real”, borrow from Morpheus. Here, too, we can heed the Romantics, as no one has emphasised the crucial connecting bridge that language plays for intelligence and culture more than Herder, who recognised, per Blanning, that:

“the vital link between part and whole, between one individual and another, between individual and community, between humans and the natural world, is language...Without language there can be no knowledge, no self-consciousness, no awareness of others, no social existence, no history.”

What happens when language becomes the domain and product of an intelligence with no self-consciousness, awareness, having no social existence or history? In compressing language into a proverbial zip file, we are in fact ceding language, and all of the thought and knowledge and emotion and beauty and humanity that language entails, to a formless, cultureless, monolith. Could anything symbolise our culture, and our cultural moment, more than this? Our vanity and banality, our vacuity and superficiality, our rootlessness and emptiness, our voyeuristic, ostensible reality. Rob Horning’s recent essay captured my anxiousness when he rather rhetorically pondered why wouldn't such simulated “versions of knowledge be good enough for the diminished truth capacities of our fallen world?”

The emergence of AI large language models is the endgame of the postmodern era. Postmodernism took the rejection of reason from Romanticism, and elaborated the culture of feeling into a primacy of hyper-subjectivity and epistemic relativism. The postmodernists, where any discernible sense could be gleaned from their “pseudo-profound bullshit”, relied on terrorising language and dismantling stable meaning from words, with a hyper-subjectivity where everything could mean anything so nothing meant anything. Their sophistry heralded in the age of “My Truth, Your Truth” epistemic relativism in which we now suffer, and now we’ve empowered AI large language models to provide us with blurry photocopies of multiple versions of truth and truthiness.

For postmodernists like Jean-François Lyotard who were concerned with the “legitimation of knowledge”, where science was just one of many “language games” that sought to legitimise itself through language, the fact that science has now produced the ultimate AI-derived language game must seem poetic. Legitimation is no longer a concern, because nothing produced by AI large language models is truly legitimate. AI gave the postmodernists their final victory over language and reality. All the language of the web can be condensed into a zip file, and from this everything, anything can be produced, and whatever that anything is means nothing in original legitimacy. The lack of substance and production devoid of value is truly postmodern.

This is why the remedy lies with the Romantics, before the cynicism of postmodernism rendered the counterpoint to excessive rationalism into an absurd subjectivist nothingness. Perhaps no single emanation of Romanticism better encapsulates the AI revolution than Mary Shelley’s ‘Frankenstein’. Victor Frankenstein’s creation, the Creature, is a product of scientific experimentation, yet his escape and exposure to the natural world filled him with awe and wonder, and his eloquent reflections portray his evolution of moral consciousness and convey that he does in fact possess a soul. The Creature’s articulate self-awareness and knowledge found expression in a form of humility, the knowledge that his maliciousness was born of his misery inflicted by Frankenstein and his ostracisation from community. When the Creature mused:

“And what was I? Of my creation and my creator I was absolutely ignorant...”

He may have been providing a forewarning. For what are we? If not ignorant of our creations. The AI monsters of our creation are born of the very same hubris that produced the Creature. But AI and large language models will never know the humility of the Creature because, unlike him, they are devoid of the ability to experience our world, to truly know of what it is they speak, and to add, as the Romantics did, the context of the human soul to knowledge.