The Oracle of Delphi famously pronounced Socrates as the wisest in all of Greece, and asserted that the full synthesis of human wisdom could be expressed in the dictum: know thyself. The Ancient Greek precinct is clouded in myths and legends all dealing with consultations, wonderment, and questions about life. It holds a place in our collective historical consciousness for that reason, and is maintained by the undying curiosity to seek wisdom and knowledge.
One of the modern candidates for Delphi is certainly large language models (LLMs), which aim to compress the totality of human knowledge into an interactive system. This kind of generative artificial intelligence (AI) continues to improve, and the development of artificial general intelligence (AGI) still remains the objective of technocrats. These LLMs are a sponge for information, holding it passively and waiting for the user to inquire about a random subject or situation. The accumulation of knowledge in this way presents many, many issues, one of which concerns the notion of wisdom. Will AI be the wisest source of the answers to life’s arcane questions?
There are several views about the concept of wisdom. Generally, the colloquial version of the word can be said to fall under a broad category of “accumulated experience”. Other perspectives on wisdom, such as rationality, epistemic accuracy, epistemic breadth, and the ability to channel this knowledge into a philosophy of living well are all important. Regardless of one’s own account of wisdom, the access to a technological sage is now a reality. Students already use LLMs in astonishingly high numbers, AI can perform certain medical analyses better than doctors, and this heightened integration of advanced technology into the foundation of daily life shows no signs of slowing down.
But how does wisdom typically accrue? One way is certainly to seek it from others—there are plenty of proclaimed experts out there. The assumption here is that objects of truth exist which are mind-independent, suspended in the air awaiting contact. Once grasped, the subject can then incorporate the insights into their life. An LLM would be the modern portal for this process, taking the vast ocean of human knowledge and history into consideration before it replies with advice to a user.
There seems to be something dubious about the above method, though. The wisdom seems a bit unearned. To go around collecting life knowledge a-la-carte seems disingenuous at best and fraudulent at worst, especially if it were only gathered by means of an LLM. Immanuel Kant claimed (as did Socrates, in a similar way) that wisdom could not be “poured into a man by others” but rather it must originate from within oneself (7:200).1 This declaration helps understand the previous skeptical intuition about LLMs—there is a lack of identity in the hunt for wisdom in such a way. Which suggests that someone must make experiential contact with the ideas themselves in order to properly espouse the beliefs.
Herein lies one of the crucial considerations for a future with AI—how will it impact the practical side to wisdom, the part that entails living?
Kant is an interesting case study here. He spent nearly all of his life in his hometown of Königsberg, adhering to a systematic routine of reading, writing, walking, and dinner parties. He wrote extensively about love, beauty, and the sublime, but was never married and probably never saw a mountain. Yet his philosophical output was extraordinary. His conduct paints an image opposite to the modern world traveler who seeks to imbibe all possible life experiences in the unsuspecting corners of every culture, every country, in an attempt to embrace all of existence first-hand. This spectrum of theoretical and practical wisdom is meant to be bridged, Kant wrote. He proclaimed that wisdom is to apply the “practical use of reason that conforms perfectly with the law” (Ibid). It is to take the moral law found in the noumenal world and actively live in accordance with it in the phenomenal world.
To do this, he outlined three precepts. The first is to think for oneself. In the 1784 essay “What is Enlightenment?”, Kant wrote about the condition of “self-incurred minority”, which is when someone is not independent in thought, but rather depends on external sources for their understanding of the world (8:35). To free oneself of this inhibition is rightfully one of the most valuable undertakings one can pursue as it is an emancipation from superstition. A serious threat is presented by LLMs in this regard. If wisdom is to originate from within the subject, does assistance from AI not violate this provision?
The second stipulation from Kant is to develop a sensus communis, or “community sense” in one’s thought. This broadens and forms objective thought rather than subjective. The former is in service of humanity and “reason as a whole” whereas the latter is private and self-serving (5:293). It is not readily apparent if an LLM could cultivate such a mindset, because it would depend on certain programming decisions.
Lastly, Kant said that consistent thought with oneself is necessary for a certain degree of wisdom. One should strive to eliminate contradictions and live in accordance with one’s beliefs and judgments. Practically, the discipline of one’s life should be such that the guiding principles can be adopted by others. Notice how this assumes the fulfillment of the second provision already—one’s actions should not be solely attached to the benefit of an individual, but of humanity.
This trinity of requirements reveals almost too many suppositions at play. The crux is that wisdom entails certain metaethical commitments that LLMs do not have baked in during their design. For instance, a Kantian ethical framework entails a generalist approach, which is that moral rationality depends on a suitable provision of principles. Such principles are rather rigid and unforgiving. Ethical maxims do not admit of exceptions. With this worldview, LLMs could plausibly offer the advice that one ought to always keep promises. Full stop. Another option is a particularist approach, which contends that moral principles are not sufficient for judgments, and might not even be defensible in the first place. Rules should not be applied to ethical cases because they do not sufficiently cover the various features and particular qualities of all moral conundrums. Proponents of this view could most likely conjure a scenario in which it would be acceptable not to keep one’s promise.
Which stance would an LLM take? Would all versions of AI offer the unequivocal principle, or would the nuances of certain cases be taken into consideration? The undetermined and ongoing debates in metaethics are merely one example of how wisdom is a fickle concept, and may be difficult to take the form of universals. This suggests a return to the individual, and a return to the idea that wisdom must emanate from within.
To view (charitably) AI as the amalgamation of the best and wisest views on life is to neglect the practical component of reason. Even if one cognitively carries around a true fact about the world, there is an intuitive concern about how one obtained it. Wisdom implies an incurred cost, along with several markers of legitimacy according to Kant. The advances of AI continue to improve the efficiency of life and can serve as great companions for humanity. But if Kant’s prized value of autonomy is to be taken seriously, thinking for oneself should maintain a paramount feature of wisdom. To have a direct link between one’s judgments and actions does not allow for a technological middleman, an intermediary in the quest for knowledge. Instead, the challenge is to summon the courage to deploy our own use of reason.
From a Kantian perspective, the former acknowledgment is the point of departure from a reliance on external influences on one’s thought, which cripples the strive for wisdom. Philosophical inquiry is born of the recognition that none of us have our own oracle, and it is an individual expedition to unriddle the composition of the world.
The brief references to Kant’s work will use the abbreviations corresponding to the German Academy edition: Kants Gesammelte Schriften.