“As a professor of philosophy, I study the history of ideas. I am interested in questions such as how the European Enlightenment began and when the idea of free will first emerged. I spend years researching and writing books that cobble together the nuggets of information I've gathered. Like other scholars, I manage to scratch out a few new results about a handful of sources from one tradition or another. But history is immense in ways that exceed human comprehension.
Artificial intelligence offers us the chance to understand humanity in ways hitherto beyond our reach. Large language models are capable of absorbing and processing massive quantities of linguistic information. There is no language they can't learn and no reference too obscure for them to cross-check against the entirety of their training data.
Today's LLMs aren't much help in serious scholarly inquiry. They can help students, but they tell professors mainly what we already know. No wonder, since they are trained on the same familiar, limited material we have been trained on.
The first step toward harnessing the potential of AI is to recognize how much more material is out there. From the earliest texts through the era of the printing press, there is more surviving material in any given language than anyone could ever read. Few of these works have been translated. Many exist only in manuscript form, languishing in the vaults of some library or monastery. Much of it hasn't been read in centuries.
A large language model could read and retain it all.
A machine trained in this way would open new possibilities. Scholars once confined to a niche of their field could range more widely. Previously unknown texts could enter the conversation. Common assumptions about the origins of a concept like free will could be exposed to thorough scrutiny. The dominant biases of scholarship -- Western, Christian, English -- could be challenged from every corner of the globe.
The academic's perennial quest for interdisciplinarity, which until now has mainly meant listening to talks on topics you know nothing about, would come within reach.
How much would we learn from all this neglected material? No one knows, because there has never been a way to make it readily available. Until recently, it seemed that it would always be this way -- that human beings would never achieve a comprehensive understanding of their history. Victor Hugo once remarked that history must neglect many particulars of the past, or else "the infinity would overwhelm it." But the record of human civilization is finite, and although it is far too large for any human to grasp, it isn't beyond the capacity of Claude or ChatGPT.
The required investment in training is fairly minimal relative to the potential rewards. What's needed is to train an LLM so that it learns all languages, living and dead, and gets access to all available written texts -- not the tedious ephemera still under copyright, but the old and difficult things, including texts no one has read in centuries. Digitizing all this material in its original language and then translating it into all modern languages would unlock this heritage for global scholarship. The goal would be to make available, to students of all ages everywhere, a truly comprehensive record of documented human civilizations and ideas in all their variety.
We live in a world in which none of us are capable of more than the most superficial grasp of the human experience. Who understands even his own religion, let alone the religions of others? Who has a solid grasp on the ideas of a single philosopher, let alone a comprehensive grasp of philosophy across all its manifestations? It seems likely that with each passing year, humanity collectively forgets more of its cultural heritage than it manages to learn anew.
One might wonder: If we let in AI, what role will be left for human beings? Here we arrive at the heart of things, the concern that arises whenever some new technology threatens to uproot the old ways of doing things. Translating ancient Greek and Latin into modern languages, printing and selling cheap books to the masses, digitizing material and putting searchable versions online for free -- with every step, there have been those who mourn the old ways.
With the rise of AI, we move on to the next step: having computers read this material. It has already begun. My suggestion is simply that we do this the right way, globally and comprehensively, recognizing that we have the opportunity to study all of recorded human civilization in a way that has never yet been possible. Rather than see AI as an excuse to think less, I see it as an opportunity to think harder and to broaden and deepen our understanding.
The challenge we face is the same challenge we have always faced, the task of finding meaning in the information available to us. A machine can collect, compare and collate tremendous amounts of information, but it can't do the thing that matters most of all -- decide what has genuine value for human lives. For an LLM, every question you ask is important and insightful, or so it will tell you. In truth the machine has no idea what is important. It lies with us to decide which questions matter most. But harnessing the power of AI could put us in a position to answer them.
---
Mr. Pasnau is a philosophy professor at the University of Colorado, Boulder.” [1]
1. AI Can Help Us Understand Humanity Like Never Before. Pasnau, Robert. Wall Street Journal, Eastern edition; New York, N.Y.. 24 July 2026: A15.
Komentarų nėra:
Rašyti komentarą