"Understanding a sentence means understanding a language. Understanding a language means understanding the form of life to which it belongs."
- Philosophical Investigations
In this series, we are diving into the history of epistemology to find eternal lessons for the AI age. These ideas have shaped Novela's digital marketing simulations.
Ludwig Wittgenstein is one of the most influential philosophers of the 20th century, yet also one of the least understood. That very paradox is very Wittgensteinian.
He can be intimidating to read. His work covers a variety of complex concepts and he writes in an aphoristic, almost mystical style.
Check out the opening to his Tractatus Logico-Philosophicus:
Wittgenstein discusses concepts that he later tells the reader cannot possibly be discussed with any substance. He aims for the purest precision, yet all too often bewilders with his selection of words.
"We are unable to give any account of the connection between 'rule' and 'application,' because we can describe this connection only by using the very terms we should have to define." Remarks on the Foundations of Mathematics
Nonetheless, I believe Wittgenstein has something profound to say to us in the AI age in two key areas:
What it will take for AI to understand language, beyond parroting what it finds in its training corpus.
Where we should focus our attention to develop new, valuable skills.
In this article, we will explore how Wittgenstein's key philosophical theories highlight the skills needed to thrive in the AI age. We will also discuss the limitations of generative AI in light of these theories.
Three key ideas from Wittgenstein for the AI age:
Language Games: In his Philosophical Investigations (published posthumously) Wittgenstein posited that language is not just a system of communication, but a series of "language games" rooted in shared practices, cultural norms, and experiences. This was a significant shift of position from his earlier work, the Tractatus. By participating in these language games, we develop an understanding of meaning and context that goes beyond the mere correlation of words with objects in the world. Senior Google engineers have cited this idea as an inspiration behind some of their major search algorithm overhauls, including Hummingbird.
Ladder of Understanding: Wittgenstein famously compared the process of understanding to climbing a ladder. As we ascend, we gain new perspectives and insights, eventually discarding the ladder once we've reached our destination. This suggests that understanding is not static, but rather an evolving process that continually adapts to our experiences and interactions.
Non-theoretical Understanding: Wittgenstein's philosophy highlights the importance of non-theoretical understanding, which involves seeing connections, recognizing patterns, and grasping subtleties in human experience. This kind of understanding is distinct from scientific, data-driven understanding, and is essential for comprehending concepts like art, music, and human emotions. In essence, he believes that we have committed a serious mistake by believing we can contain the world within scientific theories. I think it's clear what he would have made of ChatGPT's attempt to master language. On the other hand, he would be dismayed by our underestimation of the importance of non-theoretical understanding. It's not too late to change, however.
Why these ideas matter for knowledge acquisition
In the AI age, embracing Wittgenstein's ideas can help us understand the limitations of technology and foster skills that complement the capabilities of AI systems, such as empathy, cultural awareness, and contextual understanding.
By reducing the world to a purely theoretical understanding - that is to say, a world of half-baked hypotheses applied to irrelevant questions with unreliable data - we make ourselves vulnerable to replacement by AI. If we instead take a broader view and apply this scientific thinking only to appropriate questions, we can apply new frameworks to uniquely human domains.
By participating in language games and cultivating non-theoretical understanding, we can develop a deeper appreciation for the complexities of human experience and communication, which will prove invaluable in an increasingly interconnected world. Furthermore, recognizing the dynamic nature of understanding can help us become lifelong learners, adapting and evolving alongside AI technologies.
Limitations of generative AI
If Wittgenstein's theories hold true, generative AI may face certain limitations in fully understanding and interacting with humans. These limitations include difficulties in grasping context and culture, discerning genuine emotions and intentions, capturing non-theoretical understanding, fostering true creativity and innovation, and navigating ethical complexities.
These limitations should not be seen as insurmountable barriers, but rather as areas where AI researchers and developers can focus their efforts to create more human-centric AI systems.
Wittgenstein's philosophical theories, while complex and often abstract, can offer valuable insights into the skills needed to navigate the AI age and the potential limitations of generative AI. They are also instructive for those of us developing online learning experiences. Lessons must be tailored to a user's context and each pathway adapted to their ongoing skill-based needs. We must build tools to help people spot the connections between concepts, instead of blindly following templates.
Generative AI can play a role in helping us achieve those goals if we know where and when to apply it.
ChatGPT might seem to understand our language, but it does not understand the form of life to which it belongs.
"Don't, for heaven's sake, be afraid of talking nonsense! But you must pay attention to your nonsense." Ludwig Wittgenstein