“What’s the difference between having a conversation with Socrates vs ChatGPT?” If treated like an interlocutor rather than merely an answer machine, might it be possible for modern students to engage with AI in much the same way that Athenian youth might engage with Socrates? Which is more likely to “corrupt the youth”?
This line of questioning was raised by Dr. Christina George during a Q&A session at Baylor’s recent AI and Faith Symposium. Baylor’s Institute for Faith and Learning invited scholars from across the nation to deliver over 200 presentations on AI over three days. Topics ranged from whether AI should be used in elderly care to explanations of mathematical processes behind artificial “intelligence.” Throughout all the presentations I attended, Dr. George’s question followed me. In their own way, each presenter helped me to understand the unique value of human conversation.
In her presentation, Dr. George offered the beginning of an answer: “Socrates allows for his own mind to change throughout the course of many of his dialectics!” AI, however, does not have any mind to change.
Sam Byassee, a Baylor senior who chaired the Crane Scholars panel at the symposium, described AI as “a conversation partner who collates information for you to sift through.” You can use AI to enable your own thinking, but if “you want someone with whom to contemplate the realities about which you are speaking, then there is going to be a stark difference between the kinds of conversations you can have.” Since AI does not think for itself, it can never be the kind of partner, like Socrates, who grows alongside you.
However, this is not the kind of difference we can calculate and measure. Sam Byassee observed “there is something demonstrably different” between human and AI conversations, but this might not be something that “shows up directly on any of our observation charts.” Byassee’s comment expresses a common theme that I observed during the conference sessions I attended: it is clear that humans and AI are different kinds of things, even sharing certain capacities, but there is great difficulty in locating exactly where we are different.
Dr. Ian Gravagne, associate chair of Baylor’s school of engineering and computer science, gave a presentation that helped me to better understand exactly what we mean by “intelligence” when we apply that word (rightly or not) to AI. AI processes human language by turning human words into numbers (a lot of them!) through a process called “tokenization.” Then these number representations of words are assigned specific locations within a “vector space.” Through this process, AI can make mathematical associations between words. For example, the input “teacher + PhD” can output “professor” because the inputted words are related to one another insofar as they are situated near one another in the vector space. When “added together,” they direct the AI to the vicinity of “professor.”
Granting that both humans and AI possess a capacity for respective kinds of intelligence—the latter being a purely calculative kind—then the question of Socrates vs ChatGPT reveals itself as far more complex than attributing greater reasoning capabilities to one or the other.
“The fact that AI cannot be interested is perhaps one of its greatest benefits to humanity as well as one of its greatest dangers.”
In his presentation, Dr. Phillip Donnelly of Baylor’s Great Texts Department compared AI to the oracles of the ancients: Just as the ancients went to an oracle for answers and had to decide whether to trust the oracle’s advice and how to interpret it, so also moderns go to AI for a kind of “prophecy”—though this prophecy is generated by crunching numbers rather than communion with gods.
Dr. George noted that we are quick to accept what AI tells us as if it is plain fact: “It is not always approached with the same skepticism that we bring to a Google search,” or in Dr. Donnelly’s analogy, the same skepticism one might want to bring to an oracle.
One feature of Socratic dialogue is that Socrates is interested in the welfare of his interlocutors. In fact, Socrates, as history records, was perhaps too interested. By contrast, AI is, in Dr. George’s words, “not interested, whatsoever, in the implications that result from its ‘teaching.’”
The fact that AI cannot be interested is perhaps one of its greatest benefits to humanity as well as one of its greatest dangers. Insofar as AI is disinterested in people and in itself, it can be used to perform those tasks that humans might not want to do themselves—cleaning, caretaking, etc.—or would perform more effectively without error. What societal operations would we prefer be performed by humans with interests vs disinterested AIs? Do we prefer that Socrates “corrupt the youth” or ChatGPT?