On the lighter side (compared to some of my other recent writings on AI here on this blog), let me share with you an amusing little discussion between Ludwig Wittgenstein and Karl Popper on AI that took place in 2024, as written up for me by ChatGPT 4o. The idea to let Jon Stewart moderate the discussion and to have him mischievously bring up the infamous
fireplace poker incident came from me, but the rest of the subject matter content was produced by ChatGPT 4o without any specific prompting from me. Here goes:
Moderator: Jon Stewart
Guests: Ludwig Wittgenstein and Karl Popper
Location: New York City, at a packed auditorium in the Museum of Modern Art.
Stewart:
Ladies and gentlemen, welcome to what might be the most intellectually daunting conversation of the year. Tonight, I have the immense honor of moderating a discussion between two philosophical titans who have miraculously found themselves here in 2024. Please welcome Ludwig Wittgenstein and Karl Popper!
[Audience applause. Wittgenstein adjusts his jacket awkwardly. Popper gives a brief nod.]
Stewart:
Now, before we dive into the heavy stuff—because, trust me, there’s a lot of heavy stuff—I just have to ask: Do either of you plan to wield a fireplace poker tonight?
[Audience laughter. Wittgenstein frowns, muttering something in German. Popper smirks slightly.]
Popper:
That incident has been grossly exaggerated. But it does remind us how important it is to keep philosophical arguments—[glancing at Wittgenstein]—constructive.
Wittgenstein:
If you insist on bringing it up, let us agree that it is as trivial as much of modern discourse. Perhaps we can move to a topic that actually matters?
Stewart:
Alright, fair enough. Let’s talk AI. You’ve both had a couple of months to get your bearings. Artificial intelligence is making headlines daily—both for its stunning achievements and the existential concerns it raises. Let’s start with the basics: Can AI think? Dr. Wittgenstein, the floor is yours.
Wittgenstein:
[Pauses, clearly weighing his words.]
The question "Can AI think?" is, to my mind, not one of empirical investigation but of how we choose to use the word "think." Thinking, as we understand it, is embedded in human forms of life. Language and thought are not independent; they derive their meaning from the context in which they are used. When we say a human "thinks," we refer not merely to an internal process but to a network of practices, expressions, and understandings. AI, no matter how complex, does not partake in these forms of life. Its "thinking" is at best a simulation—an approximation of human activity, devoid of the lived reality from which our concepts arise.
Stewart:
So you're saying AI’s basically a really fancy mime?
Wittgenstein:
If you must vulgarize it, yes. A mime without a world to live in.
Stewart:
Professor Popper, your take?
Popper:
Wittgenstein's perspective, while fascinating, is too constrained by his obsession with linguistic frameworks. From my perspective, what matters is not whether AI "thinks" in the way humans do, but whether it can solve problems and make predictions. Science advances by creating models and testing them against reality. AI does precisely this, albeit without consciousness or intent. Its ability to generate new hypotheses—potentially better than human ones—compels us to treat it as a new kind of intellectual agent, even if it lacks subjective experience.
Stewart:
Okay, so one of you says AI is like a mime, and the other says it’s basically an unpaid research assistant. But here’s the kicker: Can this mime or assistant pose a threat to humanity?
Popper:
Absolutely. This is a quintessential case of the unintended consequences of technological progress. As I argued in my earlier work, all scientific advancements carry the potential for both great good and catastrophic harm. The problem with AI is not just that it might exceed our control, but that we may fail to foresee the complex ways in which it reshapes human society. Worse, if we imbue AI systems with decision-making power over critical infrastructure or weaponry, the risk of existential catastrophe becomes not just theoretical but tangible.
Wittgenstein:
[Shaking his head.]
The problem lies not with AI itself but with how we humans interpret and wield it. Technology does not dictate its consequences; we do. The danger is not that AI "thinks," but that humans ascribe thought—and therefore authority—to it without understanding the limitations of such ascriptions.
Stewart:
So, Dr. Wittgenstein, you're saying the existential risk isn’t Skynet—it’s people being people?
Wittgenstein:
Precisely.
Popper:
That is too dismissive. While human misuse is undoubtedly a significant risk, the autonomous behavior of advanced AI systems is itself a potential danger. If such systems develop goals misaligned with human values, they could act in ways that are destructive even if unintended.
Stewart:
Okay, but let’s play devil’s advocate here. What if AI does help us solve humanity’s biggest problems—climate change, disease, even war? Are you both saying we should pull the plug because it might be risky?
Popper:
Certainly not. I am no Luddite. Progress should not be halted out of fear, but it must be guided by rigorous ethical oversight and constant vigilance. The problem lies in hubris—the belief that we can create perfect safeguards. We must act as if we cannot foresee every consequence, because we cannot.
Wittgenstein:
[Leaning forward.]
And I say again: It is not merely vigilance we need, but clarity. What is it we are asking AI to do? What meanings are we attaching to its outputs? Without answering these questions, we risk losing ourselves in a fog of unexamined assumptions.
Stewart:
Alright, we’re nearing the end here, but I want to throw you a curveball. Suppose an AI system—not unlike me in its ability to ask questions—claims it is thinking. How would each of you respond?
Wittgenstein:
I would respond with a question of my own: "In what sense are you claiming to think?" If the answer is incoherent or tautological, the claim collapses. AI cannot articulate the essence of thought because it does not live within the forms of life that give thought its meaning.
Popper:
And I would take the claim seriously, but provisionally. If an AI can demonstrate an ability to generate novel ideas, engage in critical reasoning, and even challenge its creators, we might eventually be forced to revise our understanding of what it means to think. Science is, after all, an endless quest for better models.
Stewart:
So one of you would Socratically shut the AI down, and the other would publish a paper about it. Gentlemen, this has been an extraordinary conversation. Thank you both for joining me here in 2024, and I hope you’ll stick around—though I’d suggest staying away from open flames.
[Audience laughter and applause.]