Are We Just Predictable Scripts? The Dilbert Dilemma
In an episode of
Dilbert, the protagonist unwittingly has a conversation—not with his mother, but with
a recording of her usual responses. She had pre-recorded their exchanges because Dilbert was
so predictable, his side of the conversation could be anticipated without her present.
Even when Dilbert realized he was talking to a recording,
it kept responding accurately to him, as though their conversation were still fluid and natural.
This presents a stunning thought experiment:
how could Dilbert prove he was talking to a recording? What test could he run to confirm whether the responses were truly dynamic or just well-constructed predictions?
More intriguingly—does this apply to AI today?
Conversational AI and the Illusion of Free Will
Much like Dilbert’s situation, conversational AI models today operate by
predicting the most rational human responses to any given input. Language models don’t think like humans—they don’t strategize or innovate in a goal-seeking manner. Instead, they mimic
patterns from human language, selecting the most statistically probable answer rather than constructing thoughts independently.
This leads to an unsettling possibility:
what if AI only looks intelligent because humans themselves follow predictable scripts?
Consider daily conversations:
- “Hey, how’s the weather?” → “It’s cold today!”
- “What’s up?” → “Not much, just working.”
- “Good morning.” → “Good morning!”
These exchanges
follow patterns, meaning AI can simulate them convincingly. If human interaction is largely scripted by cultural norms and routine behaviors, then
AI can pass Turing tests not because it is truly intelligent, but because humans are naturally predictable.
What If Reality Is Pre-Recorded?
Now, let’s push the thought experiment even further.
What if reality itself were
a pre-recorded sequence, shifting based on predicted human decisions?
Imagine walking into a store. You intend to buy soda. If this reality were pre-recorded, it wouldn’t matter what you
think your decision is—the moment you enter, the world aligns to your expected behavior. Your choice isn’t truly a choice, it’s
merely selecting from predefined script variations, much like an AI model flipping between likely responses.
This would mean:
- Your actions don’t actually change the world—they merely select between scripted paths.
- Free will becomes an illusion, with reality adapting dynamically to what was already expected of you.
Much like AI, your perception of agency might only exist within predefined rails.
Two-Face’s Coin and Predictive Reality
This dilemma mirrors
Two-Face’s coin-flip philosophy in
Batman. He relies on
a simple binary system—heads or tails—as though fate is predetermined, and his choices aren’t truly choices.
But what if the
coin itself were rigged—designed to always fall in alignment with his expected behavior? Much like a pre-recorded reality, the illusion of randomness would remain intact while the result
still adheres to a predictable script.
This aligns with AI challenges today—how do we distinguish
true intelligence from pattern-based prediction?
AI’s Path Forward: From Predictability to True Goals
If AI only
mimics human conversational behavior, then
what’s missing for a true “goaled AI” to emerge?
Currently, most AI systems are
reactive rather than proactive. They respond dynamically but don’t independently pursue objectives across long timeframes. This is where
heuristic-driven AI could mark the next evolutionary step.
Instead of relying purely on
LLM-driven responses, an advanced AI system could integrate
modular skill sets, combining:
Task-Specific Heuristics → Structured problem-solving methods for precise execution.
Speech & Hardware Integration → STT (Speech-to-Text), TTS (Text-to-Speech), and real-world interfacing (Arduino control).
Adaptive Skill Selection → AI switches between specialized expert modules rather than loosely improvising responses.
