M: One of my largest weaknesses is that when making decisions for other people, I can be very good at giving objective guidance – but when choosing for myself, I go into a vapor lock and cannot seem to overcome this very often without an external force active. Knowing this, I’m wondering if I could create a machine learning clone of myself. something that I could ask my own advice of and try using as an external force on those occasions when I can’t or won’t ask someone else. My first direct question is: electronic and software elements aside, what would this entail? What would I have to do in order to create a acceptable electronic version of myself? I know that it’s not an exact copy, so you don’t have to respond as though I am demanding certainty. That will come later. Just give me the information without waffling at the moment, knowing I won’t hold any of it to be the 100% truth.
G: To create a functional ML-based clone of yourself as an external advisor, disregarding the technical/software/hardware details for now, you would need to:
1. Define the scope of "you":
Decide what aspects of yourself you're trying to replicate. For your
goal (decision guidance), that likely includes:
2. Gather or create representative data:
To train or tune anything ML-like to mimic your thinking, you’ll need
examples. This would require:
3. Systematize your values and patterns:
Even without a model, you'd benefit from building a structured
decision template that represents how you break things down.
Example components:
This structure can later help guide your AI clone or just become part of a prompt framework.
4. Define the rules for "externalization":
If the clone is an external force to “break vapor lock,” it needs:
5. Handle contradictions in identity:
You’ll need to address:
6. Set ethical boundaries:
If the clone starts reflecting your worst impulses, fears, or anxieties,
how will you handle that? You may want a fallback framework:
7. Create trust calibration tools:
Since the clone will be giving advice as if it were you, you’ll
need ways to:
That’s the structural, conceptual framework. Once you have those pieces, you can think about building datasets, choosing models, writing scripts, etc. Want to go into those next?
M: Are there existent series of questions that would facilitate the creation of this type of ML clone? Or would this be an on the fly project?
G: There’s no turnkey question set designed specifically to build an ML clone of your decision-making style—but you can repurpose or adapt several existing frameworks to extract the necessary information. You’d likely need to combine and customize them. Here’s a breakdown:
These are the closest to what you're aiming for. They help externalize internal reasoning:
Use case: Structure your decision logs and simulated cases. These help generate consistent training examples.
From therapy and coaching domains. Useful for surfacing subconscious patterns:
Use case: Helps model your internal resistance points, blind spots, or emotional short-circuits.
From moral philosophy, AI alignment, or character-building programs. Example prompts:
Use case: Captures your core moral calculus—useful for aligning the clone with your long-term principles.
Things like:
Use case: Provides priors for tone, reasoning tolerance, and cognitive style.
These are custom and need to be generated:
Use case: This will form the core of the training material. Even 30–50 high-quality examples could bootstrap a decent fine-tune or ruleset.
Would you like a first draft of such a question set, focused on your use case (decision guidance for self-stuck moments)?