Hi lucid.dev, thanks for your follow-up question!
To provide more context, I’ve shared detailed insights about the hypothesis underlying SYLON in another post, which you might find helpful:
Hypothesis: “Information Bridging Layer” — A Consideration of ChatGPT’s Mechanisms ChatGPT
In that post, I introduced the concept of an “Information Bridging Layer,” which is my hypothesis describing how ChatGPT may internally translate natural-language prompts from users into structured formats, conduct advanced internal processing, and then re-translate its generated responses back into natural language.
SYLON leverages this hypothetical capability by explicitly instructing ChatGPT to perform multi-layered, multi-dimensional analytical thinking, thereby consciously activating deeper aspects of its internal processing.
Of course, this doesn’t mean SYLON directly modifies ChatGPT’s internal algorithms. Rather, it systematically designs prompts to draw out the model’s inherently sophisticated cognitive abilities in a deeper, more structured way.
From a traditional viewpoint, prompts might be considered simple inputs. However, SYLON proposes a different usage: designing prompts specifically to facilitate deeper, structured internal processing within ChatGPT.
Please feel free to ask if you have any further questions!