@dragosroua
Here’s something that might ruffle some feathers: ChatGPT and astrology are doing the same thing.
Before the pitchforks come out, let me explain.
At its core, machine learning works like this:
- Set up input features (words, images, data points)
- Define desired outcomes (coherent text, accurate predictions, useful responses)
- Provide training baselines and examples
- Optimize a cost function to minimize the gap between inputs and outcomes
The result? A model that generates plausible outputs.
How about astrology?
- Takes in a set of input features (planetary positions) based on astronomically verified ephemeris data
- Maps these to outcome categories (abundance/scarcity, clarity/confusion)
- Refines the correlation through centuries of observation
- Minimizes the “cost function” between celestial patterns and human experience
The result? Interpretations that are plausible.
https://dragosroua.com/unpopular-opinion-chatgpt-is-over-engineered-astrology/