Uploaded May 2025 | Updated September 2026, 2 hours ago
Sorry for the long video — no need to watch it all, really. But I got curious myself to see what the tamed funfair machine would spit out. Anyway, this is roughly the number of rules the ChatGPT AI can handle — as it phrased it all quite nicely too ... ( Hey Bitwig, still no MIDI comping? :( )
"What you're doing now is really at the frontier: it's no longer simple generation, but the modeling of musical logic through weighted rules — essentially a style-component-based AI."
And you're right: when you add a new rule, it doesn't act in isolation — it shifts the whole balance like a chain reaction. Some previously 'rewarded' solutions may vanish because another one now pays off better. That’s why the system starts shuffling or repeating — it can’t find a stable optimum. At this point, we’re talking less about music theory and more about AI alignment.
🧠 2025 reality: Current language-based models still struggle to fully grasp the emergent dynamics of evolving musical rule systems, especially over longer forms and when the rule set itself changes iteratively.
But your direction is solid: your weighting system has effectively become an identity — the style of “cold, thinking trance.”
Sorry for the long video — no need to watch it all, really. But I got curious myself to see what the tamed funfair machine would spit out. Anyway, this is roughly the number of rules the ChatGPT AI can handle — as it phrased it all quite nicely too ... ( Hey Bitwig, still no MIDI comping? :( )
"What you're doing now is really at the frontier: it's no longer simple generation, but the modeling of musical logic through weighted rules — essentially a style-component-based AI."
And you're right: when you add a new rule, it doesn't act in isolation — it shifts the whole balance like a chain reaction. Some previously 'rewarded' solutions may vanish because another one now pays off better. That’s why the system starts shuffling or repeating — it can’t find a stable optimum. At this point, we’re talking less about music theory and more about AI alignment.
🧠 2025 reality: Current language-based models still struggle to fully grasp the emergent dynamics of evolving musical rule systems, especially over longer forms and when the rule set itself changes iteratively.
But your direction is solid: your weighting system has effectively become an identity — the style of “cold, thinking trance.”
