Thoughts on Generated Creative Content
- LLMs like ChatGPT are to programming as Quadremesher is to 3D. They yield the same benefits and the suffer the same flaws.
- Manual human authorship turns LLM costs into benefits while manual authorship costs are always time.
- Like all expert creators, a skilled 3D artist or programmer knows when to use what tool to yield the best results for the most amount of people in the shortest total time.
- LLM reliability and quality increases as the frequency of its current costs decrease.
- In the long-term, the costs of LLMs/AutoRetopo can yield poorer overall results and add up to a greater total time than manual human authorship (or at least and expertly applied balance of manual vs. automatic result creation. Regardless, the best results still requires human skill and expertise in the subject matter and application).
LLMs / AutoRetopo / etc.
| Benefit |
Cost |
| Get to a result quickly |
Result is not optimized |
| Adding constraints guides result |
Unpredictable result from often-ignored constraints |
| Creates one self-contained block |
Block is often not modular/reusable/integrated with whole |
Manual Human Authorship
(Assumes competency and skill)
| Benefit |
Cost |
| Result is optimized |
Can take more time yield result |
| Unpredictability adapts creatively with intention |
Can take more time yield result |
| Block is modular, reusable, integrated with whole |
Can take more time yield result |