(no title)
d-mz | 6 months ago
AI isn't a solution to this; on the contrary, whatever insufficiency exists in the original data set will only be amplified, as compression artifacts can be amplified in audio.
We also can't trust any data that's created recently, because an LLM can be trained to provide correct-looking structured data that may or may not accurately model semantic structure.
The best source of structured, tagged, contextually-rich data suitable for training or searching in the future will be from people who currently AREN'T using generative AI.
No comments yet.