Search Results
6/24/2025, 10:33:11 PM
>>508618514
>it is simply pasting other peoples work and mashing it together
It's more complicated than that. A better way of thinking about it would be to think of training data (images, text, whatever) as data points on a graph, and an LLM is the function you try to fit to the data. The model tries to fit the provided data and the model can be used to interpolate approximations that reflect the data from a particular input, but it doesn't generally duplicate exact results, just a "closest fit".
The more you narrow or massage the data, of course, the closer the model replicates it exactly - it's why most people who use LLMs to produce high quality output go through many iterations or use things like Low Rank Adaptation (LoRA) models made from smaller, more refined datasets to fine-tune output.
>it is simply pasting other peoples work and mashing it together
It's more complicated than that. A better way of thinking about it would be to think of training data (images, text, whatever) as data points on a graph, and an LLM is the function you try to fit to the data. The model tries to fit the provided data and the model can be used to interpolate approximations that reflect the data from a particular input, but it doesn't generally duplicate exact results, just a "closest fit".
The more you narrow or massage the data, of course, the closer the model replicates it exactly - it's why most people who use LLMs to produce high quality output go through many iterations or use things like Low Rank Adaptation (LoRA) models made from smaller, more refined datasets to fine-tune output.
Page 1