Why DeepSeek's new AI model thinks it's ChatGPT

Awww. Poor LLM companies worried about competing companies creating models based on … data “stolen” from their own model? Totally makes sense, since the original LLM was conceived without data, kind of like the Virgin Mary. Let’s just all pretend like OpenAI didn’t steal terabytes worth of data themselves in violation of a whole bunch of terms of services.

“Like taking a photocopy of a photocopy, we lose more and more information and connection to reality,”

This is the Achilles heel of LLMs.

With the proliferation of models, training new LLMs on the output from older LLMs becomes all but impossible to avoid.

It's the AI equivalent of inbreeding --- where defects are inevitably magnified over time.

To be the devil's advocate, what if the modern corpora that DeepSeek is trained on encourages this response? There has been a couple of years where AI responses to prompts have proliferated through the internet at large.