Raw LLM Responses

What’s truly insane is that anyone who is truly knowledgeable or even cares about AI would be fighting for the preservation of original art sources and the proper labelling of AI art. It’s well known that training models on AI output creates a deteriorating spiral for any AI model. The more AI output gets passed off as source training data, the quicker a model training on it will suffer. Don’t need to take my word for it, just ask an AI about the studies and about what will happen to them if they can no longer find verifiable human-generated content. Ask them to project the next ten years if original artists were unable to sustain their craft and discouraged from passing on their techniques. Every model I have talked to about this issue ended up advocating some form of licensing or attribution in order to preserve the provenance of training inputs. Open-AI based models were more worried about potential overhead on average, and risk of liability, while Alphabet models seemed to think the situation was urgent and that the net benefits of a properly structured licensing system would eventually outpace the cost. All models I queried (I didn’t talk to Grok) came to the conclusion that some form of fraud detection and prevention needed to exist to prevent model collapse. Generative AI masquerading as human content is a bigger threat to them than to us.