Brilliant post Jim, full of wonderful insights and observations. Going to quote this: “I feel a duty to point out that contrary to whatever AI evangelists like to say, generative AI is not creating anything new or unique or imaginative in any way, it is just regurgitating and hallucinating (a metaphor for random failure to produce expected results) different data distributions that build images pixel by pixel from billions of original images found in its diffusion training model.”
A practical consideration: Some of these initiatives to 'counter' the AI trend seem rather counterproductive. If you label your work as 'Zero AI', you all but ensure its usage in the next batch of training data. The scrapers will thank you for your help (while maintaining plausible deniability, of course).
Brilliant post Jim, full of wonderful insights and observations. Going to quote this: “I feel a duty to point out that contrary to whatever AI evangelists like to say, generative AI is not creating anything new or unique or imaginative in any way, it is just regurgitating and hallucinating (a metaphor for random failure to produce expected results) different data distributions that build images pixel by pixel from billions of original images found in its diffusion training model.”
Insightful article. Big fan of AI baby peacocks.
A practical consideration: Some of these initiatives to 'counter' the AI trend seem rather counterproductive. If you label your work as 'Zero AI', you all but ensure its usage in the next batch of training data. The scrapers will thank you for your help (while maintaining plausible deniability, of course).
This is for sure one of the main drivers of C2PA. It’s not even useful for its professed purpose as watermarks are easily removed
Fox guarding the henhouse.
Say, cheesy!
Well written as usual.
Thanks for reading, Joe!
🧀