We are born creative geniuses // but we usually lose our talent, how tech (AI) can/can’t help?
GenAI tend to produce less varied and more homogeneous outputs when trained on their own AI-generated data
Dr. George Land revealed at TEDxTucson that a creativity test designed for NASA showed that 98% of 4 to 5-year-old children exhibited genius-level imagination. However, this percentage significantly dropped to 30% at age 10 and a mere 12% at age 15. The findings suggest that traditional education may hinder creative genius, with only 2% of adults maintaining it. Land emphasizes the potential to regain creativity by distinguishing between divergent (imagination) and convergent (judgment) thinking and encouraging the former. He challenges individuals to reconnect with their five-year-old selves and exercise creativity daily, as demonstrated by an exercise involving improving a table fork.
The Failure Of Success // 13 years ago
What about AI creativity?)
AI models becoming less diverse and creative as they are increasingly trained on data generated by other AI models. It highlights research showing that language models and image generators tend to produce less varied and more homogeneous outputs when trained on their own AI-generated data. The consequences of this “AI eating its own output” are unclear, but it could lead to a collapse of AI creativity or necessitate measures to enforce diversity, such as marking AI-generated content or using more randomness in model training.
“Researchers say that, anecdotally, error-strewn AI papers are everywhere. “This is a widespread issue impacting many communities beginning to adopt machine-learning methods.” Some “argue that naive use of AI is leading to a #reproducibility crisis, while others “worry that ill-informed use of AI software is driving a deluge of papers with claims that cannot be replicated, or that are wrong or useless in practical terms.”
