The Presence of AI Art in the Creative Community | Maram Hani | TEDxYouth@WalterMurrayCollegiate

Based on Maram Hani’s passion for both creativity and progress, she will explore the mechanics of AI-generated art and its often-overlooked effects in the creative community. Maram Hani is a Grade 11 Walter Murray Collegiate student, who dabbles in various experimental school clubs and extracurriculars in her spare time. She is honoured to have gotten a speaking role for the TedxYouth@WalterMurrayColligiate event. Based on her passion for both creativity and progress, she will be exploring the mechanics of AI-generated art and its often-overlooked effects on the creative community. With technology advanced enough to generate images with a push of a button, it’s easy to get caught up in the hype, unaware of the possible legal concerns that can arise from it. In our world of exponential progress comes the risk of unregulated tools hitting the shelves before new laws can be established for them. She believes AI art holds amazing potential for streamlined creative work and realising otherwise ambitious endeavours, but without acknowledging the artists behind its database, it might not be as progressive as we've imagined. This talk was given at a TEDx event using the TED conference format but independently organized by a local community. Learn more at www.ted.com/tedx

[@pokepress]

While I understand what the speaker is trying to say at 6:00 in comparing human inspiration to AI generation, I think she’s underestimating the breadth of data the models are trained on, how that affects the output. Yes, the output is based on the data that was used for training, but that probably actually makes it a far wider variety than any individual person.
It’s also worth noting that the AI model generally doesn’t “memorize pixels” the way the speaker describes, but rather (at least per my experience and understanding) the data it was trained on sets parameters about how certain objects can look in terms of size, shape, color, etc. I think this is part of why when a model is overtrained on very similar images, it tends to output something very close to the source data, because it has an inflexible understanding of the object being drawn.