Version 1
: Received: 9 June 2022 / Approved: 10 June 2022 / Online: 10 June 2022 (04:35:14 CEST)
How to cite:
Hoorn, J.F. Erroneous AI-classification Systems Produce Creative Design Ideas, Predicting Level of Innovation Value . Preprints2022, 2022060148. https://doi.org/10.20944/preprints202206.0148.v1.
Hoorn, J.F. Erroneous AI-classification Systems Produce Creative Design Ideas, Predicting Level of Innovation Value . Preprints 2022, 2022060148. https://doi.org/10.20944/preprints202206.0148.v1.
Cite as:
Hoorn, J.F. Erroneous AI-classification Systems Produce Creative Design Ideas, Predicting Level of Innovation Value . Preprints2022, 2022060148. https://doi.org/10.20944/preprints202206.0148.v1.
Hoorn, J.F. Erroneous AI-classification Systems Produce Creative Design Ideas, Predicting Level of Innovation Value . Preprints 2022, 2022060148. https://doi.org/10.20944/preprints202206.0148.v1.
Abstract
In the mid-layers of Deep Learning systems, clustered features tend to fit multiple classifications, which are filtered out during the final stages of object recognition. However, many misclassifications remain, regarded as errors of the system. This paper claims that tagging an entity incorrectly for reasons of similarity is evidence of spontaneous machine creativeness. According to the ratings of 40 design educators and researchers, AI-generated false-class inclusions produced creative design ideas, predicting the level of innovation value. These designers were not just anybody but came from a design school in Asia with a top position on the world ranking-lists. They entered an experiment in which 20 classification mistakes were framed as early-design ideas that were either human-made or intentionally suggested by creative AI. Many examples passed the Feigenbaum variant of the Turing test with a conceptual preference to creations supposedly done by human hand.
Copyright:
This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
