What do we think - a GPT-3 experiment
What do we think is a GPT-3 artificial intelligence language model generating thoughts based on age and gender.
The language model is primed with five random blog posts annotated with age and gender from this dataset. The model has crunched through large quantities of the internet. When prompted with age and gender it generates a snippet of text based on the context it figures out from the priming examples and the given prompt.
In a way the results represent some kind of average collective discourse of age and gender ‘in the internet’. Or maybe the results are somewhat random, and I interpret them based on my own assumptions of age and gender.
What do we think was made after a two-day Artificial Intelligence Exchange workshop at TU Dresden.
Here are few examples of the output prompted with random age and male and female as gender.