Prompts – the simple writ­ten instruc­tions giv­en to an algorith­mic­ally con­trolled lan­guage or image mod­el – seem to have almost magic­al qual­it­ies: with just a few words plus “arti­fi­cial intel­li­gence,” they can gen­er­ate a seem­ingly infin­ite amount of new texts, images, and oth­er out­puts.
 
Prompts are a key ele­ment of the “rhet­or­ic of com­pu­ta­tion” (Hayles, 1999; Hayles and Hil­lis, 2011; Hayles, 2012): they not only define the para­met­ers with­in which a machine can work, but also shape the out­put of the machine. As such, prompts are a power­ful tool for shap­ing the rhet­or­ic of com­pu­ta­tion – and thus the rhet­or­ic of digit­al media more gen­er­ally.
 
The italicized sen­tences above were gen­er­ated by OpenAI’s GPT‑3 mod­el (cf. Rad­ford et al., 2020), com­plet­ing the intro­duc­tion and adding a mostly plaus­ible second para­graph. If the enchant­ment over such accur­ate text pro­duc­tion has worn off slightly, it is mainly because we have already become accus­tomed to the rap­id advance­ment of pre-trained mod­els such as GPT‑3, LaM­DA or DALL‑E 2. Wield­ing the power of bil­lions of machine learn­ing para­met­ers, prompts are used every day to pro­duce photoreal­ist­ic images of a “dragon fruit wear­ing kar­ate belt in the snow” (Saharia et al., 2022), aca­dem­ic papers “with min­im­al human input” (Gpt, Thun­ström & Stein­grims­son, 2022) or “self-help essayist”-style blog posts (Thompson, 2022).
 
Undoubtedly, mean­ing-pro­du­cing AI mod­els not only pos­sess their own com­pu­ta­tion­al or digit­al rhet­or­icity; the rhet­or­ic of prompts and how they are used stra­tegic­ally is about to trans­form and shape pub­lic com­mu­nic­a­tion fun­da­ment­ally (cf. Schäfer and Wessler, 2020). This talk will ana­lyze the rhet­or­ic­al dimen­sions of prompts, includ­ing their func­tions and modes of action with regard to the rhet­or­ic­al sys­tem, the inter­play between humans and machines with regard to algorithmic cre­ativ­ity and the mean­ing and effects of prompts with regard to the struc­ture of com­mu­nic­a­tion.
 
Lit­er­at­ure
· Gpt Gen­er­at­ive Pre­trained Trans­former, Almira Osman­ovic Thun­ström, Steinn Stein­grims­son (2022). Can GPT‑3 Write an Aca­dem­ic Paper on Itself, with Min­im­al Human Input?. HAL. ffhal-03701250f
· Hayles, N. K. (1999). How We Became Posthu­man. Uni­ver­sity of Chica­go Press. (Ref­er­enced by GPT3.)
· Hayles and Hil­lis (2011) = Nonex­ist­ent (Ref­er­enced by GPT3.)
· Hayles, N. K. (2012). How We Think. Uni­ver­sity of Chica­go Press. (Ref­er­enced by GPT3.)
· Saharia, C. et al. (2022). Photoreal­ist­ic Text-to-Image Dif­fu­sion Mod­els with Deep Lan­guage Under­stand­ing. arX­iv https://doi.org/10.48550/arXiv.2205.11487
· Schäfer, M. S., & Wessler, H. (2020). Öffent­liche Kom­munika­tion in Zeiten künst­lich­er Intel­li­genz. Pub­lizistik 65, 307–331.
· Thompson, Clive (2022). How Many Stor­ies on Medi­um Have Been Writ­ten by AI? Medi­um. https://clivethompson.medium.com/how-many-stories-on-medium-have-been-written-by-ai-8e632da3f79b