The DGPuK con­fer­ence for the "Media Lan­guage — Media Dis­course" interest group kicked off on 28.02.2024 in Karls­ruhe with an inter­act­ive read­ing and pan­el dis­cus­sion with author Jörg Piringer and oth­er guests entitled "Schöne Neue Tex­twelt" (A Brave New World of Texts). A sim­il­ar event had already taken place at the Epple­haus in Tübin­gen in the fall. This time, the ven­ue was loc­ated in the math­em­at­ics build­ing of the KIT and was already well vis­ited at the start of the event at 19:30.

The even­ing was open to every­one inter­ested in the top­ic and aimed to act­ively involve the audi­ence and reflect togeth­er about the dimen­sions of gen­er­at­ive AI. The medi­um for this was the Men­ti­meter tool, which the audi­ence used to repeatedly answer ques­tions (and also ask some them­selves), share asso­ci­ations and express opinions.

The event was mod­er­ated by the edit­or-in-chief and co-founder of Sci­ence Notes, Dr. Thomas Susanka, who asked the audi­ence to share their asso­ci­ations with gen­er­at­ive AI right at the begin­ning. With this ini­tial food for thought in mind, the read­ing part of the even­ing began. Jörg Piringer read from his book of poetry "Gün­stige Intel­li­genz" (Cheap Intel­li­gence), which he co-cre­ated with an early ver­sion of Chat­GTP. Piringer pur­chased early access to the AI for just €5.60 in 2021 and began exper­i­ment­ing with the pro­gram. Among oth­er things, he asked Chat­GTP to out­put word lists and non­sense words, which he then used to gen­er­ate poems.

He read out loud some of these poems, altern­at­ing with reflect­ive pas­sages about the co-cre­at­ive work with Chat­GTP. With­in these pas­sages, Piringer asked him­self vari­ous ques­tions, for example, what is the inten­tion of the AI? Who is actu­ally the author of the texts writ­ten? And what will the future of lit­er­ary intel­li­gences look like?

The read­ing was fol­lowed by anoth­er round of ques­tions dir­ec­ted towards the audi­ence, as well as the joint writ­ing of a prompt for Chat­GTP and an eval­u­ation of the res­ults the AI pro­duced. The audi­ence prompt was: "Write an obit­u­ary for an air con­di­tion­er in a region­al dia­lect without using the word air con­di­tion­er". The res­ult was a source of amuse­ment for those present, but it was clear that both the dia­lect and the genre of the obit­u­ary could not be imple­men­ted well by ChatGPT.

The inter­act­ive part of the even­ing was fol­lowed by a pan­el dis­cus­sion, in which Jörg Piringer was joined by Prof. Annette Leßmöll­mann, Pro­fess­or of Sci­ence Com­mu­nic­a­tion with a focus on lin­guist­ics at the KIT and research­er at the RHET AI Cen­ter, and Prof. Jan Niehues, Head of the Chair of Arti­fi­cial Intel­li­gence for Lan­guage Tech­no­lo­gies at the KIT.

Annette Leßmöll­mann opened the dis­cus­sion with the obser­va­tion that gen­er­at­ive AIs have also found their way into sci­ence com­mu­nic­a­tion — espe­cially in edit­or­i­al work — and that there is a lot of reflec­tion hap­pen­ing on the future of co-cre­ativ­ity in sci­ence journ­al­ism. "We are liv­ing in a con­stant Tur­ing test," she emphas­ized, in which it must be checked which texts could be AI-gen­er­ated and which ori­gin­ate from a human mind. This has an impact on recip­i­ents and will lead to the devel­op­ment of new cul­tur­al techniques.

When it comes to the cre­ativ­ity of arti­fi­cial intel­li­gence, Jan Niehues argues that non-cre­at­ive texts in par­tic­u­lar, which are primar­ily for­mu­list­ic in nature, can be writ­ten well by lit­er­ary AIs. In the future, the use of gen­er­at­ive AIs could reach a level sim­il­ar to that of googling inform­a­tion today, where an AI searches through exist­ing texts and cre­ates new texts from them.

Jörg Piringer sees people's focus on indi­vidu­al pref­er­ences, feel­ings and atti­tudes as a key unique selling point of humans, which is closely linked to cre­ativ­ity. While arti­fi­cial intel­li­gences work towards a broad selec­tion of imit­a­tions rather than indi­vidu­al­ity, a spe­cial fea­ture of the cre­at­ive pro­cess for Piringer is also the con­scious and unique selec­tion of cer­tain text mod­ules and pas­sages, which human authors con­stantly do dur­ing the writ­ing pro­cess (also dur­ing co-cre­at­ive writing).

After these ini­tial state­ments, the dis­cus­sion was opened up to ques­tions from the audi­ence. Those present were par­tic­u­larly inter­ested in the ques­tion of wheth­er AIs will have a per­son­al­ity in the future, which Jan Niehues answered with a highly prob­able no, as this is cur­rently only pos­sible through very tar­geted prompt­ing. Annette Leßmöll­mann added, that it is how­ever excit­ing to see how quickly we anthro­po­morph­ize com­puters and arti­fi­cial intel­li­gences, i.e. attrib­ute a human-like per­son­al­ity to them.

The leg­al ques­tion of the copy­right of co-gen­er­ated texts was also raised by those present. Accord­ing to Piringer, in his opin­ion the rights do not lie with the AI but with the human oper­at­or. He added that Chat­GTP and oth­er gen­er­at­ive AIs have been trained with the data and intel­lec­tu­al prop­erty of count­less artists without them ever being com­pensated for it.

Finally, the audi­ence ven­tured a look into the future and wanted to know how the next gen­er­a­tion, which is already grow­ing up with arti­fi­cial intel­li­gence, will per­ceive it? Jan Niehues is cer­tain that they will shape every­day life in the same way that the spread of cell phones and smart­phones did in pre­vi­ous gen­er­a­tions. Arti­fi­cial Intel­li­gence will also become highly rel­ev­ant for work, for example. In addi­tion, the dicho­tomy between nat­ur­al and arti­fi­cial intel­li­gence will likely dis­ap­pear, said Annette Leßmöll­mann. It may even be neces­sary at some point to authen­tic­ate human-pro­duced or human-veri­fied texts as such.

Fol­low­ing the read­ing, the DGPuK spe­cial­ist group con­fer­ence "Under­stand­ing and Com­mu­nic­a­tion in the Digit­al Space" of the interest group "Media Lan­guage — Media Dis­courses" began. You can find a review here.