Artificial intelligence

In arti­fi­cial intel­li­gence, a fun­da­men­tal dis­tinc­tion can be made between “strong” and “weak” AI. Weak AI are sys­tems that are mostly rule-​based and lim­ited to a clearly defined task area, such as the recog­ni­tion and pro­cess­ing of lan­guage. The major­ity of “weak” arti­fi­cially intel­li­gent sys­tems, build on the prin­ci­ple of machine learn­ing, which has become pop­u­lar espe­cially in appli­ca­tion sit­u­a­tions where large amounts of data are to be sta­tis­ti­cally eval­u­ated and searched for pat­terns – e.g. spam fil­ters, autonomous dri­ving sys­tems or med­ical diag­noses. So-​called deep learn­ing, which is expected in the future, is a sub­field or fur­ther devel­op­ment of machine learn­ing and could become the basis of “strong” AI.

“Strong” AI has intel­lec­tual skills that rely on Deep Learn­ing algo­rithms on arti­fi­cial neural net­works whose struc­ture and func­tion are mod­eled on and at least equiv­a­lent to the human brain. Their insights should not be lim­ited to one domain, but should be applic­a­ble to any num­ber of other domains.

With a large infor­ma­tion base and the struc­ture of neural net­works, Deep Learn­ing sys­tems can inde­pen­dently link learned con­tent with new con­tent. In addi­tion, they are able to make analo­gies with­out being con­fronted with cer­tain key terms in the con­text of a ques­tion. The more mean­ing­ful data such sys­tems process, the more reli­able they become. Thus, the “learn­ing effect” ide­ally leads to con­tin­u­ous self-​optimization. Deep Learn­ing algo­rithms are based on arti­fi­cial neural net­works whose struc­ture and mode of oper­a­tion are mod­eled on the neu­rons of the human brain, which means that these Deep Learn­ing sys­tems can inde­pen­dently link learned con­tent with new con­tent in inter­ac­tion with a large data­base.

More­over, they are able to infer analo­gies with­out know­ing spe­cific key terms in the con­text of a prob­lem. The more mean­ing­ful data such sys­tems process, the more reli­able they become. The “learn­ing effect” thus leads in the best case to con­stant learn­ing opti­miza­tion.

• Flex­i­ble, sim­ple design and con­fig­u­ra­tion of the auto­mated cus­tomer dia­log and work­flow for all busi­ness trans­ac­tions with the help of struc­tured and reusable text and lan­guage mod­ules
• Map­ping of fre­quently occur­ring par­tial dialogs, e.g. iden­ti­fi­ca­tion or legit­i­ma­tion processes
• Sav­ing of money as well as admin­is­tra­tive costs thanks to in-​house response to increased cus­tomer inquiries
• Process opti­miza­tion in the com­pany thanks to the effi­cient han­dling of cus­tomer con­tacts
• Auto­mated send­ing of doc­u­ments, e-​mails or SMS with, if nec­es­sary, inde­pen­dent con­tact­ing of cus­tomers
• Reduc­tion in num­ber of for­warded incom­ing calls through tar­geted for­ward­ing to rel­e­vant employ­ees
• Con­tin­u­a­tion of work with exist­ing data base thanks to exten­sion of the exist­ing infra struc­ture by only a few com­po­nents
• Reduc­tion of load peaks and bet­ter plan­ning of work vol­umes thanks to the sav­ings poten­tial for recur­ring, stan­dard­ized busi­ness trans­ac­tions