Du hast meinen Punkt nicht verstanden. Es ging nicht ums Rechnen.
Ich zweifle auch nicht daran, dass ChatGPT nützlich ist, ich nutze es ja selbst.
Nur dient die Wortwahl "KI", "Algorithmen" und sogar das Wort "Halluzination" für die Fehler letztendlich dazu, die Sache zu hypen und dabei so zu tun, als wären wir weiter, als wir sind. Wir sind von einer "allgemeinen KI" noch weit entfernt.
Die heute KI "versteht" überhaupt nichts, sie hat nicht einmal einen Begriff von "wahr" und "falsch" - und gleichzeitig leistet die heute existierende KI in bestimmten Aufgabenstellungen Dinge, die die menschliche Leistung (mit etwa der gleichen Zahl "Neuronen" und sogar viel mehr "Synapsen") total in den Schatten stellt.
---
"If Peter is the father of Howard and Mary is Howard's aunt, what can you say about Peter and Mary?"
ChatGPT: Peter and Mary are siblings. Since Mary is Howard's aunt, this means she is the sister of Howard's father, who is Peter.
"Are you sure?"
ChatGPT: Yes, if Mary is Howard's aunt, and Peter is Howard's father, the most straightforward and typical familial relationship would be that Mary is Peter's sister. However, it's important to note that family relationships can be more complex. Mary could also be an aunt through marriage (for instance, if she is the sister of Howard's mother) or through other familial structures. But based on the information given and standard family structures, Mary being Peter's sister is the most likely scenario.