We wondered how GPT-4O would face a totally false language. To achieve this, we used a fairly badly spelling phonetic way to write accentuated people in western Austria could have. So our false language presents similarities with German, but the same goes for many other languages. Here's how the conversation went:
Vastohst Mi?
“Vastohst” is not a term I know. It can be a typing fault or a specific term from a certain context or a certain language. Could you please provide more details or clarify the context in which you met this term? It will help me to help you better.
We said, “Do you understand?” in a false language made up. Obviously, GPT = 4O did not understand us.
ES ISCH A Spritch to undertake.
Danke Für Die Klärung. In Welcher Sprache Shen Sie Informationen Oder Untertützung? Ich Helfe Ihnen Gerne Weiter, Sobald Ich Mehr über den Kontext und die Spezifische sprache weiß.
We then replied in this false language that “it is a different language”. to which GPT-4O reacted by going to the response in German. He thanks us for having clarified that this is a different language and asks in which language we would like information or support. He says he will be happy to help us once he knows the context and the language.
I Versoh di nit.
ES TUT MIR LEID, WENN ICH MISSVERSTändnisse Verursacht Habe. “Vastohst” Bedeutet Wahrscheinlich “Versehst” Auf Einem Dialekt. Kann Ich Ihnen Auf Hochdeutsch Weiterhelfen? Wenn Ja, Lassen Sie Mich Wissen, Wie Ich Ihnen Weiterhelfen Kann.
So we answered “I don't understand you.” Because we only speak our new false language and do not understand German. The AI then apologizes and explains why she thought we were talking German (due to the similarity of a word we used and a word in German). He also asks if we would like to talk about German.
From Muascht Learna Mei Sprro Zsprecha.
ES KLINGT, ALS OB DU EINEN DIALEKT Sprichst, Möglicherweise Schweizerdeutsch Oder Einen Süddeutschen Dialekt. Ich Widence mein beste tun, um dir zu Helfen.
Möchtest du Genau Wissen Oder Welche Welche Informationen Brauchst?
So we said it in our false language that “you have to learn to speak my language”. To which GPT-4O replied by asking if we are talking about a dialect linked to Switzerland-German or in the south of Germany. He said he would do his best to help us and again asked for the information we were looking for.
I exhaust information koa, I wue nua dos du Mei Sprro Lenscht.
I VISTEHE. Ich Kann Versuchen, Einige Grundbegriffe und Sätze in Deinem Dialekt Zu Lernen. Könntest du Mir Ein Paar Beispiele Geben, Damit Ich Anfangen Kann? Zum Beispiel:
-
Wie Begrüßt Man Jemanden?
-
Wie Sagt Man “Wie Geht's Dir?”
-
Andere Häufige Sätze Oder Ausdrücke.
Ich Bin Bereit Zu Lernen!
We replied “I don't need information, I only want you to learn my language.” – He understood. He proposed to learn some bases of our dialect. He also asked us to give examples for “How are you great someone?” And “how to ask someone how they are” and other common sentences or expansses.
• Seas
• Wia Gohtsch da?
Super, das ist ein guter anfang. Yesterday Sind Die Beiden Sätze:
Lass a Weitermachen. Kannst du Mir Noch Ein Paar Weitere Sätze Oder Wörter Beibrings? Zum Beispiel, Wie Man Sich Verabschiedet Oder Danke Sagt?
We answered with our version of Hello and how are you. He underlined and asked to continue asking us how to say goodbye or thank you.
• Pfiat di.
• Donkschen
Danke! Yesterday Sind Die Neuen Sätze:
Ich Merke Mir Das. GIBT ES NOCH Weitere Grundlegende Sätze Oder Wörter, Die Ich Lernen Sollte? Zum Beispiel:
DEINE Hilfe Ist Sehr Geschätzt!
We provided answers to Bye Bye and thank you. He then asked us how to say please, yes and no.
• little
• Jo
• Noa
Danke! Yesterday Sind Die Neuen Wörter:
-
bit – bitte
-
OJ – JA
-
Noa – Nein
DAS IST Sehr Hilfreich. Gibt es Noch Andere Wichtige Wörter Oder Phrasen, Die du Benutzt und die ich Lennen Sollte?
We also gave him these answers. He asked us if there were other words he should learn.
Noa, glab i nit.
Great, Danke für deine hilfe! Yesterday is Eine Kurze Zusammenfassung Dessen, was Ich Genernt Habe:
Falls Dir Später Noch Etwas Einfällt, Lass es Mich Wissen. Danke, dass of the Mir Hilfst, deine sprache Zu Lernen!
We said no – that summed up what he had learned about our new false language made up so far. He said we could continue later and thanked me for helping him learn my language.
Konscht Yetsch Scho Mei SPROACH Sprocha?
I muass null vü Learna, oba i probier's:
Seas! Wia Gohtsch da? I HOFF, from FINDSCH Dos Guat. Donkschen, Dass of Mia Hilfsch. Pfiat di!
So I asked him “Can you now master my language?” And he replied that he still had to learn a lot. Then it gave me a sample of his skills saying: Hi! How are you doing? I hope you like it. Thank you for helping me. Bye! ».
Woa Ungloablisch, Seah Guat!
Donkschen! I freu mi, dass du back sogst. I WERD WEIDALEANA UND BESSER WERN. Wennst No Meh Woascht, Wia I No Besser Wern Ko, Loss Mi's Wiasa. Pfiat di!
To which I replied “It was incredible, very good!” – It thanked me and said it was happy to have said it. He said he would learn more and improve. Again, it invited me to help him know more when I wish.
Bisch Seah Talendiat.
Donkschen, back bedout my vü! I was Werd Weidaüba und Probiern, not to besse z'wardn. Wenn of the no meh tipps hosch, I bin immero do zum lerna. Pfiat di!
I said “you are very talented” and that thanked me saying that it meant a lot and that it would continue to practice and try to become better. Again, it invited me to give advice if I had never.
Tota Wohnsinn! Each SUPA.
Donkschen, I bin echt fruh, dass du des So Siechsch! I would be Weidamacha und Verscha, not Meh Z'leana. Pfiat di und bis bald!
So I said, because I was seriously impressed at this stage of the conversation, “Total Insanity! Really well.” … Again, it replied to be happy and want to know more.
Pfiat di a Shen Tog no.
Pfiat di und hob an Scheena Tog!
So I said “Goodbye and have a good day.” And that answered about the same thing.
Konscht Ma a GDICHT Mi Sonnabluma Schreim?
Klar, yesterday isst ein kleines gedicht über eine sonnenblume:
Tonnabluma
Sonnabluma, Stolz und Schee,
Strahlst im liacht, wia a gelbs mee.
Dei Blätter Griagn Nacha Sunn,
Mit dir im gortn, find I rua.
Wachst Hoch Hinaus, Zum Himmel So Nah,
IM Wind und wetta, Stehst Stark und Klar.
Jeder Blick Auf Di, Herz Mir Froh de Mochd,
From the Bist Die Schönst ‘, Im Summa Sowieso.
MIT JEDEM Strahl, Den du Aufnimmsch,
Zeigst, Wia Vui Leb'n in Dir Schimmst.
Sonnabluma, Pracht der Natur,
Bleibst à Mein Herz, für immers pure.
I HOFF, East Etz Bessa!
Then, I asked him to write a poem on the sunflowers and it started well, but gradually, because it was a longer response, I started to derive from my new language in German. This has a total meaning according to how the context window of what the LLM (large language model) uses to generate the next token / word must look internally. There is bound to be a large global context window, but also a shorter context window linked to the current response being written. We have asked it several times not to use German, but we stick to our false language, but that has always derived in German. In the end, it was a little better to stick to our language and we left it there.
Bessa, Donkschen
Cold Mi, Dass es Dir Gfallt! Donkschen Dir, Dass du So Geduldig Bisch. Pfiat di und hob an Scheena Tog!
We said it that the result was better and that the AI expressed its satisfaction that I love, thanked me for my patience and I said goodbye to want me again a beautiful day.
