Korean Cipher with OpenAI o1

so the example I'm I'm going to try out is a almost a code cracking of a badly corrupted Korean sentence so here I pasted in the The Prompt and I'm asking the model to translate this badly corrupted Korean sentence to uh English and as you can see this is not a invalid Korean uh sentence so let's start with the existing model GPT 40 and see how it does the model is just not able to understand this text which is a valid response because this is not a valid language so what's happening here U so Korean

is an interesting language in that when we form a character we can combine vowels and consonants sometimes put the consonant at the bottom and so on one way to corrupt this character is to add in some extra unnecessary consonants to it and that combination is so unnatural to native speakers so they can just when they see it uh just automatically undo that change and understand the text so this is character level corruption we can do that at the phrase level we can also do that at the sound level and so on people have come up

with various methods like this and I found that really interesting so I adopted a few of them to create this example so if you understand Korean this part that I'm highlighting now you can read it off as a and so on I'm not going to read off the whole thing but this is the idea Koreans can read it but the models find it so difficult to understand so now let's go on to our new model uh o1 preview and see if reasoning can help solve this problem so I typed in the same thing unlike GPT

40 this Model start thinking uh thinking through this problem before outputting the answer so it's decoding the garble text so that's actually the U the right approach because this is I gave a a translation task but the under L task is actually decoding this problem so it started off the right path and then I'm examining this text deciphering the text deciphering is actually the right Verve to use here enhancing the translation and then actually it starts unpacking some part of it so here boom and so on this is already an decrypted part of this and

then once the model figures it out that part everything else is just easy enough so it does the other sentence too and so let's close this thought so it thought for 15 seconds the final translation the model output is no translator on Earth can do this but Koreans can easily recognize it there is a method of encrypting hungo by inputting various transformations of vowels and consonants it creates a way to make it look different on the surface it can even confuse AI models I think this is a perfect translation of the sentence so this illustrate

how general purpose reasoning models like o1 preview can help seemingly unrelated questions like this which is almost like a code cracking so um I hope this illustrate how reasoning can be a powerful tool for for solving your problems