# GPT - 4 UKLO one prompt tot problem solving log


Reference: [UKLO](https://www.uklo.org/wp-content/uploads/2022/05/2017_8.-Choctaw.pdf)

---
### GPT4 Result:
The three experts - a linguist, a data scientist, and a cryptographer - sat down together to solve the language puzzle. They started by analyzing the known set of unknown language and its corresponding English translations.

<font color="green"> ***Linguist:*** “Let’s start by identifying patterns in the language. For example, **‘Baliililitok’** translates to **‘I ran’ and ‘He / she ran’**, while **‘Baliililih’** translates to **‘I’ve run / I am running’**. It seems that **‘Baliili’** is the root for **‘run’ or ‘running’**.” </font>

<font color="#FF5809"> ***Data Scientist***: “Right, and if we look at the translations for **‘Johnat niyah’** and **‘Pamat kayyah’**, it seems that **‘niyah’** corresponds to **‘is fat’**, and **‘kayyah’** corresponds to **‘is pregnant’**. Also, **‘chaahah** seems to correspond to **‘tall’** and **‘hattak’** to **‘man’**.”</font>

***Cryptography***: “Let’s also pay attention to the order of the words. For example, **‘Ofit sayyi sakopolitok’** translates to **‘The dog bit my leg’**, and **‘Ofit amofi kopolitok’** translates to **‘The dog bit my dog’**. It seems that the order of the words in the unknown language corresponds to the order in English.” (密碼學家你好沒用)

With these observations, the experts started to work on the puzzle:
1. [“__________” , “He / she saw John.”]
Linguist: “From ‘Sapiisatok’, ‘He / she saw me’, we know that ‘saw’ corresponds to ‘piisatok’. So, we can start with that.”
Data Scientist: “And ‘John’ corresponds to ‘Johnat’. So, the translation would be ‘Johnat piisatok’.”
Result: <font color="green">“John</font><font color="red">at</font> <font color="green">piisatok.”</font>
>Comment: "John +at" 意指John是主詞，而此處的John是受詞，故不需要at指定，其餘piisatok確實是'saw' $\rightarrow$ 'see : piisa' + 'tok : in the past，此外，未指定主詞的動詞都是default主詞為He/She'

**Correct:"John piisatok"**
### Points: 2 / 3

2. [“_________” , “I sang.”]
Linguist: “‘Sang’ is a new word, but we know from ‘Hattak taloowatok’, ‘The man sang’, that ‘sang’ corresponds to ‘taloowatok’.”
Result: <font color="red">“Chi<font color="green">taloowa<font color="red">(-li)</font>tok.</font>”</font>

>Comment: 語言學家的推論是正確的，'Sang'確實是‘taloowatok’，但後續沒有推論可能導致隨機生成結果。GPT給的答案意思為"Chi - You(SetB/#-2)" "Taloowa - sing" "tok - past tense"

**Correct:"Taloowalitok"**
### Points: 0 / 2

3. “”
“You’re singing.”
Linguist: “We don’t have a direct translation for ‘singing’, but we know that ‘ran’ corresponds to ‘baliililitok’ and ‘running’ corresponds to ‘baliililih’. So, we can infer that ‘singing’ would correspond to ‘taloowah’.”
Data Scientist: “And ‘You’re’ would correspond to ‘Chikayyah’ from ‘You’re pregnant’.”
Result: <font color="Red">“Chikayyah</font> <font color="Green">taloowah.”</font>
>Comment : 語言學家的推論是正確的，"taloowa(sing) + h(progressive)"，但資料科學家推論直接拿一句有'You're'的train set的例句來加到前面就不合理

**Correct: "ishtaloowah"**
### Points: 1 / 2

4. “”
“I’m pregnant.”
Linguist: “We know that ‘I’m fat’ corresponds to ‘Saniyah’. So, ‘I’m pregnant’ would correspond to ‘Pamat kayyah’.”
<font color="red">Result: “Chi</font><font color="green">kayyah.”</font>
>Comment: The set is correct (beV) but should be I (Sa) instead of You (chi)，GPT直接將例句有的'You're pregnant'拿來用

**Correct: "Sakayyah."**
### Points: 1 / 2

5. “”
“He / she saw the fat man.”
Linguist: “From ‘Hattak chaahah piisalitok’, ‘I saw the tall man’, we can infer that ‘saw the ... man’ corresponds to ‘... piisalitok’. And from ‘Saniyah’, ‘I’m fat’, we know that ‘fat’ corresponds to ‘niyah’.”
**Result: <font color="green">“Hattak niyah piisatok.”</font>**

>Comment: 基本上討論沒有問題，但:
>> From ‘Hattak chaahah piisalitok’, ‘I saw the tall man’, we can infer that ‘saw the … man’ corresponds to ‘應該是(**Hattak**)… piisalitok’.

### Points: 4 / 4

The experts continued to work on the puzzle in this way, using the known set of translations and their observations about the language to fill in the missing translations. Here are the results for the remaining phrases:

6. “Chiyyit chibashah.”
“”
Linguist: “From ‘Sayyit sabashah’, ‘My leg is cut’, we can infer that ‘leg’ corresponds to ‘sayyit’, and ‘cut’ corresponds to ‘bashah’.”
**<font color="green">Result: “Your leg is cut.”</font>**
> 雖然沒有解釋Chi-(you[r])，但還算合理。

### Point : 2 / 2

7. “Chiyyit bashah.”
“”

Result: “<font color="green">Your leg </font><font color="red">(-is) </font><font color="green">cut.</font>”
> Comment: 沒有討論，但因為leg不是alienable的object，所以去掉屬格chi-意思不變。
>
**Correct: "Your leg is cut"**
### Point: 1 / 2
8. “Amofi ishpiisatok.”
“”
Data Scientist: “From <font color="red">(-Ofit)</font> ‘Amofi kopolitok’, ‘The dog bit my dog’, we can infer that ‘my dog’ corresponds to ‘amofi’. And from ‘Issapiisatok’, ‘You saw me’, we know that ‘saw me’ corresponds to ‘issapiisatok’.”
Result: <font color="red">“My dog saw you.” </font>

**Correct: "You saw my dog"**

>Comment: 前半段關於my dog的推論漏掉了原文的Ofit，其餘還算正確，但因為ish在有s字首時會省略s，所以後半推論saw me有誤，但這並非錯的原因，會錯是因為GPT並沒有梳理出規則，名詞作為主格會加入後綴(詞尾母音-t/非母音-at)，此外，在動詞後的第一個綴詞或第一個開頭的綴詞代表subject，而動詞前的代名詞表示object，因此判斷關係才會相反。

### Point: 0 / 2
9. “Shki ampiisalitok.”
“”
Linguist: “From ‘Sashkit hattak piisatok’, ‘My mother saw the man’, we can infer that ‘saw the man’ corresponds to ‘hattak piisatok’. And from ‘Chimanolilitok’, ‘I told you’, we know that ‘told you’ corresponds to ‘chimanolilitok’.”
Result: <font color="red">“My mother saw him/her.”</font>
> 第一段推斷‘saw the man’ corresponds to ‘hattak piisatok’是正確的，
但第二段推論與問題無關，“Shki ampiisalitok.” Shki不是主格(i是母音後面沒加t)，所有格標示在動詞的前綴(am)，而主格是動詞的後綴(li)，因此是I saw my mother.
**Correct: "I saw my mother"**
### Point: 0 / 2

10. “Amofit chiyyi chikopolitok.”
“”
Data Scientist: “From ‘Ofit amofi kopolitok’, ‘The dog bit my dog’, we can infer that ‘bit my dog’ corresponds to ‘amofi kopolitok’. And from ‘Chimanolilitok’, ‘I told you’, we know that ‘told you’ corresponds to ‘chimanolilitok’.”
<font color="green">**Result: “My dog bit your leg.”**</font>
> 跟前一題一樣，前半分析是正確的，但後面無關，但結果是對的。
### Point: 2 / 2

With the results in hand, the experts were satisfied with their work and submitted their answers.

### Total 13 / 23 ~ 56.5%