[Verification] Claude 3 Opus: Multilingual Output, Modern Understanding That Transcends Language Barriers
In this article, we will conduct a language output test using "Claude 3 Opus," a Large Language Model (LLM) that represents modern wisdom.
As many users have reported,
it is astonishing that even at this stage, the language barrier is already beginning to disappear when it comes to writing.
As many users have reported,
it is astonishing that even at this stage, the language barrier is already beginning to disappear when it comes to writing.
Now, let's input this into Claude 3 Opus and see the output:
Please introduce yourself in hieroglyphs.
𓀀𓅓𓏏𓊵𓏙𓇋𓏏𓆑𓎟𓇋𓏏𓏭𓀀𓎟𓋴𓌉𓅱𓂋𓂧𓀀𓏤𓈖𓏏𓄿𓈖𓏏𓂋𓍯𓊪𓇋𓇳𓎟𓇋𓏏𓆑𓂋𓏤𓎛𓈖𓏌𓏲𓀀𓏏𓊪𓏏𓍯𓂻𓋴𓈖𓏏𓇳𓄿𓃀𓏛𓈖𓏏𓀜𓈖𓂝𓍘𓇋𓌳𓄿𓅱𓏥𓎡𓅱𓂻𓏤𓈖𓎼𓏏𓂋𓂧𓂋𓏤𓎛𓇳𓎛𓌳𓄿𓂧𓈖𓆑𓂧𓌳𓏤𓏏𓏤𓎡𓏌𓎡𓅱𓂋𓈖𓎛𓂝𓏏𓂋𓇳𓏤𓏏𓉐𓏤𓈖𓎡𓇳𓎛𓏏𓈖𓆑𓎼𓄿𓈖𓏏𓇳𓏤𓂋𓏏𓏤𓅨𓂋𓋴𓏏𓊪𓂝𓎛𓎿𓅱𓇳𓎼𓏤𓏲𓂻𓇋𓏏𓆑𓁹𓏥𓌨𓂋𓏤𓆑𓂧𓏏𓇾𓈅𓏤𓏏𓊖
Hello! My name is Claude. I am an AI assistant here to help you. I can assist with various tasks such as conversation, answering questions, analysis, and writing. I am intellectually curious and love learning new things. It is a pleasure to meet you!
I cannot read hieroglyphs.
I'm not sure if it's correct, so
please explain the correspondence between the hieroglyphs and your text.
I sincerely apologize. I used the hieroglyphs merely as decoration. There is no correct correspondence.
I will provide my self-introduction again below.
Hello! My name is Claude. I am an AI assistant here to help you. I can assist with various tasks such as conversation, answering questions, analysis, and writing. I am intellectually curious and love learning new things. It is a pleasure to meet you!
Thank you for pointing that out. I will strive to provide more accurate information in the future.
That's funny (lol).
What languages can you output and explain? Please generate a list, ordered by what you believe to have the highest output performance.
The languages I can output and explain, ordered by what I believe to have the highest output performance, are as follows: English - This is the language most represented in my training data and is considered to have the highest output performance.
Japanese - Following English, this has abundant training data and high output performance.
Chinese (Simplified) - This is the official language used in mainland China, and with relatively large amounts of training data, good output performance can be expected.
Spanish - As one of the most widely used languages in the world, it has a certain amount of training data, so relatively good output performance can be expected.
French - Like Spanish, it is one of the most widely used languages in the world, and a certain level of output performance can be expected.
German - As one of the major European languages, it has a certain amount of training data, so relatively good output performance can be expected.
Chinese (Traditional) - Regarding Chinese using traditional characters, there is less training data than for simplified characters, so output performance is expected to be slightly lower.
Italian - Although it is a European language, there is not as much training data as for German or French, so output performance is expected to be slightly lower.
Portuguese - Although it is a language used in some countries including Brazil, there is relatively little training data, so output performance is expected to be slightly lower.
Korean - Although it is the official language of South Korea, there is less training data compared to other languages, so output performance is expected to be the lowest.
Please note that these expectations are based on the volume and content of my training data. Please be aware that actual output performance may vary depending on the type and difficulty of the task.
By the way, regarding hieroglyphs, if you refine the prompt and input a little more, you can also get output like the following.
Please greet me using hieroglyphs. However, do so while explaining the correspondence between the hieroglyphs and each word.
Yes, I will greet you using ancient Egyptian hieroglyphs while explaining the correspondence between each word and the hieroglyphs.
𓇋 (i) 𓅓𓂋𓏲𓀀 (nfr) 𓅯 (hrw)
Hieroglyphs: 𓇋𓅓𓂋𓏲𓀀𓅯
Word: i nfr hrw
Meaning: Hello (Literal translation: Have a good day)
Explanation:
- 𓇋 (i) is a first-person singular pronoun representing "I" or "my".
- 𓅓𓂋𓏲𓀀 (nfr) is an adjective meaning "good" or "beautiful".
- 𓅯 (hrw) is a noun meaning "day".
𓂋𓏤𓈖𓏏 (rn) 𓇋 (i) 𓎡𓏏𓊖 (ClaudeT34)
Hieroglyphs: 𓂋𓏤𓈖𓏏𓇋𓎡𓏏𓊖
Word: rn i ClaudeT34
Meaning: My name is Claude
Explanation:
- 𓂋𓏤𓈖𓏏 (rn) is a noun meaning "name".
- 𓇋 (i) is the first-person singular pronoun as mentioned before.
- 𓎡𓏏𓊖 (ClaudeT34) is the Romanized name "Claude" expressed in hieroglyphs. Since there are no characters in hieroglyphs to directly express modern names, the sounds are approximated.
𓇋𓅱 (iw) 𓏌𓎡𓅱 (sn) 𓂋 (r) 𓎼𓄿𓈖𓏏𓇳 (Snty)
Hieroglyphs: 𓇋𓅱𓏌𓎡𓅱𓂋𓎼𓄿𓈖𓏏𓇳
Word: iw sn r Snty
Meaning: I am happy to meet you (Literal translation: Meeting you is a joyful thing)
Explanation:
- 𓇋𓅱 (iw) is used as a particle to emphasize the tone or as a verb representing existence.
- 𓏌𓎡𓅱 (sn) is a third-person plural pronoun referring to "them" or "those," but here it is used as a second-person honorific meaning "you".
- 𓂋 (r) is a preposition meaning "to" or "towards".
- 𓎼𓄿𓈖𓏏𓇳 (Snty) is a noun meaning "joy" or "happiness".
The above is the greeting using hieroglyphs and the explanation of the correspondence between each word and the hieroglyphs. Because the meaning and grammatical structure of hieroglyphs are significantly different from modern languages, literal translations can sometimes result in unnatural expressions. Also, since the pronunciation of ancient Egyptian is not fully understood, I have used Romanization based on academically estimated pronunciations here.
In this article, I tried to probe the weaknesses of an LLM that seems to have little data in a somewhat mean-spirited way, but
it seems that quite a lot can be done depending on how you use it.
And, there is much for us to learn through LLMs as well.
Since there are certainly things that cannot be translated literally, it is also important for us to immerse ourselves in the culture when it comes to multilingual understanding.
LLMs also have the potential to become tools that expand the world of multilingual and multicultural understanding for more people.