It was on October 28, 2024, and I was preparing for an Igbo Language lecture in honor of the late writer Chinua Achebe (when the lecture goes on air soon, it will be available here).
At one point during the preparation, I needed to write the word ‘brand’ or ‘business’ in Igbo. So, I approached Meta AI (Llama 3.2) to assist me in doing that.
I asked: “What is the Igbo word for ‘business’?” I probably thought it was too ambitious of me to start with”brand”.
Surprisingly enough, when I submitted the query, I saw the AI chatbot answering it. Then, all of a sudden, it withdrew the answer and gave me the following excuse:
“I don’t understand Igbo yet, but I’m working on it. I will send you a message when we can talk in Igbo”
But, I wouldn’t take that…
I was very furious at that lame excuse and regretted why I didn’t take a screenshot of the answer I saw scrolling in some seconds before.
So, I insisted that the chatbot must restore the answer. Look at the dialogue…
Meta AI (Llmam 3.2) translates the word 'business' to Igbo |
You can see me challenging the humanoid to believe in himself and write in Igbo. He did and even thanked me for the encouragement. He got the translation almost perfectly, including even the punctuations!
I’m hoping for the day he can write ✍️ Igbo with the full and correct accentuation marks. For example, it would be able to translate ‘business’ as ‘àzụ́máhịá’.
Also, though satisfactory, the bot’s analysis that followed the translation however didn’t quite hit the mark. ‘Trade’ is not exactly correct as ‘azụma’. It’s a bit tricky, but I hope there will be Igbo academics who can contribute in solving the impasse.
Next, I asked Llama 3.2 to translate ‘brand’ to Igbo
I was expecting him to fail woefully this time. Fortunately, I was wrong!
This is the answer I got:
Meta AI translates the word 'brand' into Igbo |
The 3 translations of the word ‘brand’ are so rich, expressing different nuances of it so well. None of the translations could fit into the usage I was seeking, as in “we are supporting local brands”. Yet, each of them was insightful enough for a bot made in America that just complained about not being able to understand Igbo!
Llama 3.2 is even attempting the accentuation already!
I must say that I’m impressed at the bot’s attempt at using the accentuation marks already in its analysis of the word ‘brand’ as you can see👆. The accentuation mark on ‘àha’ is wrong. It should have been ‘áhà’ or he should have left it unaccented in the context.
As for ‘àzụmahia’, the bot got the first syllable correctly accented but failed to accent the rest of the syllables. It's getting better.
But, it didn’t end there…
Instead, I was to have a more running battle with the bot. As you can see, after translating ‘brand’, he returned to his former behavior when I asked him to translate ‘idea’: It gave an answer, withdrew it, and gave me the excuse that it didn’t “understand Igbo”. Na wah oh🫢
Conclusion
I will bring up the rest of the conversation and the deep lessons I learned from it in a later post (probably tomorrow or next). You will see how AI is teaching us to be who we have always been made to be. It’s really edifying to know.
Meanwhile, this little attempt can encourage the rest of us to use the AI Revolution to get our various local languages more standardized while helping to expand the capacity of the tech to serve us better. It will definitely help a lot if more of our local languages can hop on the AI train this early. So, kindly share this.
READ ALSO: Could This Be The Most Memorable Gift For Mom Or Dad This Christmas?
Comments
Post a Comment
Let's share. Contribute your thoughts to this discussion below...