Note: Go here to read the previous entry about this translation work.
When I started this project, I reached out to Did You Know Gaming? and the people credited in their video to see if they could provide me with their translation of the diary. They never got back to me, but the source for their original uploaded scan did. After I completed the AI translation and began working on an article, they sent me the translation work that DYKG had sourced.
I was very curious to compare this human translation to the AI-generated one I’d been working on. I wanted to see just how accurate the chatbots had been—and, more importantly, to gain greater confidence in my understanding and takeaways. On initial examination, I noticed some differences in language and tone, as well as a few concerning inconsistencies.
Phase 6 🔗
I moved each entry of the human translation into my spreadsheet for easier cross-comparison. Overall, the texts were similar; the general gist of each entry was captured by the AI translation. But there were some inconsistencies that troubled me. I wanted to write an article about my findings, and some important details were not quite the same in each translation.
What was even more confusing was that, contextually, the AI translation sometimes seemed more accurate. Here are some examples:
Original Text:
Tソリダスのキャラ設定を行なう。ソリダスを絶対悪にしないために「愛国者達」という国防を陰でぎゅうじる年寄り連中を登場させる。
GTP4-o Translation:
I worked on the character settings for Solidus. To avoid making Solidus an absolute villain, I decided to introduce a group of elderly people called “The Patriots,” who secretly control national defense.
Human Translation:
I worked on Solidus’ character background. I definitely don’t want to make Solidus evil, so I’ll include a group of old guys called the Patriots who are gnawing away at national security in the shadows.
Is Solidus not evil, or simply not the final villain? With my familiarity with the game, I know the latter to be true. But was there perhaps a point in development when the story was different? I continued through the text, listing out inconsistencies, focusing on those that were essential to my article or had major implications for the meaning of the text.
When I had gathered all the key differences, it was time to reach out to Jolie.
Phase 7 🔗
Jolie Menzel is a Design Director with an extensive resume, having worked for Ubisoft in Osaka, Japan, among other places. Now back in the U.S., Jolie has maintained their Japanese language skills. Their experience makes them an ideal reviewer for this diary, as they, like Kojima, are an experienced developer.
I asked Jolie to review both the inconsistencies and the full quotes I plan to use in my article. Out of the nine listed inconsistencies, the AI was deemed correct six times, the human translation twice, and once, both were wrong.
Conclusion 🔗
Before reaching out to Jolie, I tried to find contact information for the translator credited by DYKG, but I couldn’t locate any details. It’s unclear how they approached the work; translating the diary is a massive undertaking, so it’s possible they focused on what would be most entertaining for the YouTube channel. They may have also relied heavily on something like DeepL to handle much of the work, fine-tuning the text afterward. I don’t know.
Regardless, I think a strength of AI is that it can wear both the hat of a game director and translator, just like Jolie. All that said I don’t think I would ever have felt comfortable posting my article if I hadn’t gotten the human translation for verification.