Ruth McCreery
General Comments
“Complete, accurate, and as natural as possible”: those are the goals set for those tackling this translation. “Complete” basically asks that you be thorough and take care not to skip things (especially the head-scratching bits). “Accurate” can have many meanings (we have all encountered someone who insists など must be always translated, and must be translated as “etc.”), but “as natural as possible” clarifies that the translator is expected to produce English that communicates the content of the Japanese text in an appropriate style. That requires implementing the plain but not simple Fred Uleman process: “Read a little, think a little, write a little.”
The finalists largely did well on all three steps in the process, but some could have given more thought to the subject matter, particularly the terminology for the burgeoning field of natural language processing AI. Plus, for a result to be “as natural as possible,” thought needs to be given to what a similar article in an English magazine would look like and particularly what an effective title might be. A title is, after all, intended to catch a reader’s attention. (I recommend doing the title last, after you’ve digested the article and perhaps understood any little twists that the title might contain.)
A natural translation also needs to cope with those twists in the body of the text. In this case, after focusing on AI’s capabilities, threats, and weaknesses, the author ends by strongly advocating drastic changes in the (Japanese) educational system. That, I would guess, is something the author cares about deeply and should be treated as the potent conclusion to his article.
E05’s translation of the title is rather clumsy. “The economy of ‘idea to product’ made possible by AI” would be far more effective with a little rearranging, as in E46’s “The Instant Idea-to-Product Economy and the AI Behind It.” E20’s “Instant Materialization” is rather puzzling: is this about AI or Star Wars? E42 has not made the idea-to-product connection. E46 produced an accurate and compelling title: “The Instant Idea-to-Product Economy and the AI Behind It.”
While how to identify the author of this article is perhaps less critical, the usual English order would be the name followed by position or qualification. “Economist Tomohiro Inoue” reverses it. (What to do about name order is always an issue, often to be decided by a publication’s style sheet or editor’s prejudices.)
In the first paragraph, E05’s “articles” as products somewhat misses the gravitas of 論文. Yes, a 論文 in a scholarly journal is an article, but the “What’s up this weekend” article in the newspaper is not a 論文. E20’s “theses,” however, goes too far in the other direction. “Essay,” as chosen by E42, E46, and E51, works better. In the same list of products, E05’s and E20’s “programs” is mere un-katakana-ization without bearing in mind that, in a list that includes essays and novels, the English “programs” is rather vague. “Computer programs,” as chosen by the other three finalists, would be more appropriate. E05 has also managed to translate the single-sentence paragraph as one long English sentence that is rather difficult to read. A natural translation would chop it up a bit, as the other four finalists did. A further problem is how to describe those images: 写真 as “photographic” or “photos” is problematic, since the English term refers to “drawing with light,” using one of several technologies to capture an actual image. Promotional materials for image-generative AI systems do claim they can produce photos, but AI creations are better described as “photorealistic,” as E46 and E51 concluded. E42 further confuses things with a grammatical error that makes it seem as though AI is coming up with the prompt as well as generating the image.
In the next paragraph, the output would be e-books, not “electronic documents.” Moreover, the stress in the Japanese is on being able to sell them; that is weakened by E05’s “and also sell them.” The other four finalists maintained the correct emphasis.
In the following paragraph, the main problem is 小物や家具, a rather odd pairing if rendered literally. “Objects both small and large such as furniture” (E42) expresses the difference in scale, as does E51’s “tools, accessories, and household items.” Both also make it clear that the products in question are ordinary objects, not amazing new inventions.
The next paragraph repeats the problem in the title. E05’s “idea to product” omits the すぐ of アイデアをすぐ形に and the 即 of アイデア即プロダクト. People have been turning ideas into products for centuries; it is the speed of E46’s “instant idea-to-product” economy that needs to be expressed. But not by “instant materialization,” please. In stating how, thanks to AI, anyone can become a creator, choosing “poor at art” or “lack artistic ability” would seem to be a rather free interpretation of 絵 in 絵が苦手、given that “art” has a much broader meaning than drawing or painting. But here pushing beyond the literal improves the author’s argument, making this sentence consistent with 美術者 in the next sentence. The E05 version of the next sentence is concise but does not make it clear that “those of average skill” must be those hoping to be professional artists, not all of us. The other finalists handle this more successfully.
Then we come to a turning point: AI as a threat rather than a blessing. With only one in five applicants scoring a job as an artist or designer now, and with AI becoming ever cleverer, jobs will become harder to find. Please note, however, that it is not “1 in 5 people” or “1 artist in 5” but one in five applicants for such jobs. The 有効求人倍率 early in the paragraph is not a familiar term or one that moves the argument forward, since the author goes on to explain that the 0.18 ratio means one in five applications secures a job in that field. I was surprised that none of the finalists reversed the order in that paragraph and began with the one in five information, providing the 0.18 selection ratio info at the end. More daringly, I would consider simply omitting the selection ratio part, since it is redundant and, by introducing an unfamiliar term, slows down the argument.
The discussion of the threat to non-artistic workers poses two problems: how to parse the list of occupations and what to do with 士業. The latter refers to occupations whose titles end with 士 and for which professional certification is required. The list can thus be broken down into office workers, professionals such as accountants and tax preparers, and academics and researchers like the author. Finalists E42, E46, and E51 all sorted out the list correctly. “Professionals” for 士業 could be omitted, since that category (of which there are several groupings) does not correspond to the broader and more vague meaning of “professional” in English. In translating many texts, the translator would need to add an explanation for 士業; here, since it does not recur and spending ink on it would not contribute to the reader’s understanding of the argument, it can be omitted.
In the following sentence, stating that AI will not trump humans (thank you E46!) in every way, the になる in 上回るようになる tells us that the author is looking towards the future, not stating current conditions. E20 missed that.
As the author moves on to discussing AI’s limitations, we run into another problematic list: 「意志」「体験」「価値判断」. The first two correspond neatly to English terms, but if 価値判断 is translated as “value judgement,” it refers to the result of applying judgement, not a quality or ability. “Ability to make judgements” is rather wordy; just plain “judgement” might be a better choice. In the next sentence, the technical term 偏差値 refers, in statistics, to the number of standard deviations above or below the mean. In the world of Japanese education, however, it simply means someone’s test scores, and there is no need for the translator to get into statistical terminology. I also would question whether off-the-chart test scores signify membership in the intellectual elite. Indeed, in the conclusion to this essay, the author clearly opposes that interpretation. “Super-elite test takers”? And then what about the passivity of the 指示待ち人間? E51’s interpretation of both terms in an employment context makes sense given the earlier discussion of AI as threats to human jobs, but may be a slight overreach. E20 expresses the clever contrast the author has set up succinctly (if “reactive” could be replaced by, for example, “passive.”) Given that passivity, AI lacks the will to start up a new business or launch a new project: note that 企業するgenerally means “starting up a new business.” 事業 can be anything from a business to an activity, but since the author has mentioned both 起業 and 事業, their translations should differ.
All the finalists did a good job on our “agoraphobic shut-ins.” But what do we do with the quotation marks around 「体験」? The author is telling us that AI cannot, by definition, have first-hand or personal experience of anything. All the finalists successfully avoided repeating the quotation marks in their translation. “Own experience(s)” or “personal experience” or even “bodily experience” would clarify that the issue is something the entity has actually done.
All the finalists dealt rather skillfully with the paragraph about AI’s inability to come up with ideas and create a work. E46’s “slew of novel ideas” fits beautifully with the action of the messy but powerful human brain. Both E42 and E51, however, used “brainstorm,” which usually implies several people (or here, perhaps, AIs) tossing ideas around.
The next sentence says creating original works would be 難しい. Since the author has established that that AI simply cannot do it, “difficult” seems rather weak. But consider how often you have received 「難しい。。。」in response to a request, and you may find E46’s “finds itself at a loss” or E20’s blunt “cannot” both consistent with the author’s argument and true to the Japanese.
All the finalists handled the Tokyo Tower in van Gogh’s style well, but some struggled with the 「そもそもの」style. Here the nuance is less “its own” than “from scratch,” as in E05.
With the discussion moving on to human abilities, 絵画のような昔ながらのメディア led to several issues. First, “picture” is a much broader term than 絵画. A more appropriate translation would be “painting,” since the next sentence mentions oil paints. 昔ながらのメディア does not suggest “looking at the history of old media”; the sense is that working in an age-old/traditional medium makes creativity difficult. In the next sentence, the author has deliberately used a parallel structure: 油絵具もカメラもテクノロジーであり、絵画も写真も一種のメディアである。It is helpful to retain that structure: “Oil paints and cameras are both technologies, and paintings and photographs are both types of media.”
All the finalists rendered the next sentence nicely, with E51 using especially vivid language. Humans fumbling around sounds very plausible. E51’s “by the same token” in the next sentence also works well—and would be better without the “however.”
Now the author is approaching his goal: arguing that, in the age of AI, what people need is creativity, cultivated through education that does not turn children into “test-taking robots.” Note that ending with 「だろう」does not imply uncertainty. He is inviting the reader to concur with his argument. “Is likely in need of drastic reform” or “may be what is most in need of” misses the point.
E05
This translator has produced a readable text that is largely accurate but misses some significant nuances. E05 came up with the title “The economy of ‘idea to product’ made possible by AI,” which misses the significance of 即 in the author’s catchy AIが可能にする「アイデア即プロダクト」の経済. “The economy of ‘idea to product’” has been in operation for centuries, if not millennia; it is AI, plus tools such as 3D printers, that can make the instant idea-to-product economy possible.
The statement that “The now much-discussed language-generation AI ChatGPT can not only answer any question” made me think about あらゆる. Can ChatGPT “answer any question”? If we assume “answer” means “answer correctly,” that is clearly false. If we take あらゆるto mean “all sorts of questions,” leaving ありとあらゆるto cover “each and every,” we have a more plausible assertion of AI’s abilities. (And “computer programs” would be better than “programs.”)
In the next sentence, E05 rendered 相応する as “appropriate,” which works nicely, and also used “photographic image” instead of “photo” or “photograph,” having recognized that AI-generated images, however photorealistic, are not what we call photographs.
But why “electronic documents,” when “e-books” are all over the place? In “Soon, by connecting with devices such as 3D printers, individuals will be able to make things like accessories or furniture instantly,” the translator has made it the individuals connecting to those printers, for example, through a sloppy bit of English.
The next sentence repeats the omission of 即 in the translation of the title. Yes, “immediately” later in the sentence gets the point, but the author’s catchy “instant idea to product” notion should not be overlooked. E05 then handles the discussion of AI reducing job prospects in art and design well, though I would prefer ignoring the 等 in 「美術家・デザイナー等」 over translating it as “etc.” Here, the existence of other job categories to which that ratio applies is not relevant to the argument, and “etc.” is almost better avoided.
In parsing the list of other occupations likely to be threatened by AI, E05 has misinterpreted the list. E05’s translations should be in three groups: administrative staff, professionals such as accountants and tax consultants, and teachers and researchers. Also, since 士業 does not correspond to the English definition of “professions” and its meaning is not relevant here, it could be skipped. Note that E05 has correctly assigned the “teacher and researcher” category to the author, although the author might prefer to think of himself as an “academic” or “educator” than a “teacher.”
E05 stayed with the literal “value judgement” in listing AI’s lacks, despite its being an outcome of using judgement, not an ability, but neatly avoided explaining the statistical basis of the Japanese test score term. The “human awaiting instruction” works, and the description of the “super indoor type” is powerful.
“Beach-combing” is rather broader than 潮干狩り, but since the point is the actual physical activity in a physical setting, not edible outcomes, it works. The culminating sentence, “In short, an AI would find it difficult to create a work with originality,” is effective.
E05 was one of the two finalists who capitalized van Gogh’s name correctly. And “from scratch” is a terrific take on そもそもの.
The final paragraph could be simplified by using “cram” instead of “cram information,” but the final phrase, “must undergo drastic reform,” effectively communicates the writer’s passion.
As these comments suggest, E05 came up with some effective solutions.
E20
In general, E20 needs to work more on the “write a little” part of the translation process, to communicate how he or she has understood the text effectively.
“Instant materialization,” in the title and later in the article, definitely gets the sense of 即 but seems too SF-ish, especially since it omits the “idea” part. E20 did do a pleasing job on labeling the author. In the first sentence, “answering any questions” is clumsy English, apart from exaggerating AI’s powers, and “theses” is a bit narrow, and lofty, for 論文、which often are essays without the academic implications of a thesis. (And could we have “computer programs” instead of “programs”?)
The next sentence begins well with “Similarly,” an interpretation of 一方 that makes more sense than “on the other hand.” A little more thought might have helped in the next sentence, in which AI is generating “photos” of “decent quality.” Here “quality” would normally refer to image quality, not whether the image suits the prompt given. Also, these are photorealistic images, not actual photographs. I like the idea of an image of a couple flirting, but question whether it would be too narrow for 戯れる. Playing? Fooling around?
E20 handled the e-book sales nicely and correctly connected AI with the relevant contraptions to produce items, but たちどころに would be more “on the spot” or “instantly” than “on their own.” But materializing ideas? The translation continues well through the sentences about the job-to-applicants ratio and its future.
When we get to the implications for white-collar workers, E20 also had trouble parsing the list. “Clerical staff” is good, but the folks like the author are both academics and researchers. But “It does not mean, however, that AI surpasses humans in every aspect” is in the present tense, while the author is clearly, in 上回るようになる, considering the future.
Getting down to AI’s lacks, we again find “value judgement,” which is a result of applying judgement, not an ability. “Real experience,” however, conveys the bodily nuance of 体験：実際に自分の身をもって経験すること. “Reactive,” as the opposite of “proactive,” does work, if blandly, and “super-indoor types” and “web surfing” are good. In “Never having experienced activities like clamming or campfires, they never create novels or movies based on their real experience,” “their own experiences” might be more effective. And then we come to “generative AIs cannot originate works.” I suspect that if you told ChatGPT to “originate a work of fiction,” for example, it would come up with something—but that work would lack originality. Basically, as an active verb, “originate” generally means “start” rather than “create.” Here, as in “instant materialization,” E20 has chosen a translation that seems succinct and clever but misses the point.
E20’s handling of the final sentence clearly communicates the author’s view that education must “be reformed dramatically,” but the previous part of the sentence is odd as English: why not “drives children to cram” instead of “cramming”? And what is a “scoring machine”?
E42
This finalist came up with some effective translations but started out awkwardly with “The AI-driven economy of ideas and instant products.” “AI-driven” works, but the connection between ideas and instant products is not made. In the first sentence, “on everyone’s lips” is a vivid way to express 話題, but then E42 ends up with “among other things,” here an unnecessarily literal translation.
E42 goes on, by having “typing” modify “AI,” to have the AI typing in the prompts. “That coincide with that input” is good, but the result would be photorealistic images, not photos.
In the next sentence, why write “create content like picture books” instead of “create picture books”?
Next, E42 pushes the user aside and has AI, paired with 3D printers, single-handedly creating things.
E42 handles the threat to art-related occupations well, though the final “this figure will only lower” is awkward. “Decline”? “Worsen”?
E42 sorted out the list of occupations and deftly omitted the “professionals.” “Judgement of value” stayed on the list of lacks, but “first-hand experience” is a good translation of 体験 here.
The rendering of the 「スーパー偏差値エリート」 sentence is disturbing both because of the term “high deviation scores” and because the translator is claiming that AI have those scores, rather than understanding the metaphor the author is using. The overly wordy next sentence, with “likened,” does suggest something of a metaphor in operation, but states that AI “are likened to” (by whom?) rather than “could be likened to.” “Proactively” in the following sentence is good. Including both the 起業 and 新規事業 is not critical, though it would not have hurt.
In discussing those limitations, E42 uses “brainstormed,” which usually refers to a process in which several people (or AIs?) are engaged.
“However, it can’t create style in the first place” could be rewritten as “It can’t, in the first place, create style,” which indicates AI’s essential inability rather than whether it can create a style of its own or from scratch. Also, “create style” does not fit with, for example, “in the style of van Gogh.” “A style” implies a countable noun, as in the styles of van Gogh, Pissarro, and Hokusai. “Style” as an uncountable noun could be replaced by “elan” or “luxury,” for example.
The translation of the section about hope for continued human creativity works well. The final sentence, with its “point-grabbing machines,” could be less literal but works quite well.
E46
“The Instant Idea-to-Product Economy and the AI Behind It” is an excellent title, and E46 labels the author nicely. The first sentence is also very effective, with “all manner of questions” a great way to handle あらゆる. (The “as well” at the end could be cut.)
In the next sentence, “on the other hand” implies a contrast with the first sentence that does not exist in the Japanese. “Similarly” (E20), “while” (E05), or “meanwhile” (E51) work better for the 一方. E46 then handles the image generation well, with photorealistic images that match the prompts. E46 also deals with the e-books and the creation of objects well, except that たちどころに is more “on the spot” or “instantly” than “on their own.”
E46 has awarded office workers certificates in misinterpreting the list of occupations. The “(such as myself)” is, however, very effective in capturing the author’s tone. And “trump human workers in all regards” is also a vigorous turn of phrase. E46 then goes on to use “volition, experience, and judgement” for AI’s lacks. “Judgement” works well, but I would have liked to see “real,” “first-hand,” or even “personal” added to “experience.” The phrase “crème de la crème of the intellectual elite” is clever, but (in the author’s world) are those who score well on tests the intellectual elite?
E46 was the only finalist to distinguish 起業 and 新規事業 well, and E46 describes the effects of AI’s limitations carefully. “Super shut-in” is excellent, but “digging for shellfish when the tide is low” overexplains 潮干狩り and makes me wonder if E46 has ever gone clamming. “Coming up with a slew of novel ideas” is powerful and accurate, and “finds itself at a loss” is a deft way to handle the 難しい。This translator goes on to handle the rest of the text well, with “text-taking robots” a brilliant solution to those 点取りマシーンに. Still, the “is likely in need of drastic reform” weakens the author’s statement. Here, the final だろう is not expressing uncertainty but inviting the reader to agree with him.
E51
This finalist’s translation of the title, “AI and the Birth of the Idea-to-Product Economy,” is excellent, the author is labeled correctly, and “abuzz over ChatGPT” is excellent. E51 also handled the turn to image generation smoothly, using “meanwhile” for the transition and describing the output as “photorealistic images.” The discussion of the impact on jobs begins well with “conversely” and the current job-applicant ratio is explained succinctly. E51 also sorts out the white-collar job list and dares to drop “professionals,” though confusingly adding “professions” at the end.
In the discussion of AI’s lacks, E51 uses “intention, experience, and judgement.” “Real experience” might have been better.
In the next sentence, E51 has connected the glorious scores and the passive behavior to employees and employee candidates. That is reasonable, given the earlier discussion of declining job possibilities, but might be rethought, given that the author is going on to discuss AI capabilities and lacks more broadly.
Replacing “oil paints” with “paintbrushes” is daring but effective. 絵画も写真 are, however “paintings” and “photographs,” not the actions of painting and engaging in photography.
The next several sentences work well, but the final sentence needs more thought. “May be what is most in need of drastic reform” seriously weakens the author’s statemen.
（Continued to Part 4 by Jim Davis）