LAST week’s column looked at how machine translation (MT) has—and has not—improved. Free services like Bing and Google Translate can give quick-and-dirty, mostly-correct translations for tourists and the curious most of the time. For professional uses, machine-translated material must be post-edited for both accuracy and style. With restricted subject matter, MT systems can be trained to choose the best translations for words with multiple meanings. This is why (for example) the European Commission uses MT extensively. The legalistic language of the European Union may be impenetrable to outsiders, but the narrow range of bureaucratic language makes translating it much easier.

All this is getting better as computers get faster, storage cheaper and software smarter. But MT has improved only gradually, not in the revolutionary leaps and bounds seen in other fields. It is an example of the truism that machines find it easy to do things humans find hard (vast maths problems), and yet find it hard to do things humans find easy (language, natural movement). (Some experts have already begun pouring cold water over last weekend’s reports that a machine has finally passed the artificial-intelligence “Turing Test”, for example.)

No one knows how MT will look in 25 years. But that doesn’t stop us from guessing. At the TAUS conference on MT in Dublin last week, Johnson was invited to debate with Nicholas Ostler on the lingua franca of the future. Mr Ostler is a historian of the long sweep of languages’ lives. His most recent book, “The Last Lingua Franca” (reviewed here), laid out the arguments he presented in Dublin. English is actually shrinking, in percentage terms, as a mother tongue. (Other languages’ speakers are having more babies.) And foreign languages associated with dominating groups (first colonial Britain and then hyperpower America, in this case) can stir resentment, so it is not guaranteed that people in the future will always want to learn English.

Meanwhile, Mr Ostler has high hopes for MT. All of that increased computing power should mean that, for the vast majority of the world’s people, the quick-and-dirty translations available from the likes of Google can only get better. In the long run, MT will be a better option for most people than slaving over learning English for years. Most people, after all, spend most of their lives working and living in their native language.

Johnson presented the case for English (a predictive case, not a hoped-for one): English has a reach and penetration unlike any language in history. It is now spoken by twice as many non-natives as natives, increasingly shedding its association with America and Britain. (When a Swede negotiates with a Brazilian taxi driver, or a Hungarian attends a conference in Poland, they are not thinking of American foreign policy when they pragmatically use English.)  Schools are introducing English at earlier and earlier levels: Denmark is beginning English in first grade, and Zurich has chosen to teach pupils English before French, the second-biggest language of Switzerland. As much as it may chagrin French-speakers, such decisions are entirely practical and can be expected in ever greater numbers.

The effect of this, Johnson predicts, is that the lock-in English now enjoys will only get stronger. Crucially, English will begin to be taken for granted. Every child will one day get it in school (as every child in China now does). They will hear so much English in early years that acquisition of a decent fluency will be easier and easier. Technology isn’t only helping machine translation. It is giving children around the world television, music and movies in English. Ambitious families will ensure that their kids see as much as possible, as early as possible, so they speak English not just competently, but fluently and comfortably. And those kids will increasingly choose English themselves. It opens up social networks; there’s a lot more on Twitter if you speak English. English even opens up games: youngsters round the world learn English to chat while they play online games like Minecraft and Worlds of Warcraft. This early and frequent exposure to English will mean the effort to learn, which Mr Ostler describes, begins to seem a lot less wearisome. It is simply part of the environment, something that billions of children will know a decent bit of before they even begin their first class. And as they progress up the grades, incentives will kick in, as pupils hear again and again how many doors English will unlock for them. Add to that an ever-strenghtening network effect: the more people who speak English, the more useful it is to speak English. It is hard to see what will stop this momentum in the next few decades.

In the course of Johnson’s discussion with Mr Ostler, we agreed on many things, and found a certain synthesis. English is still a language of elites, those well educated or in the kind of well-paid globalised professions that required it. Machine translation has come a long way in the past decade. For many people who are born, live and die without ever leaving their home regions, MT will be good enough for the few times in their lives they need to interact with foreigners. Speech recognition has got a lot better, making slow and carefully enunciated speech decently (if not brilliantly) translatable.

Johnson’s closing, though, was this: MT is improving with written texts, but it has a very long way to go in interpreting speech. As English gains ground, neither speech recognition nor MT will come far enough to replace the loud, unstructured conversations businesspeople have as they hash out deals in a noisy hotel bar. Who is so confident in MT that they would rely on it for a job interview? What about a first date? And, if successful, the subsequent marriage? No one can say when MT will be reilable for interpreting the quick, context-dependent and unstructured mess that is live human speech. A generation, at least—a generation in which English’s incredible penetration around the world will only deepen.

Johnson must confess: he narrowly lost the audience vote. Of course, there need be no single winner: both MT and English clearly have a future. But contests and debates are nonetheless illuminating, not to mention fun. So share your predictions in the comments.