Geocurrents is a fun blog. I recommend it.
But they can, like all blogs, post some ill-thought-out stuff. They have recently posted a putative takedown of Larry Summers’ argument that universities should cut the resources that they use to teach foreign languages.
The argument: English is widely-spoken and pretty much all you need for business. Add that to the ever-increasing efficiency of machine translation, and the case for expending mass resources to teach foreign languages disappears. Thus, we should cut back.
I was looking forward to a solid criticism of this argument. Instead, Geocurrents avoids Summers’ substantive points and then at the end undercuts itself. Since I think that Summers is probably wrong, I find this frustrating.
Let’s proceed, in a lesson about How Not To Build an Argument:
(1) The author points out that most of the world does not, in fact, speak English. Since Summers never argued that they did, it is hard to see the relevance of this point.
(2) The author then points out that a lot of immigrants to the United States speak a language other than English. Of course, their children lose those languages super-fast. Moreover, Summers never claimed that there were no non-English-speakers in the United States, and I have never anybody justify the resources spent on foreign-language teaching by saying that we need to communicate better with immigrants. If that is the author’s argument, then she needs to make it explicitly.
(3) Finally, after two non sequiturs, the author states that machine translation is imperfect. This is perfectly true. Some students of mine recently presented me with a Pemex integrated service contract machine-translated from Spanish: it was so incomprehensible that we went back to the Spanish and translated it ourselves. But ...
First, machine translation is getting better all the time. Summers’ argument is about the future. I want to know why the author of the Geocurrents post believes that translation will be as bad in 2050 as it is now! And I am superlatively annoyed that she does not tell me.
Second, the author then presents a frighteningly good and clear machine translation of a Russian poem. No, it isn’t as poetic as the human translations. But it is completely comprehensible. And since Summers is talking about “doing business in Asia, treating patients in Africa or helping resolve conflicts in the Middle East,” she needs to explain why the level of machine competence shown by her own example is not sufficient to those tasks.
At which point she also needs to explain why machine translation won’t become good enough by 2050. (Or, heck, by 2015!) But I repeat.
In short, the riposte does absolutely zero to grapple with Summers’ contention. It is an example of how not to construct an argument, because it doesn’t actually construct an argument.
Comments are more than welcome. How would you argue against Summers?