Oh give us a break, Douglas. Yeah, the guy is a genius and all, but this should be titled "The Shallowness of Douglas Hofstadter and dozens of straw-man arguments." No one who knows the first thing about computers, AI, or machine translation would assert that it is remotely close to human-level translation, or that computers have any sense of meaning. This statement is ridiculous:
Before showing my findings, though, I should point out that an ambiguity in the adjective “deep” is being exploited here. When one hears that Google bought a company called DeepMind whose products have “deep neural networks” enhanced by “deep learning,” one cannot help taking the word “deep” to mean “profound,” and thus “powerful,” “insightful,” “wise.” And yet, the meaning of “deep” in this context comes simply from the fact that these neural networks have more layers (12, say) than do older networks . . . No one is exploiting anying, or trying to fool people into thinking the programs have "depth" in the sense he means. "Deep neural networks" and "deep learning" are narrowly defined technical terms. No one who has studied AI would be fooled by them, any more than you might think subatomic quarks really are charming, as in "delightful." This is also ridiculous: It’s hard for a human, with a lifetime of experience and understanding and of using words in a meaningful way, to realize how devoid of content all the words thrown onto the screen by Google Translate are. It’s almost irresistible for people to presume that a piece of software that deals so fluently with words must surely know what they mean. It isn't hard for me! Because I studied linguistics at Cornell, I have translated many documents, and I written lots of programs. It isn't just easy for me; it is second nature. When I look at a Google translation, I can tell at a glance what it was doing and how it made a "mistake" (a misnomer in this context). I know what deep networks are, and I know that AI presently has no sense of meaning whatever -- the experts are working on that. I was fully aware of every problem and limitation with machine translation discussed in this article. However, despite these problems, Google translate does a pretty good job with weather reports, patents, and electrochemical papers. Not novels, for crying out loud! What would you expect?? The "intelligence" of the Google-plex supercomputer, and all the other supercomputers, is roughly on par with the intelligence of a mouse, or the collective intelligence of a colony of bees. It is hundreds of thousands, or millions of times, less than human intelligence. So it can only perform a crude imitation of human cognition. AI does exceed human abilities in a narrow range of problems, such as playing Go, recognizing faces, or determining that a young woman who shops at Target is pregnant before her father realizes that fact. It is not surprising that an intelligence far less than ours can exceed ours in some ways. The collective intelligence of a bee colony can solve many problems much better than we humans can, such as: finding and collecting nectar, constructing remarkably effective honey storage devices (honeycombs), and cooling hives on hot days. A giant human brain that can translate language is not particularly good at finding nectar. Machine translation is nothing like human translation because the machines are still a million times less intelligent than we are, and they are missing many crucial aspects of our intelligence. Perhaps, in the future, as the technology improves, these problems will lessen. But I doubt machines will ever think the way we do, and language is natural behavior evolved to work with our brains. Not with synthetic silicon-based thinking machines. It is remarkable that Google has managed to teach something resembling a mouse to do anything remotely resembling translation. It works pretty well, considering how difficult that is.
