Clearly, getting machine-like reading comprehension is easy: just react instinctively to the lede, utterly ignore the core of the article as TL;DR, put words in author's mouth for good measure. That is not quite the kind of equivalence that the field is striving for: "humans can't read, machines can't either, QED."

Douglas R. Hofstadter is a tad more than a translator. Try Google.

Your response simplifies a pretty specialized occupation. Translating a paragraph is not simple math. Most of the time ‘understanding’ natural language is AI complete. Translation is not just looking up words but also considering culture and history. So maybe don’t be so sure that with the correct ‘magical’ technique we can achieve human level translation...

Do you speak a second human language that’s not closely related to your first?

English and Russian, say?

Because I find most data scientists that do tend to think language translation may be one of the last things we solve. I think it requires hard AI, honestly.

Maybe look at the progress we have made in last twenty years? Yes, there are examples of machine translation failing miserably, but it has also reached a level which allow people to actually use it for communication (as evidenced in real life and mentioned in the article). I think that's already quite an achievement.

We should not be too harsh on computers. Humans are not perfect either. Instead of judging machine translation based on few random cases, we should be evaluating against number of human translators of various experience levels.

The level of communication it enables is thin. Far from even simple conversations and relies heavily on humans having a “theory of non-mind” of the machine.

I speak three: Russian (native), English (mistaken for native speaker by native speakers) and Hebrew (average). I may not be an expert in AI though (couple of courses, hobby projects and worked alongside AI researchers on the non-ML side of some apps). Still, looking at the methods we already have and recent successes (like Yandex's linguistic work), it doesn't like as hard a problem as hard AI with human-like cognition.

Could you please not post uncivil or unsubstantive comments to HN? We're trying for better than that here.