"holy grail of language learning is the ability to learn a language outside the country that speaks it, to a level of fluency that puts the user within 30 to 45 days of advanced reading, speaking, and vocal comprehension "
As someone who has learned several languages to near native fluency and majored in linguistics at the University, I think the OP is right on in his estimate of the goal. But the biggest problem software in this area is facing is that language learning is not a one way street. You can't memorize thousands of words in all the contexts in which they occur and expect to be able to have any competency, at least for speaking & vocal comprehensione. Even when the software provides dialoges and attempts to simulate conversations (ie. "Pimsleur" and others), it fails because its the same over and over, mostly outside of real world contexts, where background noise, dialects, and other tasks are involved. A new language learner has to be able to respond rapidly and fluidly in spoken conversation over a variety of subjects if he/she wants to be able to function at the level the OP is talking about, and current software simply doesn't provide the degree of realism in these types of situations.
IMHO, in order for any software to be able to achieve the above stated goal, I think it would need to first have a very good humanlike AI. why? So you can have a "human partner" in conversations over a variety of subjects, where the learner is forced to quickly respond and interact with the partner/software. Otherwise, you're left with all the current options which are at best subpar in comparison to actually living and interacting with native speakers in the said country.
I had a friend move to Argentina and in order to improve his Spanish, he found a group that pairs fluent Spanish speakers learning English with fluent English speakers learning Spanish for one-on-one conversations. It was pretty informal in that the pair could meet wherever and whenever they wanted. Half the time they spoke in English, half the time they spoke in Spanish.
Perhaps software that scheduled live video chats of similar conversations between learners could help with the problem you bring up? Given that most language software aims at multiple languages anyway, seems like the only challenge is technical with the video chat (latency might be a problem although audio is pretty good too) and the scheduling interface. And I suppose enough users in each language willing to do such conversations.
While we can't give you conversational interaction with a human-level AI yet, I've been working for the past couple of years on a project for The Language Flagship (http://www.thelanguageflagship.org/) to try to solve the real-world context problem by creating a searchable database of re-usable multimedia resources for language learning, so that you can always give a learner new stuff. It's kinda like Duolingo, teaching you to translate the web, but with audio, images, and video along with the text. It's mostly vaporware still, but we do have some really nice tools for teachers for making playlists, video captions, transcripts, and annotations.
Additionally, there's a startup here in Utah that we've been working with called MovieMouth, which produces software and auxilliary content for teaching language comprehension via popular movies (they get around copyright problems by only providing playlist files and original content like test questions, and requiring that end-users have independent access to the video media to feed into their software). They've been running beta tests, teaching English with Pixar movies, down in Chile .
I couldn't agree more. My startup, Nulu ( http://www.nulu.com/ ) , is focused on two big things: bringing people new and interesting content every day (we use news content as the basis for learning) and super high quality support (native speaker translation done at the phrase level, as well as professional voice artist audio for the whole story).
Check it out, and let me know whether it's useful to you! We do English for Spanish speakers and Spanish for English speakers right now, with many more languages to come.
As someone who has learned several languages to near native fluency and majored in linguistics at the University, I think the OP is right on in his estimate of the goal. But the biggest problem software in this area is facing is that language learning is not a one way street. You can't memorize thousands of words in all the contexts in which they occur and expect to be able to have any competency, at least for speaking & vocal comprehensione. Even when the software provides dialoges and attempts to simulate conversations (ie. "Pimsleur" and others), it fails because its the same over and over, mostly outside of real world contexts, where background noise, dialects, and other tasks are involved. A new language learner has to be able to respond rapidly and fluidly in spoken conversation over a variety of subjects if he/she wants to be able to function at the level the OP is talking about, and current software simply doesn't provide the degree of realism in these types of situations.
IMHO, in order for any software to be able to achieve the above stated goal, I think it would need to first have a very good humanlike AI. why? So you can have a "human partner" in conversations over a variety of subjects, where the learner is forced to quickly respond and interact with the partner/software. Otherwise, you're left with all the current options which are at best subpar in comparison to actually living and interacting with native speakers in the said country.