Yeah yeah, and the input hypothesis has it’s own critics about what exactly “comprehensible input” even is.
…but let’s face it: either extreme (only input, only output) is clearly stupid.
Predictive models require a feedback loop; if you don’t have any means of correcting the predictions of good output, you’ll never refine your model.
This is true of AI models, and it’s true of humans. It can’t not be true.
The point is that you can only learn meaningfully from annotated inputs, like tv, or like talking to people. The input just has to be at the right level.
...
Should people start with listening instead of talking to people? Probably not. Probably they should do both. Why not do both? It's crazy to suggest you should only one or the other.
Probably, the feedback loop in deep immersion of a language, where you have it around you all the time, in a country where that is the native language is the best possible way to encounter lots of examples at lots of different levels to learn from. Probably, you get better input from speaking, because it means people talk to you at a level similar to what you're already at.
Is it about actually speaking? No.
…but it’s patently ridiculous to suggest you could go to another country and just sit silent and listen for three months.
This is pretty interesting tech, and frankly, amazing that it works at well.
Is it perfect? No. Is it another useful tool? It certainly is.
Comprehensible input has a pretty straight forward definition. It's language before or just after the end of understanding. Maybe 80-90 percent known words so you can get the context of what's being said.
If you can't grab hold of context of what's being said you won't really pick much up.
That's why total immersion isn't effective for beginners.
…but let’s face it: either extreme (only input, only output) is clearly stupid.
Predictive models require a feedback loop; if you don’t have any means of correcting the predictions of good output, you’ll never refine your model.
This is true of AI models, and it’s true of humans. It can’t not be true.
The point is that you can only learn meaningfully from annotated inputs, like tv, or like talking to people. The input just has to be at the right level.
...
Should people start with listening instead of talking to people? Probably not. Probably they should do both. Why not do both? It's crazy to suggest you should only one or the other.
Probably, the feedback loop in deep immersion of a language, where you have it around you all the time, in a country where that is the native language is the best possible way to encounter lots of examples at lots of different levels to learn from. Probably, you get better input from speaking, because it means people talk to you at a level similar to what you're already at.
Is it about actually speaking? No.
…but it’s patently ridiculous to suggest you could go to another country and just sit silent and listen for three months.
This is pretty interesting tech, and frankly, amazing that it works at well.
Is it perfect? No. Is it another useful tool? It certainly is.