If language evolved to allow us to exchange information, how come most people cannot understand what most other people are saying? This perennial question was famously addressed in the Old Testament story of the Tower of Babel. … The real puzzle is that the greatest diversity of human societies and languages arises not where people are most spread out, but where they are most closely packed together. Papua New Guinea is a classic case. That relatively small land mass – only slightly larger than California – is home to between 800 and 1000 distinct languages, or around 15 per cent of all languages spoken on the planet. This linguistic diversity is not the result of migration and physical isolation of different populations. Instead, people living in close quarters seem to have chosen to separate into many distinct societies, leading lives so separate that they have become incapable of talking to one another. Why? …
Actually, the Simplified Chinese character set was a created as a government effort specifically to increase the literacy rate by making the writing system easier to learn.
Reading a Smithsonian article reminded me: why do you say that this is a problem when the 20th and 21st century have witnessed the deaths of many languages with thousands of language extinctions estimated in the future?
"The general consensus is that there are between 6000[2] and 7000 languages currently spoken, and that between 50-90% of those will have become extinct by the year 2100.[1]" http://en.wikipedia.org/wik... or: "The top 20 languages spoken by more than 50 million speakers each, are spoken by 50% of the world's population, whereas many of the other languages are spoken by small communities, most of them with less than 10,000 speakers."
Isn't this exactly what you would expect on economic and coordination grounds? With an increasingly developed global economy, small languages lose their benefits compared to the opportunity costs, with ever more coordination on a few very widespread languages and a single lingua franca, while medium size languages poke along and the top languages enjoy outsized benefits (consider how many people speak any form of English even after the fall of the English Empire).
>A maximally general ontology would be "stuff exists". The more general you make it, the less informational content it has. How much use is something with minimal informational content?--PDJ
Yes, we obviously don't want something with minimal informational content. But that's not what I was suggesting.
Recall the context of what we were talking about. That was minds/artificial intelligence. I wanted an ontology sufficiently general to cover the domain 'minds', NOT an ontology concerned with the more nebulous domain 'reality'. So that does change the game somewhat.
We can then ask what ontology is maximally general within the context of the domain 'intelligent minds' and this question is one which doesn't fall into the trap of minimal information content.
To sum up: We looking for a language sufficiently general to fully describe the domain 'intelligent minds'. This is all we need, because this would include all thoughts that minds could ever think about reality.
"Not one based on a fixed ontology with neccessary definitions, but an open-ended one based on prototypes."
A maximally general ontology would be "stuff exists". The more general you make it, the less informational content it has. How much use is something with minimal informational content?
>People have been trying to come up with unequivocal thought-languages for centuries, and none of them managed to transcend their own world view.
The problem was that none of these people were me. Are you sure I haven't already done it? ;)
>You can't start from omniscience.
You can't start from a blank slate either.
>*General* intelligence has to be fluid, flexible intelligence, not intelligence that's constrained to one ontology..
But that's exactly why you need a sufficiently general built-in ontological framework to start with. Not one based on a fixed ontology with neccessary definitions, but an open-ended one based on prototypes.
"I think Robin writes off those language standardization efforts as being just "Sometimes a nation will push hard to get everyone in the nation to speak the same basic language, in order to strengthen national solidarity".
I don't see the point. It might have that *motivation*, but then it still might have the side-effect of enhancing communication.
Sort of. Influential writers tend to write in a formal, official language rather than a street slang. OTOH, given academic specialisation, highly educated writers will also tend to write for an audience of specialists, with jargon to match.
"The probability stuff scarcely scratches the surface of the deep ontology problem. The categorizations and logical pinpointings are always PRIOR to probabilistic calculations. "
Strongly agree. The LW approach is misplaced effort. They obsess about numbers and calculations, but that is no use until you have got the conceptual map sorted.
"An entirely new general-purpose 'language of thought' needs to be built up from scratch, starting from a finite set of ontological 'prims', which constitute the basic 'building blocks' of thought. "
Strongly disagree. People have been trying to come up with unequivocal thought-languages for centuries, and none of them managed to transcend their own world view. (A 17thC version had God as a fundamental category. Do you see LW buying that?)You can't start from omniscience. *General* intelligence has to be fluid, flexible intelligence, not intelligence that's constrained to one ontology..or one way of calculating probabilities, for that matter.
Sort of. Actually a lot of them are rather rarified concepts. And a fair chunk of it is caused by simply no knowing that there is an existing term. Ie re-invention of wheels rather than ingroup-signalling. Although the latter might explain why they don't switch to the established terms.
". Sometimes a nation will push hard to get everyone in the nation to speak the same basic language, in order to strengthen national solidarity"
I think that is actually usual, not "sometimes". Pretty much everything we call a language is a set of dialects, and pretty much every nation teaches a standard or official dialect in schools. Standard Spanish is Castillian, standard German is the Prussian dialect, and so on.
In China the push for standardizing the script was to restore order (Qing). Several languages used the same written script, and they (the confucians anyway) wanted each object to have a unique and proper name, else political chaos would take over.
In the 20th century script was modernized as to make reading/writting easier.
Even today China is far from being a (spoken) unilingual country.
My usage of the word 'language' is in the most general sense to mean a representational system of manipulating concepts and logical relations. So this would include mathematics, data modelling and ontology, 'languages' far more powerful than ordinary natural languages used for speaking.
Even such a low-level task as object recognition can be reinterpreted as a language problem in this more general sense, by using the notion of 'logical pinpointing' (moving from abstract to ever more specific systems of logic) and viewing computation as a model (or simulation) of reality (with the program and data models constituting the 'language').
If you want to signal high status than you need a costly signal.
If you just want to signal that you are a fan of a certain band, wearing a T-Shirt of the band is a very clear signal. No observer would think that you are lying about being a fan of the band because it's not valuable to lie.
As far as I understood the purpose of the Chinese/Japanese efforts was to make it easier to become literate in those languages.It wasn't about preventing people from speaking other languages.
When it comes to pushing to get everyone to speak the same language I rather think about events in countries like Turkey, where Kurdish people get forced to speak some Turkish.
Actually, the Simplified Chinese character set was a created as a government effort specifically to increase the literacy rate by making the writing system easier to learn.
Reading a Smithsonian article reminded me: why do you say that this is a problem when the 20th and 21st century have witnessed the deaths of many languages with thousands of language extinctions estimated in the future?
"The general consensus is that there are between 6000[2] and 7000 languages currently spoken, and that between 50-90% of those will have become extinct by the year 2100.[1]" http://en.wikipedia.org/wik... or: "The top 20 languages spoken by more than 50 million speakers each, are spoken by 50% of the world's population, whereas many of the other languages are spoken by small communities, most of them with less than 10,000 speakers."
Isn't this exactly what you would expect on economic and coordination grounds? With an increasingly developed global economy, small languages lose their benefits compared to the opportunity costs, with ever more coordination on a few very widespread languages and a single lingua franca, while medium size languages poke along and the top languages enjoy outsized benefits (consider how many people speak any form of English even after the fall of the English Empire).
>A maximally general ontology would be "stuff exists". The more general you make it, the less informational content it has. How much use is something with minimal informational content?--PDJ
Yes, we obviously don't want something with minimal informational content. But that's not what I was suggesting.
Recall the context of what we were talking about. That was minds/artificial intelligence. I wanted an ontology sufficiently general to cover the domain 'minds', NOT an ontology concerned with the more nebulous domain 'reality'. So that does change the game somewhat.
We can then ask what ontology is maximally general within the context of the domain 'intelligent minds' and this question is one which doesn't fall into the trap of minimal information content.
To sum up: We looking for a language sufficiently general to fully describe the domain 'intelligent minds'. This is all we need, because this would include all thoughts that minds could ever think about reality.
"You can't start from a blank slate either."
And as for false dichotomies...:-)
"Not one based on a fixed ontology with neccessary definitions, but an open-ended one based on prototypes."
A maximally general ontology would be "stuff exists". The more general you make it, the less informational content it has. How much use is something with minimal informational content?
--PDJ
>People have been trying to come up with unequivocal thought-languages for centuries, and none of them managed to transcend their own world view.
The problem was that none of these people were me. Are you sure I haven't already done it? ;)
>You can't start from omniscience.
You can't start from a blank slate either.
>*General* intelligence has to be fluid, flexible intelligence, not intelligence that's constrained to one ontology..
But that's exactly why you need a sufficiently general built-in ontological framework to start with. Not one based on a fixed ontology with neccessary definitions, but an open-ended one based on prototypes.
"I think Robin writes off those language standardization efforts as being just "Sometimes a nation will push hard to get everyone in the nation to speak the same basic language, in order to strengthen national solidarity".
I don't see the point. It might have that *motivation*, but then it still might have the side-effect of enhancing communication.
Sort of. Influential writers tend to write in a formal, official language rather than a street slang. OTOH, given academic specialisation, highly educated writers will also tend to write for an audience of specialists, with jargon to match.
"The probability stuff scarcely scratches the surface of the deep ontology problem. The categorizations and logical pinpointings are always PRIOR to probabilistic calculations. "
Strongly agree. The LW approach is misplaced effort. They obsess about numbers and calculations, but that is no use until you have got the conceptual map sorted.
"An entirely new general-purpose 'language of thought' needs to be built up from scratch, starting from a finite set of ontological 'prims', which constitute the basic 'building blocks' of thought. "
Strongly disagree. People have been trying to come up with unequivocal thought-languages for centuries, and none of them managed to transcend their own world view. (A 17thC version had God as a fundamental category. Do you see LW buying that?)You can't start from omniscience. *General* intelligence has to be fluid, flexible intelligence, not intelligence that's constrained to one ontology..or one way of calculating probabilities, for that matter.
Sort of. Actually a lot of them are rather rarified concepts. And a fair chunk of it is caused by simply no knowing that there is an existing term. Ie re-invention of wheels rather than ingroup-signalling. Although the latter might explain why they don't switch to the established terms.
". Sometimes a nation will push hard to get everyone in the nation to speak the same basic language, in order to strengthen national solidarity"
I think that is actually usual, not "sometimes". Pretty much everything we call a language is a set of dialects, and pretty much every nation teaches a standard or official dialect in schools. Standard Spanish is Castillian, standard German is the Prussian dialect, and so on.
I don't know for Japan.
In China the push for standardizing the script was to restore order (Qing). Several languages used the same written script, and they (the confucians anyway) wanted each object to have a unique and proper name, else political chaos would take over.
In the 20th century script was modernized as to make reading/writting easier.
Even today China is far from being a (spoken) unilingual country.
My usage of the word 'language' is in the most general sense to mean a representational system of manipulating concepts and logical relations. So this would include mathematics, data modelling and ontology, 'languages' far more powerful than ordinary natural languages used for speaking.
Even such a low-level task as object recognition can be reinterpreted as a language problem in this more general sense, by using the notion of 'logical pinpointing' (moving from abstract to ever more specific systems of logic) and viewing computation as a model (or simulation) of reality (with the program and data models constituting the 'language').
If you want to signal high status than you need a costly signal.
If you just want to signal that you are a fan of a certain band, wearing a T-Shirt of the band is a very clear signal. No observer would think that you are lying about being a fan of the band because it's not valuable to lie.
As far as I understood the purpose of the Chinese/Japanese efforts was to make it easier to become literate in those languages.It wasn't about preventing people from speaking other languages.
When it comes to pushing to get everyone to speak the same language I rather think about events in countries like Turkey, where Kurdish people get forced to speak some Turkish.
Yes, but "often go out of their way to show off how X they are" is essentially the wordy phrase "signalling" was created to replace.
You don't update priors, you update posteriors! *mutters about insufficient group loyalty*
I do agree that "reconsider" is a hell of a lot less pretentious than "update", humans aren't AI's after all.