[extropy-chat] i-language again

[Spike]

> Damien Broderick

> > >Overview for "spike"
> > >
> > >The noun "spike" has 9 senses in WordNet...

> Harvey Newstrom
> 
> Excellent(1)!
> 
> ¿ Butperomaisaberma whyporquépourquoiwarumperchè notnonicht
> inventinventeinventezerfindeninventi aun 
> languagelengualanguesprachelingua
> thatesoceladasquello everybodytodostoutjederognuno canlataludoselatta
> readleídobidongelesencolto ?
> 
> -- 
> Harvey Newstrom

The notion is to take advantage of our computers to do
what computers do so very well: to learn and remember stuff.
This would allow our brains to do what brains do so well:
learn one language early in life, then avoid the
educational overhead required to pick up others.

The system I propose does require computers to translate.
What I have been thinking about for some time is a slight 
variation on each of our native languages, that are as 
close to the original language as possible, yet would be
universally machine translatable.

Stay with this line of reasoning: I think we are drawing
close to a possible solution.  It would require some sort
of universal dictionary, in which the definitions all
appear as pictures.  This is something that could never
have been done before the internet existed.

Example, spike.  (1) picture of a railroad spike.
(2) picture of a athletic shoe with cleats.
(3) picture of graph with a sudden upturn.
...
(11) picture of Spike Jones the musician, etc.

Then create in this universal graphical dictionary
a list of equivalent words in various columns: 
German in column 1, French in 2, English 3, etc.  Then 
anyone with a computer would be able to unambiguously
translate any text into any other language, without
loss of meaning provided the original writer would number 
each of the ambiguous terms (about a quarter of the words
I would estimate).

If a language has no word for a certain concept, it
merely needs to borrow one from any other language.
We do that now all the time: languages borrow technical
terms from whoever invents them.  Rocket science is
*filled* with German terms for instance.  

This kind of writing would be laborious for sure, but 
text in this form would be universally accessible to 
anyone who understands *any* language, but more importantly, 
text that has been de-ambiguationed is accessible by AI.
Text in this form will educate AI, and it will bootstrap
from that to all the rest of the text.

Perhaps text in the appropriate form would contribute
to the probability that the first AI is friendly.

Eliezer, this thread cries out for your comments.

spike