Previous entry : Extreme Markup Languages 2004
Next entry : Trackback doesn't work for me
July 19, 2004To Martin, more on the unbearable inevitability of discretization
These are additional details, inspired by Martin Lessard's comments .
I don't think you are too liberal artistically to use my semantic web search engine
, as you so kindly put it. ;-) I don't think pre-empirical sciences (i.e. literature, theology, politics, etc.) can't benefit from a not-so-distant, intelligent semantic web. I even believe that they will be able to use it with no more technology than the rational
users require. What I do believe (and meant) is :
- A semantic web query engine able to identify from free, unstructured, unindexed, markup-less text,
poems that have bird as metaphor for freedom
is not simply intelligent, but also more intelligent than the vast majority of humans at this very task. - A query like
get all the poems that have bird as a metaphor for freedom
can mean anything fromget all the poems that were intended to have bird as a metaphor for freedom
toget all the poems that could be understood to have bird as a metaphor for freedom, at some point in time (100 years ago?, now?, 100 years from now?)
. Such a query cannot logically discriminate any text that hasbird
in it. What you want, then, is for the engine to assume the intent of the user from use case statistical inference, which is pretty easy to do. But you also induce the user in error on the correct meaning ofpoem that have bird as a metaphor for freedom
, for which the correct answer isany text, in any language, that refers to birds
. Just like the pre-CSS graphic designers used HTML table positioning, your usage of the semantic web is wrong, as far as semiotics go. A rational query would beget all text that have been identified as poems and have been said to use birds as a metaphor for freedom
which is an extremely easy thing to parse and process. - The mathematical newspeak does not reduce expressiveness (whatever that means). On the contrary, it significantly increases communication efficiency.
- If the minimal standard set (TCP/IP, HTTP and Unicode) is universal enough for you, the semantic web cannot be considered to be intrinsically fragmented : it is rather a uniform cyber-representation of a fragmented world. In the worst-case scenario, cross-fragment semiosis is always possible through ontological harmonization, which also seems to be mechanizable.
All in all, for a query such as what does God want me to do now?
I would prefer my query engine to answer go get laid
than according to the Bible...
.
Human aren't specifically good at logic statement (btw Google is taking advantage of it) : in a semantic web query engine one might put "bird freedom" and expect tons of birds metaphore poems. From the point of view of the engine these expectations make no sense at all as "poem" or "metaphore" concepts wasn't explicit.
We human do think in context. Query field , as a matter of fact, does scrap all context : you ask us to be specific in your query field. At that point I find it is cheating : we give part of the answer to the "AI". Human use every bit of his environment to extract context and understand the on-going communication (a wink, a body language, etc).
How semantic web may cope with that? By asking user to be more specific? Your exemple is quite clear on that. But by lacking rationality in our query, I think this shows we, human, express our rationality on a different level. At a matter of fact, by asking the use of "mathematical newspeak" you're increasing comprehension (by a high magnitude) between human and the machine. Isn't like cheating? I mean shouldn't AI be able to catch us on the intelligence race, therefor be able to understand context?
I understand I have to specified what I mean by "context" and why Semantic web need it. For now I may just ask myself if (logic-query impaired) human can publish on the web documents that isn't full of non-logic assertions that may bug an Agent's report. Because we do use natural language and from a logic point of view it isn't always using univoque meaning (ex.: inclusive OR vs Exclusive OR doesn't exist in French ("ou") but contemporary logic doesn't clearly differentiate it.
Posted by: Martin Lessard on July 20, 2004 11:48 PMHuman aren't specifically good at logic statement (btw Google is taking advantage of it) : in a semantic web query engine one might put "bird freedom" and expect tons of birds metaphore poems. From the point of view of the engine these expectations make no sense at all as "poem" or "metaphore" concepts wasn't explicit.
We human do think in context. Query field , as a matter of fact, does scrap all context : you ask us to be specific in your query field. At that point I find it is cheating : we give part of the answer to the "AI". Human use every bit of his environment to extract context and understand the on-going communication (a wink, a body language, etc).
How semantic web may cope with that? By asking user to be more specific? Your exemple is quite clear on that. But by lacking rationality in our query, I think this shows we, human, express our rationality on a different level. At a matter of fact, by asking the use of "mathematical newspeak" you're increasing comprehension (by a high magnitude) between human and the machine. Isn't like cheating? I mean shouldn't AI be able to catch us on the intelligence race, therefor be able to understand context?
I understand I have to specified what I mean by "context" and why Semantic web need it. For now I may just ask myself if (logic-query impaired) human can publish on the web documents that isn't full of non-logic assertions that may bug an Agent's report. Because we do use natural language and from a logic point of view it isn't always using univoque meaning (ex.: inclusive OR vs Exclusive OR doesn't exist in French ("ou") but contemporary logic doesn't clearly differentiate it.
Posted by: Martin Lessard on July 20, 2004 11:49 PMSorry for the previous double comment.
I went further in my reply:
Gödel's Theorem has been used to argue that a computer can never be as smart as a human being because the extent of its knowledge is limited by a fixed set of axioms, whereas people can discover unexpected truths ...
More on my (French) post at :
http://zeroseconde.blogspot.com/2004/07/godel-et-le-web-semantique.html