Monday, September 13, 2010

Talking to the Great-Beyond



At the risk of dating myself (in the non-romantic sense), I can remember a world sans-internet, with the first twinkles of information crawling down the not-so-super-highway (message-boards) albeit in a very second hand way.
Now to Google something is a verb in it's own right, and something most of us do regularly. New advances in data acquisition and management are parse, and we expect more and more in return for simple, inexplicable, incomprehensible, inappropriate and and down-right useless queries.
So it is not often that I become amused and engaged with a new way to query the great cloud-in-the-ether. Yet today, I spent (ok, maybe this is more procrastination talking) a fun while 'interacting' with a search engine. Well, no t actually a search engine, but a data analysis and access engine. And I began asking it questions. Questions about itself. I was trying to converse with an algorithm...

'It' is WolframAlpha (www.wolframalpha.com). If you ask Wolfram what it is, it replies that it is 'a computational knowledge engine.' This gives a sense of Wolframs state of mind; adequate and to the point, but not terribly enlightening. Those behind Wolfram (ie, his 'father' Stephan Wolfram) go a little further:

"Wolfram|Alpha's long-term goal is to make all systematic knowledge immediately computable and accessible to everyone. We aim to collect and curate all objective data; implement every known model, method, and algorithm; and make it possible to compute whatever can be computed about anything. Our goal is to build on the achievements of science and other systematizations of knowledge to provide a single source that can be relied on by everyone for definitive answers to factual queries."

So, the ideal is that WolframAlpha, and his siblings (Beta is in gestation) are able to not only provide matches to key phrases, but will summarise, compute, interpret and display data. A suggested example of this is to query two stocks, or companies (eg enter "Apple Microsoft") or countries (eg Australia USA) and you are presented with a summary comparison of the two.
Where this is supposed to become really powerful is for analysis of more integrated data that is freely available online, but little accessed (census data, accounts, stocks, shares, etc) however that is well and truly outside my area of exploration (if someone out there is more insightful than me, let me know!). So instead I settled down to plying the poor thing with demeaning questions (I imagine this is exactly how Marvin first started complaining about the pain in all his right diodes). But in fact WolframAlpha is so designed that asking sensible (or less so) questions is actually intuitive, in that the response is generally to provide some sort of coherent, or at least ball park, answer. Thus my questions grew from tentative "who are you" to the more subtle and tongue in cheek "What’s the speed of an unladen swallow?" and eventually trying to catch him out "is there a god."
Wolfram's ability to answer direct questions varies with the precision and extrapolation of the query, and while some questions are answered in a glib fashion suggesting anticipation by the programmers, other are reduced to component parts and return less than specific results. Still others return the interesting response "Philosophy Ideas/Human Discourse: Development of this topic is under investigation..."
So could we be seeing an attempt to pass the Turing Test??

------------------------------------
Background: the Turing Test was proposed in 1950 by Alan Turing as a way to determine when artificial intelligence has reached that of human cognition. A human user (judge) is blindly engaged in natural language conversation with in turn a human and computer companion. If the user is unable to determine which is human and which is computer, the computer is judged to have passed the test.
Although not an ideal or ultimate test of AI, the test has been backed with a $100,000 prize for the first success which has enhanced it's appeal and credibility, such that it is still the ultimate focus of many AI developments. As yet (2010), it remains unawarded.
------------------------------------

As far as I (in my limited capacity) know, this would not be the first time the idea of using the internet cloud to pass the test has been attempted (something tells me Google may have tried), and it seems the most likely path to success, with the ability to consult, analyse, and model responses upon billions of online examples far superior to any attempt to rely on a library of words and rules pre-determined by programmers.

However, this is not really taking advantage of WolframAlpha's abilities, and obviously many of the answers have been carefully provided for geeks such as myself, rather than being the result of natural language responses from Wolfram himself. But it is, none the less, amusing.
However, it got me thinking about what is it that makes us want to squeeze some sentient-like behaviour out of a non physical being? Why ask stupid questions, or those that I already know or think I know the answer to, of a line of code...?
Hmmmmm.... Anyone for religion...?


Things to ask WolframAlpha:

Are you alive?
When were you born?
Are you my friend?
What is your favourite colour?
What is the weather like?
What’s the speed of an unladen swallow?

What is the meaning of life?
Who is Luke's father?
Who is your daddy and what does he do?
How many chucks can a wood chuck chuck?
How many roads must a man walk down?
Is Elvis alive?
Is there a god?

1 comment:

  1. Meh... Why bother? Don't need no stinkin' Internet. The answer is '42'...

    ReplyDelete