A Brit tries to single-handedly create A.I.

[msbae]

http://www.independent.co.uk/life-s...uld-change-the-internet-for-ever-1678109.html

Wolfram Alpha will not only give a straight answer to questions such as "how high is Mount Everest?", but it will also produce a neat page of related information – all properly sourced – such as geographical location and nearby towns, and other mountains, complete with graphs and charts.


The real innovation, however, is in its ability to work things out "on the fly", according to its British inventor, Dr Stephen Wolfram. If you ask it to compare the height of Mount Everest to the length of the Golden Gate Bridge, it will tell you. Or ask what the weather was like in London on the day John F Kennedy was assassinated, it will cross-check and provide the answer. Ask it about D sharp major, it will play the scale. Type in "10 flips for four heads" and it will guess that you need to know the probability of coin-tossing. If you want to know when the next solar eclipse over Chicago is, or the exact current location of the International Space Station, it can work it out.

So, it's a software equivalent to the local Super Nerd who's brain is filled with esoteric terminology and trivial minutiae? Why doesn't he just hire me to find information for him?

The engine, which will be free to use, works by drawing on the knowledge on the internet, as well as private databases. Dr Wolfram said he expected that about 1,000 people would be needed to keep its databases updated with the latest discoveries and information.


He also added that he would not go down the road of storing information on ordinary people, although he was aware that others might use the technology to do so.

Good thing my profile at MySpace & Facebook is set to private.

Wolfram Alpha has been designed with professionals and academics in mind, so its grasp of popular culture is, at the moment, comparatively poor. The term "50 Cent" caused "absolute horror" in tests, for example, because it confused a discussion on currency with the American rap artist. For this reason alone it is unlikely to provide an immediate threat to Google, which is working on a similar type of search engine, a version of which it launched last week.

"We have a certain amount of popular culture information," Dr Wolfram said. "In some senses popular culture information is much more shallowly computable, so we can find out who's related to who and how tall people are. I fully expect we will have lots of popular culture information. There are linguistic horrors because if you put in books and music a lot of the names clash with other concepts."

The world is better off without knowing anything about that drug-peddling thug.

He added that to help with that Wolfram Alpha would be using Wikipedia's popularity index to decide what users were likely to be interested in. With Google now one of the world's top brands, worth $100bn, Wolfram Alpha has the potential to become one of the biggest names on the planet.

Not if it's using a questionable source like Wikipedia.

Looks like we have a ways to go before Skynet finally arrives...

 

[All Seeing Eye]

But it's not an AI, it's just a catalogue of information.

 

[msbae]

But it's not an AI, it's just a catalogue of information.

A Mad Scientist has to start somewhere...

 

[All Seeing Eye]

But it's not an AI, it's just a catalogue of information.

A Mad Scientist has to start somewhere...

They started with the calculator.

 

[msbae]

The Abacus =/= the beginnings of AI.

 

[Lindley]

Sounds like it's just taking Latent Dirichlet Allocation (topic discovery in a set of documents) to the next logical step.

 

[Garrovick]

Interesting, but not what I'd consider A.I. To me A.I. would not only be able to recite dictionary/encyclopaedia information but also gather information about the user from gathering their day to day activities. But I don't think that "tri-corder" sensor capability exists yet.

 

[Asbo Zaprudder]

Meh, Wolfram reinvents the wheel yet again, and claims it as his own.

 

[Deckerd]

I have to say using the internet as its primary source of information is only going to be impressive if it can get into a lot of protected datasets.

 

[CuttingEdge100]

I think this is a bad idea as it could easily be used to aggregate and collect huge amounts of data on innocent people. Whether Dr. Wolfram intends to do this or not is irrelevant. Others will.

I have read an article on this (possibly this article) in which people suggested it could open an entire world of artificial intelligence, and a self-organising internet.

Is this really necessary or even a good idea? I can think of all kinds of things that can go wrong with Artificial Intelligence...


CuttingEdge100

 

[Meredith]

Sounds like a "BDA" Big Dumb AI, kinda like the computer on the enterprise.

 

[Robert Maxwell]

Yeah, this is not even close to "AI." It's just lexical analysis mixed with data mining and a little computation. Like a self-building expert system, except he says it will take about 1000 people to maintain it. I'd be a lot more impressed if it could do all this stuff in a completely automated fashion and not require anyone to deliberately feed it data.

You could probably find 95% of the data people are looking for just by Googling the actual question. I mean, "How tall is Mount Everest?" Some really simple parsing could pull that from probably tens of thousands of different pages. Unless it will have access to very specific and obscure data, there is basically nothing new here.

Mostly, I'm just unimpressed that he thinks an expert system that requires 1000 people to maintain is "innovative."

 

[msbae]

Yeah, this is not even close to "AI." It's just lexical analysis mixed with data mining and a little computation. Like a self-building expert system, except he says it will take about 1000 people to maintain it. I'd be a lot more impressed if it could do all this stuff in a completely automated fashion and not require anyone to deliberately feed it data.

You could probably find 95% of the data people are looking for just by Googling the actual question. I mean, "How tall is Mount Everest?" Some really simple parsing could pull that from probably tens of thousands of different pages. Unless it will have access to very specific and obscure data, there is basically nothing new here.

Mostly, I'm just unimpressed that he thinks an expert system that requires 1000 people to maintain is "innovative."

It must have been a slow news day.

 

[Robert Maxwell]

That, and Wolfram is a good self-promoter.

 

[CuttingEdge100]

Still, this system obviously has the risk of being used to aggregate data on innocent people. Even though Wolfram won't allegedly do it, others will.

Anyone with half a brain would realize that


CuttingEdge100

 

[FordSVT]

^Google does pretty much everything this claims to.