"Computational Knowledge Engine"

[Ex-User (5841)]

I'm debating whether or not this is officially cooler than Wikipedia (oh yeah, I said it). Behold!:


http://www.wolframalpha.com/


If only Wikipedia could do calculus for you (for free at that):

http://www.wolframalpha.com/input/?i=d/dy((3y^7-22x)(e^(x))/x)

Works out differential equations as well.

 

[scorpiomover]

Neato.

 

[PhoenixRising]

This... is... AWESOME I typed in "calculus fundamental theorem", and the way it was presented on this site made it really easy to understand. Makes me want to give another try at higher mathematics.

Definitely better than Wikipedia.

 

[Architect]

Wolfram Alpha has been around for a long time, I think originally it was a paid service. It's interesting technology, however in the past we haven't had good luck with expert systems such as this. Most of the modern technology companies who are using AI, such as Google, are using brute force methods with large data sets. So far that's proving to be a more successful strategy.

 

[Ex-User (5841)]

Wolfram Alpha has been around for a long time, I think originally it was a paid service. It's interesting technology, however in the past we haven't had good luck with expert systems such as this. Most of the modern technology companies who are using AI, such as Google, are using brute force methods with large data sets. So far that's proving to be a more successful strategy.

Troof^

As an example of demonstrable testimony to the above truth, I just googled Wolfram Alpha and read a Wikipedia article about it

 

[Architect]

Troof^

As an example of demonstrable testimony to the above truth, I just googled Wolfram Alpha and read a Wikipedia article about it

One could counterpoint with the fact that Google searches are actually expert systems due to crowd sourced page ranking .... furthermore Wikipedia is an expert system that I find far more useful than Alpha. Fuck, I'm killing my own argument ...

No just kidding - Watson beats the hell out of Alpha, is not an expert system, and is able to understand inference and language subtleties. While they haven't released technical details it's clearly a large data set with NN/tree search thrown at it.

 

[scorpiomover]

Wolfram Alpha has been around for a long time, I think originally it was a paid service. It's interesting technology, however in the past we haven't had good luck with expert systems such as this. Most of the modern technology companies who are using AI, such as Google, are using brute force methods with large data sets. So far that's proving to be a more successful strategy.

Google's approach, is a bit of brute force. It is not unlike the idea of making a memory-intensive program work better, by buying more memory, rather than writing more efficient code. At $30 for 1024 extra Mb of RAM, that's about 1 hour's salary for a coder, and it takes a lot longer than that to code it. If that fails to work, ask a coder to change the algorithm to another existing algorithm and try that.

That approach will eventually bottom out, when we can no longer get the info we need, no matter how much RAM, CPUs and algorithms wirtten by programmers that we throw at it. So when we reach that stage, then we'll need a whole new type of programmer, one that isn't consistent at all with the current set of programming paradigms.

Probably be something that 99.9% of programmers will say cannot work.