R2's Day: One, singular sensation....

Republibot 2.0

Sorry about that title.  I'm a little off today- very little sleep due to the Big Bad Wolf huffing and puffing all night in the form of a megastorm...

(Hey- why didn't the Wondertwins try that one?  Form of: A Megastorm!)

 

Anyway.  Slap happy.  Tired.  Good news is, my house is made of brick.  Very large bricks.  So large that they actually generate an optical illusion which goes something like this- you look at my house from a distance of 50 feet or so, and it looks like a small cottage that looks like it could hold a bedroom, maybe two.   As you get closer, your brain will shift and see things the way they are- it's a 3000 square foot house.  Because your experience tells you that bricks are a certain size, you underestimate the size of the house until you realize that the bricks in THIS house have about four times the volume of typical bricks.

 

Or, as we like to say, "It's bigger on the inside"

So, I was reading this bit by Bob Cringely (yes, I call him "Bob") on Big Data becoming Artificial Intelligence.  (It's short.  I'll wait.)

Done?  Good.

Anyway, it finally struck me- the Singularity that R1 has been preaching about for-ev-er.  You wonder why we put him in suspended animation?  He was monomaniacal.   (Greek pun!)

But I finally 'got' it.

The way Google makes intelligent systems (and note, I am NOT using scare quotes around 'intelligent'.  (Except just now)) is that it shows the system some correlations, allows the system to code for those correlations, shows it some more, the system modifies its own code so that it can correlate faster and more efficiently- lather, rinse, repeat about a bazillion times.

 

This is why Google Translate is getting better and better- because it allows you to pick the best translation for your situation, it also uses your input to present other users with a better translation.

 

Have you ever used Google Image Search?  At times, it presents some odd choices... but it does that to try and test it's own assumptions.  If you upload an image of a duck as the item to search for, you will get ducks.  Lots of ducks.  You may also get a few dogs that share the coloration of the duck... and some other things that don't make much sense until you look at the form and color of the photo that you uploaded to start your search.

It's not only pushing pixels- it's trying to understand.  (Again, no scare-quotes)

We were trying to do artificial intelligence all wrong.  By trying to create smart systems by using human language rules, we put several roadblocks in the way of AI.  Apparently, we just needed to figure out a way to get a system to make correlations, and provide a self-scoring mechanism for making those relationships clearer and better.

 

There's an Asimov (I believe, but I'm too sleepy to look it up) story where the phone system achieves sentience when it had enough connections to become self-aware.

Oddly enough, it may be Google (the search engine, not the company) that may be intelligent, because it has crafted sufficient connections between search terms.

How do you become smarter?  You seek out knowledge...

... you search for it.

 

 

Tags: