Google's systems may now be smarter than its engineers

[pctech]

http://www.theregister.co.uk/2013/11/15/google_thinking_machines/

[Technical Ben]

Sorry. IMO tools is tools. Peoples is people. It's like saying "I'll build this car so fast it will take over the world". A cars fastness does not make it "better" or more powerful than a person. Likewise, a machines "intelligence" is not going to cause a problem any time soon. It's not physically possible for it to "take over" any more than it is for a car to "take over".

But like the car, it can do some things very well. Cars go much faster than we can run. Calculators give results faster than most mathematicians. These computers are just very fast calculators. I've not see reason to consider a single one can analyze and build an understanding to work with and act on information. They are simply very sophisticated sorting machines.

The risk is, when you think a sorting machine can give you the same results a human can. Like decide what to have for dinner! [BzzzktZip] Sylent Green is mostly good for you, our statistics say. Plus Jedwood is great music. [/kuplunk] 

[pctech]

As you can see from the article Ben they are saying that they are not quite sure how it is learning to do certain things.

[Technical Ben]

Yeah, I read that as "we don't understand the statistical data", but by all means they programed the software. It's equivalent to letting it collate the cat pictures based on colour, and going "gosh, it got it right, cat pictures!"
The underlying mathematics might be that cats are mostly a certain colour. It's an "unknown method", but not "intelligence" beyond what they already programed.

Seen too much buzz and hype to believe their claims. That, and knowing the actual limits of the system (maths ). To the very least, any action a person can take can be automated. But I've not seen "learning" automated yet.

[pctech]

Will be interesting to keep an eye on though

[Gary]

We build the machines we code the machines, then complex machines do odd things that were not in the original parameters. Of course as time goes on and machines get faster and more complex, like computers that write their own code to deal with crashes of shut down part of their network if something fails and re route themselves or that need new ways to deal with data etc maybe we will start to see the birth of something, who knows how the spark of AI may work. Well its fun to day dream 

[pctech]

In all honesty I think if true AI appears anywhere it'll probably be in the network security or server management field as that is where it is most needed to counter threats.

[Gary]

In all honesty I think if true AI appears anywhere it'll probably be in the network security or server management field as that is where it is most needed to counter threats.

When you get systems as vast as google little glitches like this do make you wonder, Mitch.

[pctech]

Yes, indeed, Google also had a project on the go called Spanner that allowed it to automatically shift workloads between data centres in the event of a failure.

I think we are a bit off the Eagle Eye/Terminator/I, Robot scenario at the moment but I do sincerely hope some bright spark doesn't decide to shift defence responsibility to Google 

[Simon]

Don't given them ideas!  They probably already have more intelligence on us all than the CIA.   

[pctech]

Yes indeed, and in the 90s people thought MS were the bad ones.